Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
Application 18/984,259 filed on 12/17/2024 with provisional application 63/611,791
filed on 12/19/2023 has been examined. In this Office Action, Claims 1-20 are currently pending.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is
directed to an abstract idea without significantly more.
Claim 1 recites: (Step 2a, Prong One)
applying multi-label classifiers to the unlabeled access data based on a comparison of
multi-label classifications.
The limitation of applying multi-label classifiers to the unlabeled access data based on a comparison of multi-label classifications, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. That is, other than reciting a generic computer-implemented method, nothing in the claim element precludes the step from practically being performed in the mind. For example, but for the computer-implemented method language, “applying” in the context of this claim encompasses the user manually determining/classifying generic “access data” using generic “comparisons” and “classification” steps. Similarly, the limitation(s) of
performing; grouping; grouping; generating; determining; determining; and comparing, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. For example, but for the computer-implemented method language, performing; grouping; grouping; generating; determining; determining; and comparing in the context of this claim encompasses the user manually receiving generic “access data” and classifications/groupings and performing generic “frequency analysis” and corpora generation steps. If a claim limitation, under its
broadest reasonable interpretation, covers performance of the limitation in the mind but for the
recitation of generic computer components, then it falls within the “Mental Processes” grouping
of abstract ideas (concepts performed in the human mind (including an observation, evaluation,
judgment, opinion)).
Further, these concepts also recite “Certain Methods of Organizing Human Activity”; (such as
commercial or legal interactions (including agreements in the form of contracts; legal
obligations; advertising, marketing or sales activities or behaviors; business relations) where
performing generic applying/classifying steps of generic access data using generic classifications and frequency analysis is a method of human activity in commercial or legal interactions.
Accordingly, the claim recites an abstract idea.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites one additional element – using a computer implemented method to perform both the performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. The computer implemented method in both steps is recited at a high
level of generality (i.e., as a generic processor performing a generic computer function of
“applying”) such that it amounts no more than mere instructions to apply the exception using a
generic computer component. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of a computer implemented method to perform both the performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. The claim(s) is/are not patent eligible.
Referring to claim 2, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein determining the first multi-label classifications based on the access-data corpora comprises: generating n-grams for the unlabeled access data; determining tagging scores for the n-grams, wherein the tagging scores are based on the access-data corpora; and
applying labels to the unlabeled access data with label probabilities based on the tagging scores”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein determining the first multi-label classifications based on the access-data corpora comprises: generating n-grams for the unlabeled access data; determining tagging scores for the n-grams, wherein the tagging scores are based on the access-data corpora; and applying labels to the unlabeled access data with label probabilities based on the tagging scores” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract
idea into a practical application because it does not impose any meaningful limits on practicing
the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein determining the first multi-label classifications based on the access-data corpora comprises: generating n-grams for the unlabeled access data; determining tagging scores for the n-grams, wherein the tagging scores are based on the access-data corpora; and applying labels to the unlabeled access data with label probabilities based on the tagging scores” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 3, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract
idea into a practical application because it does not impose any meaningful limits on practicing
the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 3, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract
idea into a practical application because it does not impose any meaningful limits on practicing
the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 4, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein determining the tagging scores further comprises assigning weights based on the access-data corpora, and wherein the weights are assigned to: the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein determining the tagging scores further comprises assigning weights based on the access-data corpora, and wherein the weights are assigned to: the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract
idea into a practical application because it does not impose any meaningful limits on practicing
the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein determining the tagging scores further comprises assigning weights based on the access-data corpora, and wherein the weights are assigned to: the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 5, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the n-gram dictionaries comprise: dictionaries of positive n-grams;
dictionaries of negative n-grams; and dictionaries of n-grams to be discarded”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the n-gram dictionaries comprise: dictionaries of positive n-grams; dictionaries of negative n-grams; and dictionaries of n-grams to be discarded” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract
idea into a practical application because it does not impose any meaningful limits on practicing
the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the n-gram dictionaries comprise: dictionaries of positive n-grams; dictionaries of negative n-grams; and dictionaries of n-grams to be discarded” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 6, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein identifying the high frequency occurrences in the access data is based on a high frequency threshold”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein identifying the high frequency occurrences in the access data is based on a high frequency threshold” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract
idea into a practical application because it does not impose any meaningful limits on practicing
the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein identifying the high frequency occurrences in the access data is based on a high frequency threshold” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 7, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein identifying the low frequency occurrences in the access data is based on a low frequency threshold”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein identifying the low frequency occurrences in the access data is based on a low frequency threshold” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract
idea into a practical application because it does not impose any meaningful limits on practicing
the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein identifying the low frequency occurrences in the access data is based on a low frequency threshold” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 8, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the similarity algorithm is a cosine similarity algorithm”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the similarity algorithm is a cosine similarity algorithm” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract
idea into a practical application because it does not impose any meaningful limits on practicing
the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the similarity algorithm is a cosine similarity algorithm” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Claim 9 recites: (Step 2a, Prong One)
applying multi-label classifiers to the unlabeled access data based on a comparison of
multi-label classifications.
The limitation of applying multi-label classifiers to the unlabeled access data based on a comparison of multi-label classifications, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. That is, other than reciting a generic “access-data classifier”, nothing in the claim element precludes the step from practically being performed in the mind. For example, but for the “access-data classifier” language, “applying” in the context of this claim encompasses the user manually determining/classifying generic “access data” using generic “comparisons” and “classification” steps. Similarly, the limitation(s) of
performing; grouping; grouping; generating; determining; determining; and comparing, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic or unspecified computer components/”access-data classifier”. For example, but for the “access-data classifier” language, performing; grouping; grouping; generating; determining; determining; and comparing in the context of this claim encompasses the user manually receiving generic “access data” and classifications/groupings and performing generic “frequency analysis” and corpora generation steps. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic unspecified computer components/” access-data classifier”, then it falls within the “Mental Processes” grouping of abstract ideas (concepts performed in the human mind (including an observation, evaluation, judgment, opinion)).
Further, these concepts also recite “Certain Methods of Organizing Human Activity”; (such as
commercial or legal interactions (including agreements in the form of contracts; legal
obligations; advertising, marketing or sales activities or behaviors; business relations) where
performing generic applying/classifying steps of generic access data using generic classifications and frequency analysis is a method of human activity in commercial or legal interactions.
Accordingly, the claim recites an abstract idea.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites one additional element – using a “access-data classifier” to perform both the performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. The “access-data classifier” in both steps is recited at a high level of generality (i.e., as a generic “access-data classifier” performing a generic computer function of “applying”) such that it amounts no more than mere instructions to apply the exception using a generic unspecified computer component/”access-data classifier”. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of a “access-data classifier” to perform both the performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic unspecified computer component/”access-data classifier”. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. The claim(s) is/are not patent eligible.
