Detailed Action
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 .
Claims 1-17 are pending for examination. Claims 1, 16, and 17 are independent.
Response to Amendment
The office action is responsive to the amendments filed on 08/28/2025. As
directed by the amendments claims 1, 4, 7-8, 11-12, and 15-17 are amended.
Response to Arguments
Applicant's arguments filed 08/28/2025 have been fully considered but they are not fully persuasive.
Applicant arguments regarding 35 U.S.C. § 112:
Examiner response: Applicant’s arguments, see Pages 6-7 of Remarks, filed 08/28/2025, with respect to 1-16 have been fully considered and are persuasive. The 35 U.S.C. § 101 has been withdrawn.
Applicant arguments regarding 35 U.S.C. § 101:
Examiner response: Applicant’s arguments, see Page 6 of Remarks, filed 08/28/2025, with respect to 4, 7, and 9-15 have been fully considered and are persuasive. The 35 U.S.C. § 101 has been withdrawn.
Claim 17 is directed to non-statutory subject matter.
Applicant arguments regarding 35 U.S.C. § 102/103:
Applicant submits that Li does not disclose "wherein: the ensemble [that includes the first classifier in a plurality of classifier] includes a transformer that generates an output corresponding to the embedding, and the output is shared for all the plurality of classifiers in the hierarchy of the ensemble." In particular, although Li discloses in [0084] "[a] transformation mechanism 1108 next maps the set of position-modified embeddings into transformer output vectors," Li only discloses a single classifier. [0079] ("the topic-detecting system 110 can be implemented as a multi-class classifier."). Thus, since the Li system deploys only a single classifier, it is not the case that Li discloses that "the output [of the transformer] is shared for all the plurality of classifiers in the hierarchy of the ensemble," as recited in the claims. Therefore, withdrawal of the Section 102 rejection is requested.
Remaining Prior Art Rejections
Since none of the other references applied in this office action overcomes this deficiency in Li, withdrawal of the remaining prior art rejections is requested.
Examiner response: Examiner respectfully disagrees, Li also discloses multiple classifiers in para 0080 stating “The topic-detecting system 110 can implement the multi-class classifier using any type of neural network, such as a convolutional neural network (CNN), a transformer based neural network, etc., or any combination thereof.” Para 0088 also states “The classification mechanism can perform this task using a softmax layer , a neural network , a support vector machine ( SVM ), etc.”, which disclose a plurality of classifiers. Li does not explicitly disclose the plurality of classifiers in a hierarchy. Hill is incorporated to explicitly disclose a hierarchy of classifiers (See updated 103 rejection below).
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.
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
Regarding Claim 17:
Claim 17 is 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 directed toward " A computer-readable storage medium" without expressly excluding transitory signals. Therefore, it appears that the claim is directed to signals per se, which is not patent eligible subject matter under 35 USC 101.
Claim Rejections - 35 USC § 103
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1-6, 8-10, 12-13, and 15-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li et al. (US 2021/0382944 A1, hereinafter "Li") in view of Hill et al. (US 2020/0193265 A1, hereinafter "Hill").
Regarding Claim 1
Li discloses: A computer-implemented method for determining a knowledge graph ([Para 0105 and Fig 16] disclose a computer implemented method.), comprising the following steps:
determining an embedding for a sequence of tokens of an instance ([Para 0082-0085 and Fig 11] describes transforming input tokens in a web document into a set of input embeddings.);
determining a first classification for the embedding at a first classifier ([Para 0043, 0056, Fig 8, and Fig 11-12] describes determining a condition probability (i.e., first classification) for topics in the web document (i.e., embedding). [Para 0061] describes using a model to compute the conditional probability (i.e., at a first classifier) belonging to an ensemble that includes the first classifier in a plurality of classifiers ([Para 0080, 0088-0089, and Fig 11] Para 0080 states “The topic-detecting system 110 can implement the multi-class classifier using any type of neural network, such as a convolutional neural network (CNN), a transformer based neural network, etc., or any combination thereof.” This describes multiple machine learning models (i.e. a plurality of classifiers).)
determining if the first classification meets a first condition; ([Para 0057-0058, 0103, Fig 8, and Fig 11-15] describes the conditional probability meeting a prescribed threshold value (i.e., meets a first condition).)