Referring to claim 10, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the unsupervised natural language processing module is configured to:
generate n-grams for the access data; determine tagging scores for the n-grams, wherein the tagging scores are based on the access-data corpora; and apply labels to the access data with label probabilities based on the tagging scores”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the unsupervised natural language processing module is configured to: generate n-grams for the access data; determine tagging scores for the n-grams, wherein the tagging scores are based on the access-data corpora; and apply labels to the access data with label probabilities based on the tagging scores” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the unsupervised natural language processing module is configured to: generate n-grams for the access data; determine tagging scores for the n-grams, wherein the tagging scores are based on the access-data corpora; and apply labels to the access data with label probabilities based on the tagging scores” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 11, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the access data comprise entitlement data, and wherein the unsupervised natural language processing module is configured to determine the tagging scores based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the access data comprise entitlement data, and wherein the unsupervised natural language processing module is configured to determine the tagging scores based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the access data comprise entitlement data, and wherein the unsupervised natural language processing module is configured to determine the tagging scores based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 12, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the unsupervised natural language processing module is configured to assigning weights based on the access-data corpora, and wherein the weights are assigned to:
the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the unsupervised natural language processing module is configured to assigning weights based on the access-data corpora, and wherein the weights are assigned to: the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the unsupervised natural language processing module is configured to assigning weights based on the access-data corpora, and wherein the weights are assigned to: the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 13, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the access-data corpora comprise: dictionaries of positive n-grams;
dictionaries of negative n-grams; and dictionaries of n-grams to be discarded”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the access-data corpora comprise: dictionaries of positive n-grams; dictionaries of negative n-grams; and dictionaries of n-grams to be discarded” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the access-data corpora comprise: dictionaries of positive n-grams; dictionaries of negative n-grams; and dictionaries of n-grams to be discarded” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 14, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the high frequency occurrences in the access data are based on a high frequency threshold”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the high frequency occurrences in the access data are based on a high frequency threshold” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the high frequency occurrences in the access data are based on a high frequency threshold” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 15, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the low frequency occurrences in the access data are based on a low frequency threshold”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the low frequency occurrences in the access data are based on a low frequency threshold” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the low frequency
occurrences in the access data are based on a low frequency threshold” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 16, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the similarity algorithm is a cosine similarity algorithm”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the similarity algorithm is a cosine similarity algorithm” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the similarity algorithm is a cosine similarity algorithm” steps to perform both the aforementioned performing; grouping; grouping; generating; determining; determining; and comparing; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Claim 17 recites: (Step 2a, Prong One)
applying multi-label classifiers to the unlabeled access data based on multi-label classifications.
The limitation of applying multi-label classifiers to the unlabeled access data based on multi-label classifications, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. That is, other than reciting a generic computer-implemented method, nothing in the claim element precludes the step from practically being performed in the mind. For example, but for the computer-implemented method language, “applying” in the context of this claim encompasses the user manually determining/classifying generic “access data” using generic “comparisons” and “classification” steps. Similarly, the limitation(s) of performing; grouping; grouping; generating; determining; determining; and comparing, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. For example, but for the computer-implemented method language, performing; grouping; generating; determining; determining; and applying in the context of this claim encompasses the user manually receiving generic “access data” and classifications/groupings and performing generic “frequency analysis” and corpora generation steps. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas (concepts performed in the human mind (including an observation, evaluation, judgment, opinion)).
Further, these concepts also recite “Certain Methods of Organizing Human Activity”; (such as
commercial or legal interactions (including agreements in the form of contracts; legal
obligations; advertising, marketing or sales activities or behaviors; business relations) where
performing generic applying/classifying steps of generic access data using generic classifications and frequency analysis is a method of human activity in commercial or legal interactions.
Accordingly, the claim recites an abstract idea.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites one additional element – using a computer implemented method to perform both the performing; grouping; generating; determining; determining; and applying steps. The computer implemented method in both steps is recited at a high level of generality (i.e., as a generic processor performing a generic computer function of “applying”) such that it amounts no more than mere instructions to apply the exception using a generic computer component. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of a computer implemented method to perform both the performing; grouping; generating; determining; determining; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. The claim(s) is/are not patent eligible.
Referring to claim 18, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “generating n-grams for the unlabeled access data; determining tagging scores for the n-grams, wherein the tagging scores are based on the access-data corpora; and
applying labels to the unlabeled access data with label probabilities based on the
tagging scores”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “generating n-grams for the unlabeled access data; determining tagging scores for the n-grams, wherein the tagging scores are based on the access-data corpora; and applying labels to the unlabeled access data with label probabilities based on the tagging scores” steps to perform both the aforementioned performing; grouping; generating; determining; determining; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “generating n-grams for the unlabeled access data; determining tagging scores for the n-grams, wherein the tagging scores are based on the access-data corpora; and applying labels to the unlabeled access data with label probabilities based on the tagging scores” steps to perform both the aforementioned performing; grouping; generating; determining; determining; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 19, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and
stopword frequencies in the entitlement data” steps to perform both the aforementioned performing; grouping; generating; determining; determining; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on:
entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data” steps to perform both the aforementioned performing; grouping; generating; determining; determining; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Referring to claim 20, (Step 2a, Prong One) this further merely performs an additional abstract
mental step of “wherein determining the tagging scores further comprises assigning weights based on the access-data corpora, and wherein the weights are assigned to: the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies.”.
(Step 2a, Prong Two)
This judicial exception is not integrated into a practical application. In particular, the claim only
recites the additional elements of “wherein determining the tagging scores further comprises assigning weights based on the access-data corpora, and wherein the weights are assigned to: the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies” steps to perform both the aforementioned performing; grouping; generating; determining; determining; and applying steps. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
(Step 2b)
The claim does not include additional elements that are sufficient to amount to significantly
more than the judicial exception. As discussed above with respect to integration of the abstract
idea into a practical application, the additional element of using “wherein determining the tagging scores further comprises assigning weights based on the access-data corpora, and wherein the weights are assigned to: the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies” steps to perform both the aforementioned performing; grouping; generating; determining; determining; and applying steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
The claim(s) is/are not patent eligible.
Additionally, Claims 9-16 are rejected under 35 U.S.C. 101 because the claimed invention is
Directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the
four categories of patent eligible subject matter because independent claim(s) 9 does not recite
statutory computer hardware/processors (only generic “access-data classification system” without even generically described computer processing elements, for example a “computer processor”) without limitation and thus the claim(s) is/are directed to a signal per se and/or mere information in the form of data, and dependent claims 10-16 do not correct this deficiency.
The specification paragraph The claim(s) does/do not fall within at least one of the 0023 suggest a software only implementation.
See generally guidance on the New Form Paragraphs for Subject Matter Eligibility Rejections
under the 2019 Revised Patent Subject Matter Eligibility Guidance (¶ 7.05.01 Rejection, 35
U.S.C. 101, Nonstatutory (Not One of the Four Statutory Categories);
Available via:
https://www.uspto.gov/sites/default/files/documents/form_para_for_2019peg_20190108.pdf
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Roy et al., US Pub. No. 2022/0171925 A1, in view of Doyle et al., US Pub. No. 2020/0019561 A1, in view of Wright et al., US Pub. No. 2023/0351194 A1, in view of Wu et al., US Pub. No. 2023/0102892 A1.
As to claim 1, Roy discloses
a computer-implemented method,
(Roy [0005, 0061-0062])
comprising:
performing a frequency analysis on unlabeled access data to identify high frequency
occurrences and low frequency occurrences in the unlabeled access data;
(Roy teaches frequency classes and determining lower/higher word frequencies i.e. to identify high frequency occurrences and low frequency occurrences in the unlabeled access data” see [0029] At step 206 of the process 200, the system 100 determines word frequency of each of the words in V and N. Word frequency of a word indicates/represents frequency of occurrence of the word in the input data. The system 100 may use any known word counting techniques for counting the number of occurrences of each of the words, in the input data, and thereby determine the word frequency. Based on the determined word frequency, the system 100 may assign each of the plurality of words to corresponding frequency class. For a word w, corresponding class is denoted as w,S, where S takes value V or N depending on parent set of w, and i denotes a class number.;
See also [0038-0044] [0038] If w1 has lower word frequency than w,, then w, may be a domain word. …[0044] If word frequency of the word w1 is higher than the word frequency of w,,)
grouping the high frequency occurrences in a first grouping;
(Roy teaches different word frequency classes, i.e. “high frequency occurrences in a first grouping/ low frequency occurrences in a second grouping” [0029] . The system 100 may use any known word counting techniques for counting the number of occurrences of each of the words, in the input data, and thereby determine the word frequency. Based on the determined word frequency, the system 100 may assign each of the plurality of words to corresponding frequency class. For a word w, corresponding class is denoted as w,S, where S takes value V or N depending on parent set of w, and i denotes a class number))
grouping the low frequency occurrences in a second grouping;
(Roy teaches different word frequency classes, i.e. “high frequency occurrences in a first grouping/ low frequency occurrences in a second grouping” see [0029])
generating access-data corpora based on the first grouping, the second grouping,
(Roy teaches different categories based on word frequencies, i.e. “generating access-data corpora based on the first grouping, the second grouping” see [0007] The plurality of words are classified to a set of dictionary words (V) and a set of nondictionary words (N), via the one or more hardware processors. Further, word frequency of each word in V and N is
determined via the one or more hardware processors, wherein the word frequency of a word represents frequency of occurrence of the word in the input data. The nondictionary words in N are then segregated as domain terms and noise terms, based on the determined word frequency, via the one or more hardware processors. Further, the plurality of words in V and N are classified into a plurality of pre-defined categories, via the one or more hardware
processors, wherein the plurality of pre-defined categories comprises a plurality of error categories and a plurality of non-error categories; See also [0029])
Roy does not disclose:
and
an initial dictionary using a similarity algorithm, wherein the initial dictionary is manually created
from the unlabeled access data, and wherein the access-data corpora comprise n-gram
dictionaries;
However, Doyle discloses:
and
an initial dictionary using a similarity algorithm, wherein the initial dictionary is manually created
from the unlabeled access data,
(Doyle teaches manual labelling and supervised machine learning dictionary creation using cosine similarity, i.e. “an initial dictionary using a similarity algorithm, wherein the initial dictionary is manually created from the unlabeled access data,”
See [0149] The labeled data 300B may be generated by human reviewers or one or more generations of the one or more data models. For example, domain experts may manually identify and label some terms of interest from a set of digital contents of a software model to form labeled data (e.g., a partially labeled set of digital contents) in some embodiments;
See also [0232] A dictionary or a data structure including unique tokens may be optionally generated at SOSA. The unique tokens in this dictionary or data structure will be sent to a
word embedding or term embedding module that transform these unique tokens into corresponding vector representations. Prior to actually transforming these unique tokens, the
word embedding or term embedding module or the artificial intelligence modules therein may be trained with one or more training instances at 510A.;
see also [0242] These analogical reasoning tasks iteratively calibrate the word embedding or term embedding modules under training in a supervised, unsupervised, or reinforcement learning environment until the desired accuracy is achieved.;)
and wherein the access-data corpora comprise n-gram dictionaries;
(Doyle [0145] The plurality of tokens may further include n-grams (e.g., unigrams, bigrams, trigrams, four-grams, five-grams, etc.) extracted from the data of interest in some embodiments.)
It would have been obvious to one having ordinary skill in the art at the time of the effective filing date to apply dictionary generation as taught by Doyle to the system of Roy since it was known in the art that machine learning text processing systems provide a word embedding or term embedding model (a software model) where the word embedding or term embedding model for a dictionary or data structure, which may then be pruned where tokens having a predetermined number of times of occurrences in the dictionary, tokens that are equivalent to each other, etc. may be pruned from the dictionary or data structure where the computational resource requirement (e.g., memory footprint, processor cycles, etc.) is roughly proportional to the product of the number of tokens to be vectorized and the degrees of freedom where pruning the dictionary at may thus further conserve computational resources. (Doyle [0239]).
Roy/Doyle do not disclose:
determining, by an unsupervised natural language processing module, first multi-label
classifications for the unlabeled access data, wherein the first multi-label classifications are
based on the access-data corpora;
determining, by a supervised neural network module, second multi-label classifications
for the unlabeled access data, wherein the second multi-label classifications are based on the
access-data corpora and a manually-labeled subset of the access data;
however, Wright discloses:
determining, by an unsupervised natural language processing module, first multi-label
classifications for the unlabeled access data, wherein the first multi-label classifications are
based on the access-data corpora;
(Wright teaches an unsupervised learning neural network generate groupings of the input data and multiple labels see para. [0081] For instance and in some embodiments of the AI
program 502, the program may be configured to perform unsupervised learning, in which the machine learning program performs the training process using unlabeled data, e.g., without known output data with which to compare. During such unsupervised learning, the neural network may be configured to generate groupings of the input data and/or determine how individual input data points are related to the complete input data set ( e.g., via the front-end program 504)
see also [0122] Many other types of scored labels may also be assigned, and multiple scored labels may be assigned to individual clients, where each separate score identifies a
value of the client in a particular category of business parameter)
determining, by a supervised neural network module, second multi-label classifications
for the unlabeled access data, wherein the second multi-label classifications are based on the
access-data corpora and a manually-labeled subset of the access data;
(Wright also teaches supervised/human identifications see [0134] When the human analyst 810 identifies groupings and associates them to clusters in the output data 806, the group type/label information is added to the output data 806 and stored in a database 814. The database 814 may be used for additional analysis by others, and may be used for the updated training (in supervised learning mode) of the machine learning system 804.; see also [0071] Neural networks may perform a supervised learning process where known inputs and known outputs are utilized to categorize, classify, or predict a quality of a future input.)
It would have been obvious to one having ordinary skill in the art at the time of the effective filing date to apply supervised/unsupervised label generation as taught by Wright to the system of Roy/Doyle since it was known in the art that machine learning text processing systems provide a machine learning system which includes a clustering algorithm configured to recognize clusters (purchase transaction patterns, social affiliations, transactional relationships to
other groups of clients, etc.) in multi-variate data-initially based on unsupervised learning using the only the transaction data, and later based on supervised learning which improves the clustering algorithm's ability to identify specific commonalities, relationships and life events. (Wright [0187]).