adding to the knowledge graph a first link between a first node of the knowledge graph representing the instance and a node of the knowledge graph representing the first classification when the first classification meets the first condition and not adding the first link when the first classification does not meet the first condition ([Para 0057-0060, 0103, Fig 7-9. and Fig 15] describes establishing a link between topic nodes representing the web document (i.e., representing the instance) (e.g., parent node A) and representing when the conditional probability meets the prescribed threshold (i.e., first condition) (e.g., child node B when P(A|B) > 0.5 ). ([Para 0057-0058 and Fig 8] also describes not meeting the threshold which discloses not adding a link.)), wherein:
the ensemble includes a transformer that generates an output corresponding to the embedding ([Para 0080, 0082, 0084, 0088-0089, and Fig 11] describes a transformation mechanism 1108 that maps embeddings into transformer output vectors (i.e. generated output).), and
the output is shared for all the plurality of classifiers ([Para 0080, 0082, 0084, 0088-0089, and Fig 11] describe a classification mechanism 1122 (i.e. classifiers) operates on the output of the transformation unit 1110 (i.e. the output shared).) in
Li does not explicitly disclose: a plurality of classifiers arranged according to a hierarchy;
However, Hill discloses in the same field of endeavor: a first classifier belonging to an ensemble that includes the first classifier in a plurality of classifiers arranged according to a hierarchy ([Para 0013 and Fig 4] “the root dialogue classification node comprises an ensemble of distinct machine learning models, each of the distinct machine learning models being trained to perform a distinct dialogue intent classification tasks, wherein the ensemble produces multiple distinct intent classification predictions including a distinct intent classification prediction for each one of the plurality of distinct dialogue classification nodes downstream of the root dialogue classification node.” Examiner interprets the root or parent node as a first classifier with child nodes representing a plurality of classifiers in a hierarchy (also see Fig 4 and Para 0080-0083).); the plurality of classifiers in the hierarchy of the ensemble. ([Para 0013, 0080-0083, and Fig 4] describes a root node classifier and downstream classifiers (i.e. plurality of classifiers in a hierarchy).)
It would have been obvious to a person of ordinary skill in art before the effective filling date of the invention to implement the function of ensemble downstream classification nodes taught by Hill into the method of generating graph structures disclosed by Li to disclose a plurality of classifiers arranged according to a hierarchy. The modification would have been obvious because one of the ordinary skills of the art would be motivated to utilize the feature of downstream classifiers taught by Hill as all the references are in the field of machine learning. A person of ordinary skill of the art would have been motivated to perform the combination for being able to perform more granular classification (Para 0080, Hill).
Regarding Claim 16
Li in view of Hill discloses: A device configured to determine a knowledge graph, the device a hardware processor configured to ([Para 0105, and Fig 16], Li): (Claim 16 is a device claim that corresponds to claim 1 and the rest of the limitations are rejected on the same ground)
Regarding Claim 17
Li in view of Hill discloses: A computer-readable storage medium on which is stored a computer program including computer readable instructions for determining a knowledge graph, the instructions, when executed by a computer, causing the computer to perform the following steps ([Para 0105-0110, and Fig 16], Li): (Claim 17 is a computer-readable storage medium claim that corresponds to claim 1 and the rest of the limitations are rejected on the same ground)
Regarding Claim 2
Li in view of Hill discloses: The method according to claim 1, further comprising the following steps: determining a second classification at a second classifier ([Para 0054, 0077-0078, 0083-0084], Hill describes distinct state classification nodes is configured to perform a distinct classification task (i.e., second classification) based on conversational data input and each state classification to implement one or more distinctly trained machine learning classifier (i.e., at second classifier).); determining if the second classification meets a second condition ([Para 0053-0061, 0103, and Fig 6-9] Li describes determining a second conditional probability such as Topic C (e.g., P(A|C)). [Para 0057-0058, 0103, Fig 8, and Fig 11-15], Li describes the second conditional probability meeting a prescribed threshold value.); adding to the knowledge graph a second link between the first node of the knowledge graph representing the instance and a second node of the knowledge graph representing the second classification when the second classification meets the second condition and not adding the second link when the second classification does not meet the second condition. ([Para 0057-0060, 0103, Fig 7-9. and Fig 15], Li describes establishing a second link between topic nodes representing the web document (i.e., representing the instance) (e.g., parent node A) and representing when the conditional probability meets the prescribed threshold (i.e., second condition) (e.g., child node C when P(A|C) > 0.5 ). [Para 0057-0058 and Fig 8] describes not meeting the threshold which discloses not adding a link.)