Roy/Doyle/Wright do not disclose:
comparing the first multi-label classifications and the second multi-label classifications;
and
applying multi-label classifiers to the unlabeled access data based on the comparison of
the first multi-label classifications and the second multi-label classifications;
however, Wu discloses:
comparing the first multi-label classifications and the second multi-label classifications;
(Wu teaches applying/refining classifiers based on multiple classification probabilities and classify the text data based on the probabilistic labels, i.e. “comparing the first multi-label classifications and the second multi-label classifications”
See [0024] Inputting the text data along with the probabilistic labels allows the transformer-based machine learning algorithm to refine itself ( e.g., train itself) to classify the text data based on the probabilistic labels. The training process generates (refined/trained) classifiers that
may be used to classify text data input into the algorithm into one of the six text data categories;
See also [0025] The trained classifiers may then be implemented by the transformer-based machine learning algorithm (or another transformer-based machine learning algorithm) in a
classification process on text data from any source (such as chat messages, email messages, text messages, social media applications, etc.; see also Fig. 6 item 606: “Determine a set of probabilistic labels for the at least one text data category by applying the generated labelling functions to the unlabeled text data using an unsupervised machine learning algorithm”)
and
applying multi-label classifiers to the unlabeled access data based on the comparison of
the first multi-label classifications and the second multi-label classifications;
(Wu teaches using label probabilities for rules/conditions for fine tuning classifiers, i.e. “applying multi-label classifiers to the unlabeled access data based on the comparison of the first multi-label classifications and the second multi-label classifications” See [0047] For example, in one contemplated embodiment, masking of a portion of sequential text data may be determined based on a combination of text data category label probabilities for two or more categories. As further examples, a hierarchical algorithm, a conditional algorithm, or a rule-based algorithm may be applied to the text data category label probabilities to determine whether to mask a portion of sequential text data;
See also [0038-0039] [0038] In certain embodiments, transformer-based machine learning algorithm module 410 receives unlabeled text data with probabilistic labels from probabilistic label generation module 120 and tunes (e.g., refines) itself to generate one or more trained classifiers for classifying unlabeled text data… [0039] In various embodiments, transformer-based machine learning algorithm module 410 may implement various steps of encoding, embedding, and applying functions to fine tune ( e.g., "train") itself and refine its classifier (s) to provide accurate predictions of categories for the unlabeled text data with probabilistic labels that ; see also Fig. 6 item 608: “Generate one or more classifiers for a transformer-based machine learning algorithm by applying the set of probabilistic labels to the unlabeled text data in the transformer-based machine learning algorithm where the one or more classifiers are generated to classify text data into the at least one text data category”).
It would have been obvious to one having ordinary skill in the art at the time of the effective filing date to apply classification/label based refinement as taught by Wu to the system of Roy/Doyle/Wright since it was known in the art that machine learning text processing systems provide a refined trained classification algorithm for real-time applications such as text messaging or chat messaging where the trained transformer-based machine learning algorithm in real-time applications may reduce the need for manual oversight in such applications
and provide increased accuracy over manual oversight and the trained transformer-based machine learning algorithm may reduce or eliminate the need for manual review of large sets of data to identify privacy-sensitive data or security-sensitive data for masking or redaction where reducing or eliminating the need for manual review may increase the efficiency of identifying and masking or redacting privacy-sensitive data or security-sensitive data in large sets of
data or real-time interactions (Wu [0026]).
As to claim 2, Doyle as modified discloses the computer-implemented method of Claim 1, wherein determining the first multi-label classifications based on the access-data corpora comprises:
generating n-grams for the unlabeled access data;
(Doyle [0145] The plurality of tokens may further include n-grams (e.g., unigrams, bigrams, trigrams, four-grams, five-grams, etc.) extracted from the data of interest in some
embodiments.; see also [0252] This process flow illustrated in FIG. SC may be applied to tokens more than just unigrams. For example, a similar approach may adopt multi-grams or n-grams (e.g., phrases, sentences, etc.))
determining tagging scores for the n-grams, wherein the tagging scores are based on
the access-data corpora; and
(Doyle teaches confidence/classification scores, i.e. “tagging scores” see [0194] These one or more data models may then be executed to identify and classify terms from the set of digital contents. The identification and classification of a term involves the computation of classification confidence levels or scores of the accommodated classification measures or metrics.)
And Wu as modified discloses:
applying labels to the unlabeled access data with label probabilities based on the
tagging scores
(Wu teaches probability scores for labelling i.e. “applying labels to the unlabeled access data with label probabilities based on the tagging scores” [0024] Inputting the text data along with the probabilistic labels allows the transformer-based machine learning algorithm to refine itself ( e.g., train itself) to classify the text data based on the probabilistic labels. The training process generates (refined/trained) classifiers that may be used to classify text data input into the algorithm into one of the six text data categories.).
As to claim 3, Doyle as modified discloses the computer-implemented method of Claim 2, wherein the unlabeled access data comprise entitlement data, and wherein determining the tagging scores is based on: entitlement description frequencies in the entitlement data; entitlement name frequencies in the entitlement data; and stopword frequencies in the entitlement data
(Doyle teaches using term frequencies/ name/stop word frequencies see [0186] In these embodiments illustrated in FIG. 4B, a plurality of classification measures or metrics may be identified at 402B. These classification measures or metrics may include, for example, the frequency of a word
appearing within N words of another word, the frequency of appearance or familiarity of a term in the corpus generated from the set of digital contents or from another corpus;
see also [0188] This contextual measure or metric may thus be tied to the general familiarity of a term in a specific context. For example, a term with M preceding words or characters
and/or N following words or characters that corresponds to a higher probability or frequency (e.g., a frequency value generate by an n-gram model) may indicate that this specific context in which the term is in may reduce the probability of the term being incorrectly classified
See [0195] the machine learning modules may identify any terms that correspond to a sufficiently high frequency of appearance (e.g., the definite article "the", the indefinite article "a" or "an", certain pronouns such as "he", "she", "I", etc.) and prohibit these terms from being classified; see also [0219] In addition or in the alternative, one or more reduction techniques may be applied to the data set or the normalized data set to further reduce the size. For example, punctuations may be removed. In some embodiments, one or more stop or function words or phrases ( e.g., auxiliary verbs, some pronouns such as which, what, I, you, she, he, we, etc.) and/or one or more lexical words or lexical phrases that have little or ambiguous meaning may be filtered out from subsequent processes such as vectorization and clustering.
Names of named entities (e.g., New York Times as the newspaper) may also be optionally extracted although the subsequent word embedding or term embedding processing may nevertheless learn these names.).
As to claim 4, Doyle as modified discloses the computer-implemented method of Claim 3, wherein determining the tagging scores further comprises assigning weights based on the access-data corpora, and wherein the weights are assigned to: the entitlement description frequencies; the entitlement name frequencies; and the stopword frequencies
(Doyle [0241] Customizable, adjustable weight data structures may be determined at 508B for the word embedding or term embedding module under training. One of the advantages of these techniques described herein is that unlike conventional approaches that assign an equal weight to all the tokens and thus often lead to incorrect or imprecise vectorization and clustering results, these techniques assign unequal weights to certain tokens to achieve more accurate and precise results and to enhance the computers' ability to truly understand the natural language input from users. For example, a word embedding or term embedding module may assign lower weights to tokens that are known to cause incorrect or imprecise clustering results and/or assign higher weights to tokens that are known to cause more correct or precise
clustering results during training. Another advantage of the word embedding or term embedding modules is that, unlike conventional approaches that focus on individual words (unigrams), these modules also provide the learning and hence embedding functionalities for multi-grams ( e.g., phrases, sentences, and even documents) that include morethan just the unigrams.).
As to claim 5, Doyle as modified discloses the computer-implemented method of Claim 1, wherein the n-gram dictionaries comprise:
dictionaries of positive n-grams;
(Doyle [0327] If the determination results at 804B are affirmative, these one or more actions may be identified at 806B from a repository or from an indexed data structure.)
dictionaries of negative n-grams; and
(Doyle [0329] In some embodiments where the determination results at 804B are negative, this class may be flagged for further domain expert review in some embodiments. In some other embodiments, this class leading to no identified actions and/or the other pertinent information (e.g., the original inquiry, the normalized or processed inquiry, the pertinent rules, etc.) may be gathered into a training or calibration data set or a training instance to further calibrate
the data model)
dictionaries of n-grams to be discarded
(Doyle [0219] In some embodiments, one or more stop or function words or phrases ( e.g., auxiliary verbs, some pronouns such as which, what, I, you, she, he, we, etc.) and/or one or more lexical words or lexical phrases that have little or ambiguous meaning may be filtered out from subsequent processes such as vectorization and clustering. Names of named entities (e.g., New York Times as the newspaper) may also be optionally extracted although the
subsequent word embedding or term embedding processing may nevertheless learn these names. These filtered out words or phrases may be determined not to add value or usefulness. Stop or function words and phrases contribute primarily to the grammatical structures of tokens, rather than the meanings or contents thereof.).
As to claim 6, Roy as modified discloses the computer-implemented method of Claim 1, wherein identifying the high frequency occurrences in the access data is based on a high frequency threshold
(Roy teaches frequency classes and determining lower/higher word frequencies based on set word frequencies i.e. “identifying the high frequency occurrences in the access data is based on a high frequency threshold” see [0029] At step 206 of the process 200, the system 100 determines word frequency of each of the words in V and N. Word frequency of a word indicates/represents frequency of occurrence of the word in the input data. The system 100 may use any known word counting techniques for counting the number of occurrences of each of the words, in the input data, and thereby determine the word frequency. Based on the determined word frequency, the system 100 may assign each of the plurality of words to corresponding frequency class. For a word w, corresponding class is denoted as w,S, where S takes value V or N depending on parent set of w, and i denotes a class number.;
See also [0038-0044] [0038] If w1 has lower word frequency than w,, then w, may be a domain word. …[0044] If word frequency of the word w1 is higher than the word frequency of w,,).
As to claim 7, Roy as modified discloses the computer-implemented method of Claim 6, wherein identifying the low frequency occurrences in the access data is based on a low frequency threshold
(Roy teaches frequency classes and determining lower/higher word frequencies based on set word frequencies i.e. “identifying the high frequency occurrences in the access data is based on a low frequency threshold” see [0029] At step 206 of the process 200, the system 100 determines word frequency of each of the words in V and N. Word frequency of a word indicates/represents frequency of occurrence of the word in the input data. The system 100 may use any known word counting techniques for counting the number of occurrences of each of the words, in the input data, and thereby determine the word frequency. Based on the determined word frequency, the system 100 may assign each of the plurality of words to corresponding frequency class. For a word w, corresponding class is denoted as w,S, where S takes value V or N depending on parent set of w, and i denotes a class number.;
See also [0038-0044] [0038] If w1 has lower word frequency than w,, then w, may be a domain word. …[0044] If word frequency of the word w1 is higher than the word frequency of w,,).
As to claim 8, Doyle as modified discloses the computer-implemented method of Claim 1, wherein the similarity algorithm is a cosine similarity algorithm
(Doyle teaches supervised machine learning dictionary creation using cosine similarity, i.e. “an initial dictionary using a similarity algorithm, wherein the initial dictionary is manually created from the unlabeled access data,”
See [0232] A dictionary or a data structure including unique tokens may be optionally generated at SOSA. The unique tokens in this dictionary or data structure will be sent to a
word embedding or term embedding module that transform these unique tokens into corresponding vector representations. Prior to actually transforming these unique tokens, the
word embedding or term embedding module or the artificial intelligence modules therein may be trained with one or more training instances at 510A.;
see also [0242] These analogical reasoning tasks iteratively calibrate the word embedding or term embedding modules under training in a supervised, unsupervised, or reinforcement learning environment until the desired accuracy is achieved.;
see also [0250] The objective function may be constructed so that tokens occurring in similar contexts have similar embeddings (as measured by cosine similarity); and capturing the
multiple degrees of similarity between tokens may be further enhanced by using the aforementioned analogical reasoning tasks.).
As to claim 9, Roy discloses
An access-data classification system,
(Roy [0005, 0061-0062])
comprising:
an access-data classifier,
(Roy [0005])
comprising:
an access-data corpora generator configured to:
receive access data;
(Roy [0005])
perform a frequency analysis on the access data to identify high frequency occurrences and low frequency occurrences in the access data;
(Roy teaches frequency classes and determining lower/higher word frequencies i.e. to identify high frequency occurrences and low frequency occurrences in the unlabeled access data” see [0029] At step 206 of the process 200, the system 100 determines word frequency of each of the words in V and N. Word frequency of a word indicates/represents frequency of occurrence of the word in the input data. The system 100 may use any known word counting techniques for counting the number of occurrences of each of the words, in the input data, and thereby determine the word frequency. Based on the determined word frequency, the system 100 may assign each of the plurality of words to corresponding frequency class. For a word w, corresponding class is denoted as w,S, where S takes value V or N depending on parent set of w, and i denotes a class number.;
See also [0038-0044] [0038] If w1 has lower word frequency than w,, then w, may be a domain word. …[0044] If word frequency of the word w1 is higher than the word frequency of w,,)
group the high frequency occurrences in a first grouping;
(Roy teaches different word frequency classes, i.e. “high frequency occurrences in a first grouping/ low frequency occurrences in a second grouping” [0029] . The system 100 may use any known word counting techniques for counting the number of occurrences of each of the words, in the input data, and thereby determine the word frequency. Based on the determined word frequency, the system 100 may assign each of the plurality of words to corresponding frequency class. For a word w, corresponding class is denoted as w,S, where S takes value V or N depending on parent set of w, and i denotes a class number))
group the low frequency occurrences in a second grouping;
(Roy teaches different word frequency classes, i.e. “high frequency occurrences in a first grouping/ low frequency occurrences in a second grouping” see [0029])
generate access-data corpora based on the first grouping, the second grouping,
(Roy teaches different categories based on word frequencies, i.e. “generating access-data corpora based on the first grouping, the second grouping” see [0007] The plurality of words are classified to a set of dictionary words (V) and a set of nondictionary words (N), via the one or more hardware processors. Further, word frequency of each word in V and N is
determined via the one or more hardware processors, wherein the word frequency of a word represents frequency of occurrence of the word in the input data. The nondictionary words in N are then segregated as domain terms and noise terms, based on the determined word frequency, via the one or more hardware processors. Further, the plurality of words in V and N are classified into a plurality of pre-defined categories, via the one or more hardware
processors, wherein the plurality of pre-defined categories comprises a plurality of error categories and a plurality of non-error categories; See also [0029])
Roy does not disclose:
receive an initial dictionary based on the access data, wherein the initial dictionary is manually created;