Regarding Claim 3
Li in view of Hill discloses: The method according to claim 2, further comprising: providing the embedding to the first classifier ([Para 0082-0085 and Fig 11], Li describes the web document as input embeddings provided to a classifier.); and providing the embedding and/or a hidden state of the first classifier resulting from the provided embedding as input to the second classifier. ([Para 0014, 0094, claim 1, claim 9], Hill describes a downstream classification from a parent node (first classifier) to a child node (i.e., input to second classifier).)
Regarding Claim 4
Li in view of Hill discloses: The method according to claim 3, further comprising:
determining the first link from the first node representing the instance to the node representing the first classification when the second classification meets the second condition.([Para 0014, 0094, claim 1, claim 9], Hill describes a downstream the root dialogue classification node downstream to a distinct one of the plurality of distinct dialogue classification nodes based on identifying which distinct intent classification prediction as having a highest level of confidence (i.e., a condition).
Regarding Claim 5
Li in view of Hill discloses: The method according to claim 1, further comprising:
determining the sequence of tokens of the instance ([Para 0064 0067 0079-0085, 0100 and Fig 11], Li describes a sequence of tokens of the input web document.).
Regarding Claim 6
Li in view of Hill discloses: The method according to claim 1, wherein the instance includes digital text data. ([Para 0081-0085 and Fig 11], Li describes transforming input tokens in a web document into a set of input embeddings and the web document containing words (i.e., including digital text data).))
Regarding Claim 8
Li in view of Hill discloses: The method according to claim 2, further comprising: providing the knowledge graph with nodes representing a tree of labels ([Para 0042 0057-0059,0065, 0069, 0079 Fig 7-9] Li describes the graph representing hierarchical relationships (i.e., tree) with nodes representing topics (i.e., labels). Also see fig 9.), and adding to the knowledge graph a plurality of links to nodes representing labels for the instance ([Para 0057-0060, 0103, Fig 7-9. and Fig 15] Li describes establishing links between topic nodes (i.e., labels) representing the web document (i.e., representing the instance).).
Regarding Claim 9
Li in view of Hill discloses: The method according to claim 8, further comprising: deciding based on the first classification and/or the second classification whether the instance belongs to a category represented by a node in the tree of labels or not. ([Para 0057-0060, 0103, Fig 7-9. and Fig 15] Li describes establishing links between topic nodes (i.e., labels) representing the web document (i.e., representing the instance) and discloses deciding if the web document belongs to the topic represented by a node in the hierarchy.)
Regarding Claim 10
Li in view of Hill discloses: The method according to claim 8, further comprising: assigning the first classification to a different label than the second classification ([Para 0053-0060, 0103, Fig 6-9. and Fig 15] Li discloses different topics (i.e., Topic B is different than Topic C).
Regarding Claim 12
Li in view of Hill discloses: The method according to claim 2, further comprising: providing a model ([Para 0061], Li describes using a model to compute the conditional probability.); and training the model to determine the first classification or the second classification ([Para 0039, 0050, 0091] Li discloses training a model to perform classification.).
Regarding Claim 13
Li in view of Hill discloses: The method according to claim 12, wherein the model is a neural network ([Para 0061, and Para 0080-0087], Li discloses a neural network.).
Regarding Claim 15
Li in view of Hill discloses: The method according to claim 12, further comprising:
determining a classification for an input to the trained model with a classifier at a position in a hierarchy of classifier in the model ([0057-0059, 0076, 0101, and Fig 7-9] describes the knowledge graph having a hierarchical relationship with parent nodes and child nodes. Examiner interprets parent nodes has having higher positions to child nodes (see Fig 9).); and
assigning the classification to a label that corresponds to the position in the hierarchy of a tree of labels ([Para 0057-0060, 0076, 0101-0103, Fig 7-9. and Fig 15] Li discloses parent topic nodes representing labels at higher positions and child/sub-topic nodes representing labels at lower positions (See Fig 9)).
Claim(s) 7 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li et al. (US 2021/0382944 A1, hereinafter "Li") in view of Hill et al. (US 2020/0193265 A1, hereinafter "Hill") and Zhao et al. (US 20190318202 A1, hereinafter "Zhao").