and the initial dictionary using a similarity algorithm, wherein the access-data corpora comprise n-gram dictionaries;
However, Doyle discloses:
receive an initial dictionary based on the access data, wherein the initial dictionary is manually created;
(Doyle teaches manual labelling and supervised machine learning dictionary creation using cosine similarity, i.e. “an initial dictionary using a similarity algorithm, wherein the initial dictionary is manually created from the unlabeled access data,”
See [0149] The labeled data 300B may be generated by human reviewers or one or more generations of the one or more data models. For example, domain experts may manually identify and label some terms of interest from a set of digital contents of a software model to form labeled data (e.g., a partially labeled set of digital contents) in some embodiments;
See also [0232] A dictionary or a data structure including unique tokens may be optionally generated at SOSA. The unique tokens in this dictionary or data structure will be sent to a
word embedding or term embedding module that transform these unique tokens into corresponding vector representations. Prior to actually transforming these unique tokens, the
word embedding or term embedding module or the artificial intelligence modules therein may be trained with one or more training instances at 510A.;
see also [0242] These analogical reasoning tasks iteratively calibrate the word embedding or term embedding modules under training in a supervised, unsupervised, or reinforcement learning environment until the desired accuracy is achieved.;)
and the initial dictionary using a similarity algorithm,
(Doyle teaches manual labelling and supervised machine learning dictionary creation using cosine similarity, i.e. “an initial dictionary using a similarity algorithm, wherein the initial dictionary is manually created from the unlabeled access data,”
See [0149] The labeled data 300B may be generated by human reviewers or one or more generations of the one or more data models. For example, domain experts may manually identify and label some terms of interest from a set of digital contents of a software model to form labeled data (e.g., a partially labeled set of digital contents) in some embodiments;
See also [0232] A dictionary or a data structure including unique tokens may be optionally generated at SOSA. The unique tokens in this dictionary or data structure will be sent to a
word embedding or term embedding module that transform these unique tokens into corresponding vector representations. Prior to actually transforming these unique tokens, the
word embedding or term embedding module or the artificial intelligence modules therein may be trained with one or more training instances at 510A.;
see also [0242] These analogical reasoning tasks iteratively calibrate the word embedding or term embedding modules under training in a supervised, unsupervised, or reinforcement learning environment until the desired accuracy is achieved.;)
wherein the access-data corpora comprise n-gram dictionaries;
(Doyle [0145] The plurality of tokens may further include n-grams (e.g., unigrams, bigrams, trigrams, four-grams, five-grams, etc.) extracted from the data of interest in some embodiments.)
It would have been obvious to one having ordinary skill in the art at the time of the effective filing date to apply dictionary generation as taught by Doyle to the system of Roy since it was known in the art that machine learning text processing systems provide a word embedding or term embedding model (a software model) where the word embedding or term embedding model for a dictionary or data structure, which may then be pruned where tokens having a predetermined number of times of occurrences in the dictionary, tokens that are equivalent to each other, etc. may be pruned from the dictionary or data structure where the computational resource requirement (e.g., memory footprint, processor cycles, etc.) is roughly proportional to the product of the number of tokens to be vectorized and the degrees of freedom where pruning the dictionary at may thus further conserve computational resources. (Doyle [0239]).
Roy/Doyle does not disclose:
an unsupervised natural language processing module configured to:
receive the access data;
receive the access-data corpora from the access-data corpora generator;
and determine first multi-label classifications for the access data, wherein the first multi-label classifications are based on the access-data corpora;
a supervised neural network trained by a manually-labeled subset of the access
data and the access-data corpora, wherein the supervised neural network is configured to
determine second multi-label classifications for the access data independent of the first multilabel classifications;
However, Wright discloses:
an unsupervised natural language processing module configured to: receive the access data;
(Wright teaches an unsupervised learning neural network generate groupings of the input data
and multiple labels
see para. [0081] For instance and in some embodiments of the AI program 502, the program may be configured to perform unsupervised learning, in which the machine learning program performs the training process using unlabeled data, e.g., without known output data with which to compare. During such unsupervised learning, the neural network may be configured to generate groupings of the input data and/or determine how individual input data points are related to the complete input data set ( e.g., via the front-end program 504)
see also [0122] Many other types of scored labels may also be assigned, and multiple scored
labels may be assigned to individual clients, where each separate score identifies a
value of the client in a particular category of business parameter;
see also [0071] Thus, the parameters determined from the training process can be utilized with
new input data to categorize, classify, and/or predict other values based on the new input data.)
receive the access-data corpora from the access-data corpora generator;
(Wright [0070] The input layer 262, having nodes commonly referenced in FIG. 2A as input nodes 272 for convenience, communicates input data, variables, matrices, or
the like to the hidden layer 264, having nodes 274. The hidden layer 264 generates a representation and/or transformation of the input data into a form that is suitable for generating output data.)
and determine first multi-label classifications for the access data, wherein the first multi-label classifications are based on the access-data corpora;
(Wright teaches an unsupervised learning neural network generate groupings of the input data
and multiple labels see para. [0081] For instance and in some embodiments of the AI
program 502, the program may be configured to perform unsupervised learning, in which the
machine learning program performs the training process using unlabeled data, e.g., without
known output data with which to compare. During such unsupervised learning, the neural
network may be configured to generate groupings of the input data and/or determine how
individual input data points are related to the complete input data set ( e.g., via the front-end
program 504)
see also [0122] Many other types of scored labels may also be assigned, and multiple scored
labels may be assigned to individual clients, where each separate score identifies a
value of the client in a particular category of business parameter)
a supervised neural network trained by a manually-labeled subset of the access
data and the access-data corpora, wherein the supervised neural network is configured to
determine second multi-label classifications for the access data independent of the first multilabel classifications;
(Wright also teaches supervised/human identifications see [0134] When the human analyst 810
identifies groupings and associates them to clusters in the output data 806, the group type/label
information is added to the output data 806 and stored in a database 814. The database 814
may be used for additional analysis by others, and may be used for the updated training (in
supervised learning mode) of the machine learning system 804.; see also [0071] Neural
networks may perform a supervised learning process where known inputs and known outputs
are utilized to categorize, classify, or predict a quality of a future input.)
It would have been obvious to one having ordinary skill in the art at the time of the effective filing
date to apply supervised/unsupervised label generation as taught by Wright to the system of
Roy/Doyle since it was known in the art that machine learning text processing systems provide
a machine learning system which includes a clustering algorithm configured to recognize
clusters (purchase transaction patterns, social affiliations, transactional relationships to
other groups of clients, etc.) in multi-variate data-initially based on unsupervised learning using
the only the transaction data, and later based on supervised learning which improves the
clustering algorithm's ability to identify specific commonalities, relationships and life events.
(Wright [0187]).
Roy/Doyle/Wright do not disclose:
and
wherein the access-data classification system is configured to apply multi-label
classifiers to the access data based on a comparison of the first multi-label classifications and
the second multi-label classifications;
However, Wu discloses:
and
wherein the access-data classification system is configured to apply multi-label
classifiers to the access data based on a comparison of the first multi-label classifications and
the second multi-label classifications.
(Wu teaches using label probabilities for rules/conditions for fine tuning classifiers, i.e. “apply multi-label classifiers to the access data based on a comparison of the first multi-label classifications and the second multi-label classifications” See [0047] For example, in one contemplated embodiment, masking of a portion of sequential text data may be determined based on a combination of text data category label probabilities for two or more categories. As further examples, a hierarchical algorithm, a conditional algorithm, or a rule-based algorithm may be applied to the text data category label probabilities to determine whether to mask a portion of sequential text data;
See also [0038-0039] [0038] In certain embodiments, transformer-based machine learning algorithm module 410 receives unlabeled text data with probabilistic labels from probabilistic label generation module 120 and tunes (e.g., refines) itself to generate one or more trained classifiers for classifying unlabeled text data… [0039] In various embodiments, transformer-based machine learning algorithm module 410 may implement various steps of encoding, embedding, and applying functions to fine tune ( e.g., "train") itself and refine its classifier (s) to provide accurate predictions of categories for the unlabeled text data with probabilistic labels that ; see also Fig. 6 item 608: “Generate one or more classifiers for a transformer-based machine learning algorithm by applying the set of probabilistic labels to the unlabeled text data in the transformer-based machine learning algorithm where the one or more classifiers are generated to classify text data into the at least one text data category”).
It would have been obvious to one having ordinary skill in the art at the time of the effective filing date to apply classification/label based refinement as taught by Wu to the system of Roy/Doyle/Wright since it was known in the art that machine learning text processing systems provide a refined trained classification algorithm for real-time applications such as text messaging or chat messaging where the trained transformer-based machine learning algorithm in real-time applications may reduce the need for manual oversight in such applications
and provide increased accuracy over manual oversight and the trained transformer-based machine learning algorithm may reduce or eliminate the need for manual review of large sets of data to identify privacy-sensitive data or security-sensitive data for masking or redaction where reducing or eliminating the need for manual review may increase the efficiency of identifying and masking or redacting privacy-sensitive data or security-sensitive data in large sets of
data or real-time interactions (Wu [0026]).
Referring to claim 10, this dependent claim recites similar limitations as claim 2;
therefore, the arguments above regarding claim 2 are also applicable to claim 10.
Referring to claim 11, this dependent claim recites similar limitations as claim 3;
therefore, the arguments above regarding claim 3 are also applicable to claim 11.
Referring to claim 12, this dependent claim recites similar limitations as claim 4;
therefore, the arguments above regarding claim 4 are also applicable to claim 12.
Referring to claim 13, this dependent claim recites similar limitations as claim 5;
therefore, the arguments above regarding claim 5 are also applicable to claim 13.
Referring to claim 14, this dependent claim recites similar limitations as claim 6;
therefore, the arguments above regarding claim 6 are also applicable to claim 14.
Referring to claim 15, this dependent claim recites similar limitations as claim 7;
therefore, the arguments above regarding claim 7 are also applicable to claim 15.
Referring to claim 16, this dependent claim recites similar limitations as claim 8;
therefore, the arguments above regarding claim 8 are also applicable to claim 16.
As to claim 17, Roy discloses:
a computer-implemented method, comprising:
(Roy [0005, 0061-0062])
performing a frequency analysis on unlabeled access data to identify high frequency
occurrences and low frequency occurrences in the unlabeled access data;
(Roy teaches frequency classes and determining lower/higher word frequencies i.e. to identify
high frequency occurrences and low frequency occurrences in the unlabeled access data” see
[0029] At step 206 of the process 200, the system 100 determines word frequency of each of
the words in V and N. Word frequency of a word indicates/represents frequency of occurrence
of the word in the input data. The system 100 may use any known word counting techniques for
counting the number of occurrences of each of the words, in the input data, and thereby
determine the word frequency. Based on the determined word frequency, the system 100 may
assign each of the plurality of words to corresponding frequency class. For a word w,
corresponding class is denoted as w,S, where S takes value V or N depending on parent set of
w, and i denotes a class number.;
See also [0038-0044] [0038] If w1 has lower word frequency than w,, then w, may be a domain
word. …[0044] If word frequency of the word w1 is higher than the word frequency of w,,)
grouping the occurrences into a plurality of groupings based on the frequency analysis;
(Roy teaches different word frequency classes, i.e. “high frequency occurrences in a first
grouping/ low frequency occurrences in a second grouping” see [0029])
generating access-data corpora based on the plurality of groupings,
(Roy teaches different categories based on word frequencies, i.e. “generating access-data
corpora based on the first grouping, the second grouping” see [0007] The plurality of words are
classified to a set of dictionary words (V) and a set of nondictionary words (N), via the one or
more hardware processors. Further, word frequency of each word in V and N is
determined via the one or more hardware processors, wherein the word frequency of a word
represents frequency of occurrence of the word in the input data. The nondictionary words in N
are then segregated as domain terms and noise terms, based on the determined word
frequency, via the one or more hardware processors. Further, the plurality of words in V and N
are classified into a plurality of pre-defined categories, via the one or more hardware
processors, wherein the plurality of pre-defined categories comprises a plurality of error
categories and a plurality of non-error categories; See also [0029])
Roy does not disclose:
and an initial dictionary using a similarity algorithm, wherein the initial dictionary is manually created from the unlabeled access data, and wherein the access-data corpora comprise n-gram dictionaries;
However, Doyle discloses:
and an initial dictionary using a similarity algorithm, wherein the initial dictionary is manually created from the unlabeled access data,
(Doyle teaches manual labelling and supervised machine learning dictionary creation using
cosine similarity, i.e. “an initial dictionary using a similarity algorithm, wherein the initial
dictionary is manually created from the unlabeled access data,”
See [0149] The labeled data 300B may be generated by human reviewers or one or more
generations of the one or more data models. For example, domain experts may manually
identify and label some terms of interest from a set of digital contents of a software model to
form labeled data (e.g., a partially labeled set of digital contents) in some embodiments;
See also [0232] A dictionary or a data structure including unique tokens may be optionally
generated at SOSA. The unique tokens in this dictionary or data structure will be sent to a
word embedding or term embedding module that transform these unique tokens into
corresponding vector representations. Prior to actually transforming these unique tokens, the
word embedding or term embedding module or the artificial intelligence modules therein may be
trained with one or more training instances at 510A.;
see also [0242] These analogical reasoning tasks iteratively calibrate the word embedding or
term embedding modules under training in a supervised, unsupervised, or reinforcement
learning environment until the desired accuracy is achieved.;)
and wherein the access-data corpora comprise n-gram dictionaries;
(Doyle [0145] The plurality of tokens may further include n-grams (e.g., unigrams, bigrams,
trigrams, four-grams, five-grams, etc.) extracted from the data of interest in some
embodiments.)