Regarding Claim 7
Li in view of Hill discloses: The method according to claim 2,
Li in view of Hill does not explicitly disclose: wherein the first classification and/or the second classification is a binary classification.
However, Zhao discloses in the same field of endeavor: wherein the first classification and/or the second classification is a binary classification. ([Para 0065-0067, 0204, 0218, and Fig 4] discloses a binary classification.)
Li, Hill, and Zhao are both analogous art to the present invention because both are from the same field of endeavor directed to Machine learning.
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the method for Machine Learning disclosed by Li with the Ensemble Machine learning disclosed by Hill with the method for Binary classification models disclosed by Zhao. One of ordinary skill in the art would have been motivated to make this modification in order to indicate paths through a tree (Para 0066, Zhao).
Claim(s) 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li et al. (US 2021/0382944 A1, hereinafter "Li") in view of Hill et al. (US 2020/0193265 A1, hereinafter "Hill") and Yul et al. (US 2021/0224611 A1, hereinafter "Yu").
Regarding Claim 11
Li in view of Hill discloses: The method according to claim 2, further comprising:
Li in view of Hill does not explicitly disclose: providing a data point that includes a label for the first classification and/or a label for the second classification; and training the first classifier and/or the second classifier depending on the data point.
However, Yu discloses in the same field of endeavor: providing a data point that includes a label for the first classification and/or a label for the second classification ([Para 0014, 0064-0066, 0071, 0073, 0096 and Fig 1-5] describes providing labeled testing/training data to a model for classification.); and
training the first classifier and/or the second classifier depending on the data point. ([Para 0014, 0064-0066, 0071, 0073, 0096 and Fig 1-5] describes providing labeled testing/training to identify errors and adjust parameters (i.e., training).)
Li, Hill, and Yu are both analogous art to the present invention because both are from the same field of endeavor directed to Machine learning.
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the method for Machine Learning disclosed by Li with the Ensemble Machine learning disclosed by Hill with the method for Training Machine Learning models disclosed by Yu. One of ordinary skill in the art would have been motivated to make this modification in order to reduce erroneous classifications (Para 0065, Yu).
Claim(s) 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li et al. (US 2021/0382944 A1, hereinafter "Li") in view of Hill et al. (US 2020/0193265 A1, hereinafter "Hill") and Kim et al. (US 2021/0397890 A1, hereinafter "Kim").
Regarding Claim 14
Li in view of Hill discloses: The method according to claim 12,
Li in view of Hill does not explicitly disclose: wherein the training of the model includes: determining a first loss from an output of the first classifier and a second loss from an output of the second classifier; and backpropagating the first and second losses either to train weights in the first classifier and in the second classifier depending on both the first and second losses or to train the first classifier depending on the first loss and independent of the second loss and to train the second classifier depending on the second loss and independent of the first loss.
However, Kim discloses in the same field of endeavor: wherein the training of the model includes: determining a first loss from an output of the first classifier and a second loss from an output of the second classifier ([Para 0128-0129, 00153-0154 Fig 5 and Fig 8] describes a first loss function from a first classifier and a second loss function from a second classifier.); and
backpropagating the first and second losses either to train weights in the first classifier and in the second classifier depending on both the first and second losses or to train the first classifier depending on the first loss and independent of the second loss and to train the second classifier depending on the second loss and independent of the first loss. ([Para 0128-0131, 00153-0154 Fig 5 and Fig 8] describes performing backpropagation of the total loss function (i.e., depending on both first and second loss) to train weights. )
Li, Hill, and Kim are both analogous art to the present invention because both are from the same field of endeavor directed to Machine learning.
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the method for Machine Learning disclosed by Li with the Ensemble Machine learning disclosed by Hill with the method for Training Machine Learning models disclosed by Kim. One of ordinary skill in the art would have been motivated to make this modification in order minimize loss (Para 0010, Kim).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Misra et al. (US 20210303783 A1) describes transformer and ensemble learning (Para 0039-0040). Ma et al. (US 20200311613 A1) describes transformer tensors and ensemble learning (Abstract).
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TEWODROS E MENGISTU whose telephone number is (571)270-7714. The examiner can normally be reached Mon-Fri 9:30-5:30.
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, ABDULLAH KAWSAR can be reached at (571)270-3169. 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.
/TEWODROS E MENGISTU/Examiner, Art Unit 2127