It would have been obvious to one having ordinary skill in the art at the time of the effective filing
date to apply dictionary generation as taught by Doyle to the system of Roy since it was known
in the art that machine learning text processing systems provide a word embedding or term
embedding model (a software model) where the word embedding or term embedding model for
a dictionary or data structure, which may then be pruned where tokens having a predetermined
number of times of occurrences in the dictionary, tokens that are equivalent to each other, etc.
may be pruned from the dictionary or data structure where the computational resource
requirement (e.g., memory footprint, processor cycles, etc.) is roughly proportional to the
product of the number of tokens to be vectorized and the degrees of freedom where pruning the
dictionary at may thus further conserve computational resources. (Doyle [0239]).
Roy/Doyle do not disclose:
determining, by an unsupervised natural language processing module, first multi-label
classifications for the unlabeled access data, wherein the first multi-label classifications are
based on the access-data corpora;
determining, by a supervised neural network module, second multi-label classifications
for the unlabeled access data, wherein the second multi-label classifications are based on the
access-data corpora and a manually-labeled subset of the access data; and
however, Wright discloses:
determining, by an unsupervised natural language processing module, first multi-label
classifications for the unlabeled access data, wherein the first multi-label classifications are
based on the access-data corpora;
(Wright teaches an unsupervised learning neural network generate groupings of the input data
and multiple labels see para. [0081] For instance and in some embodiments of the AI
program 502, the program may be configured to perform unsupervised learning, in which the
machine learning program performs the training process using unlabeled data, e.g., without
known output data with which to compare. During such unsupervised learning, the neural
network may be configured to generate groupings of the input data and/or determine how
individual input data points are related to the complete input data set ( e.g., via the front-end
program 504)
see also [0122] Many other types of scored labels may also be assigned, and multiple scored
labels may be assigned to individual clients, where each separate score identifies a
value of the client in a particular category of business parameter)
determining, by a supervised neural network module, second multi-label classifications
for the unlabeled access data, wherein the second multi-label classifications are based on the
access-data corpora and a manually-labeled subset of the access data; and
(Wright also teaches supervised/human identifications see [0134] When the human analyst 810
identifies groupings and associates them to clusters in the output data 806, the group type/label
information is added to the output data 806 and stored in a database 814. The database 814
may be used for additional analysis by others, and may be used for the updated training (in
supervised learning mode) of the machine learning system 804.; see also [0071] Neural
networks may perform a supervised learning process where known inputs and known outputs
are utilized to categorize, classify, or predict a quality of a future input.)
It would have been obvious to one having ordinary skill in the art at the time of the effective filing
date to apply supervised/unsupervised label generation as taught by Wright to the system of
Roy/Doyle since it was known in the art that machine learning text processing systems provide
a machine learning system which includes a clustering algorithm configured to recognize
clusters (purchase transaction patterns, social affiliations, transactional relationships to
other groups of clients, etc.) in multi-variate data-initially based on unsupervised learning using
the only the transaction data, and later based on supervised learning which improves the
clustering algorithm's ability to identify specific commonalities, relationships and life events.
(Wright [0187]).
Roy/Doyle/Wright do not disclose:
applying multi-label classifiers to the unlabeled access data based on the first multi-label
classifications and the second multi-label classifications.
However, Wu discloses:
applying multi-label classifiers to the unlabeled access data based on the first multi-label
classifications and the second multi-label classifications.
(Wu teaches using label probabilities for rules/conditions for fine tuning classifiers, i.e. “applying multi-label classifiers to the unlabeled access data based on the first multi-label
classifications and the second multi-label classifications” See [0047] For example, in one
contemplated embodiment, masking of a portion of sequential text data may be determined
based on a combination of text data category label probabilities for two or more categories. As
further examples, a hierarchical algorithm, a conditional algorithm, or a rule-based algorithm
may be applied to the text data category label probabilities to determine whether to mask a
portion of sequential text data;
See also [0038-0039] [0038] In certain embodiments, transformer-based machine learning
algorithm module 410 receives unlabeled text data with probabilistic labels from probabilistic
label generation module 120 and tunes (e.g., refines) itself to generate one or more trained
classifiers for classifying unlabeled text data… [0039] In various embodiments, transformer based machine learning algorithm module 410 may implement various steps of encoding,
embedding, and applying functions to fine tune ( e.g., "train") itself and refine its classifier (s) to
provide accurate predictions of categories for the unlabeled text data with probabilistic labels
that ; see also Fig. 6 item 608: “Generate one or more classifiers for a transformer-based
machine learning algorithm by applying the set of probabilistic labels to the unlabeled text data
in the transformer-based machine learning algorithm where the one or more classifiers are
generated to classify text data into the at least one text data category”).
It would have been obvious to one having ordinary skill in the art at the time of the effective filing
date to apply classification/label based refinement as taught by Wu to the system of
Roy/Doyle/Wright since it was known in the art that machine learning text processing systems
provide a refined trained classification algorithm for real-time applications such as text
messaging or chat messaging where the trained transformer-based machine learning algorithm
in real-time applications may reduce the need for manual oversight in such applications
and provide increased accuracy over manual oversight and the trained transformer-based
machine learning algorithm may reduce or eliminate the need for manual review of large sets of
data to identify privacy-sensitive data or security-sensitive data for masking or redaction where
reducing or eliminating the need for manual review may increase the efficiency of identifying
and masking or redacting privacy-sensitive data or security-sensitive data in large sets of
data or real-time interactions (Wu [0026]).
Referring to claim 18, this dependent claim recites similar limitations as claim 2;
therefore, the arguments above regarding claim 2 are also applicable to claim 18.
Referring to claim 19, this dependent claim recites similar limitations as claim 3;
therefore, the arguments above regarding claim 3 are also applicable to claim 19.
Referring to claim 20, this dependent claim recites similar limitations as claim 4;
therefore, the arguments above regarding claim 4 are also applicable to claim 20.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Matthews et al., US Pub. No.: US 2019/0065470 A1, teaches control platform that involves a natural language engine with a risk-based corpora, a rules engine with feature vectors from labelled change records, and topic model to generate an expected label for an additional change record based on training data generated from the labelled change records and the risk-based corpora;
Posner et al., US Pub. No.: 2020/0372030 A1, teaches Techniques for improving system performance based on data characteristics are disclosed. A system may receive updates to a first data set at a first frequency. The system selects a first storage configuration, from a plurality of storage configurations, for storing the first data set based on the first frequency, and stores the first data set in accordance with the first storage configuration. The system may further
receive updates to a second data set at a second frequency. The system selects a second storage configuration, from the plurality of storage configurations, for storing the second
data set based on the second frequency, and stores the second data set in accordance with the second storage configuration. The second storage configuration is different than the first storage configuration.
CONTACT INFORMATION
Any inquiry concerning this communication or earlier communications from the examiner should be directed to EVAN S ASPINWALL whose telephone number is (571)270-7723. The examiner can normally be reached Monday-Friday 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Neveen Abel-Jalil can be reached at 571-270-0474. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/Evan Aspinwall/Primary Examiner, Art Unit 2152 2/4/2026