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
This action is in response to the claims filed 5/2/2023:
Claims 1 – 21 are pending.
Claims 1 is independent.
Specification
The disclosure is objected to because of the following informalities:
The title of the invention is not descriptive. A new title is required that is clearly indicative of the invention to which the claims are directed.
Claim Objections
Claim 1 objected to because of the following informalities: Regarding claim 1, "A predetermined threshold" and "the first predetermined threshold" are inconsistent. As there is no other predetermined threshold recited Examiner acknowledges that the intent is clear, however, recommends amending "A predetermined threshold" to "A first predetermined threshold" for consistency. Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 3, 7-9, and 19-21 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claim 3, "one or more Miner processes" is indefinite. Miner processes appears to be a coined term, however, it is not clearly defined in the instant specification, and the instant specification explicitly states that the provided definition of Miner processes is non-limiting, such that one of ordinary skill in the art could not reasonable determine the scope of one or more Miner processes. Similarly, Examiner notes that "one or more Miner processes" has ambiguous multiplicity. It's unclear if one Miner can have multiple processes (processes of one Miner) or the claim should be read as having a Miner process for a plurality of Miners. Examiner also notes that neither "Miner" or "Miner processes" are terms of the art that a person of ordinary skill in the art would recognize. For at least these reasons the claim is seen as being indefinite. In the interest of further Examination the claim is interpreted as simply "one or more process(es)".
Regarding claims 7 and 8, "the content extraction" lacks antecedent basis. "Content extraction" is recommended.
Regarding claim 7, "preferably wherein the natural language processing model includes a Bidirectional Encoder Representations from Transformers" is indefinite. It's unclear whether or not the BERT model is required or optional in view of the language "preferably". In the interest of further Examination the claim limitation has been interpreted as being one of: a NLP model including BERT or an NLP model not including BERT".
Regarding claim 19, "to implement any one of method claim 1" is indefinite. "Any one" implies a set but then points to a single claim. In the interest of further Examination the Examiner has interpreted this as "to implement the method of claim 1".
Claims 9 and 20-21 are rejected with respect to their dependence on rejected claims 8 and 19, respectively.
Claim Rejections - 35 USC § 101
101 Rejection
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 1-21 are rejected under 35 USC § 101 because the claimed invention is directed to non-statutory subject matter.
Regarding Claim 1: Claim 1 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 1 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: Claim 1 under its broadest reasonable interpretation is a series of mental processes. For example, but for the generic computer components language, the above limitations in the context of this claim encompass natural language processing, including the following:
Identifying content (observation, evaluation, and judgement)
applying an intelligent filter to the one or more semantic context vectors to determine an output value, wherein the output value is indicative of the likelihood the content matches one or more predefined filter parameters (observation, evaluation, and judgement),
comparing the output value to a predetermined threshold (observation, evaluation, and judgement)
Therefore, claim 1 recites an abstract idea which is a judicial exception.
Step 2A Prong Two Analysis: Claim 1 recites additional elements “A non-transitory computer-readable recording medium having stored therein a program that causes a computer to execute a process”. However, these additional features are computer components recited at a high-level of generality, such that they amount to no more than mere instructions to apply the judicial exception using a generic computer component. An additional element that merely recites the words “apply it” (or an equivalent) with the judicial exception, or merely includes instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea, does not integrate the judicial exception into a practical application (See MPEP 2106.05(f)). Claim 1 also recites additional elements “receiving one or more semantic context vectors, wherein the one or more semantic context vectors relate to published content” and “transmitting a notification to a user device if the output value exceeds the first predetermined threshold, wherein the notification identifies the content for which the output value exceeds the first predetermined threshold” which amounts to gathering and outputting data which is insignificant extra-solution activity (see MPEP 2106.05(g)). Therefore, claim 1 is directed to a judicial exception.
Step 2B Analysis: Claim 1 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the lack of integration of the abstract idea into a practical application, the additional elements recited in claim 1 amount to no more than mere instructions to apply the judicial exception using a generic computer component and insignificant extra-solution activity. The gathering and outputting of data is considered well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i)).
For the reasons above, claim 1 is rejected as being directed to non-patentable subject matter under §101. This rejection applies equally to dependent claims 2-21. The additional limitations of the dependent claims are addressed briefly below:
Dependent claim 2 recites additional insignificant extra-solution activity of gathering and outputting data “receiving one or more input descriptors, wherein each input descriptor identifies the published content” as well as additional observation, evaluation, and judgement “performing content extraction on the published content to determine the one or more semantic context vectors”
Dependent claim 3 recites additional observation, evaluation, and judgement “performing content identification on one or more sources to identify the one or more input descriptors, wherein the content identification is performed by one or more Miner processes.”
Dependent claim 4 recites additional insignificant extra-solution activity “The method of claim 2, in which each input descriptor is a Uniform Resource Locator (URL)” which amounts to selection of a data type (See MPEP 2106.05(g) which is well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)).
Dependent claim 5 recites additional insignificant extra-solution activity “receiving the one or more predefined filter parameters from a user device” which amounts to gathering data (See MPEP 2106.05(g)) which is well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i))
Dependent claim 6 recites additional observation, evaluation, and judgement “the first predetermined threshold is based on a maximised Area Under Curve”
Dependent claim 7 recites additional observation, evaluation, and judgement “in which the content extraction is based on a natural language processing model; preferably wherein the natural language processing model includes a Bidirectional Encoder Representations from Transformers”
Dependent claim 8 recites additional observation, evaluation, and judgement “the content extraction is based on an ensembled model”
Dependent claim 9 recites additional observation, evaluation, and judgement “the ensembled model comprises a first model of a text classification model; and a second model of an industry classification model”
Dependent claim 10 recites additional observation, evaluation, and judgement “performing the content extraction on text of the published content to transform the text to the one or more semantic context vectors; wherein the one or more semantic context vectors include a numerical representation of a meaning of the text of the published content”
Dependent claim 11 recites additional instructions to apply the judicial exception using generic computer components “in which the intelligent filter is based on a Multilayer Perceptron”
Dependent claim 12 recites additional instructions to apply the judicial exception using generic computer components “in which the Multilayer Perceptron is formed of two or more perceptrons and comprises: an input layer to receive the input semantic context vectors; one or more hidden layers to receive a set of weighted inputs and to determine the output value based on an activation function” (this describes a standard multilayer perceptron which is a generic computer component)
Dependent claim 13 recites additional mathematical calculations and relationships “multiplying each of the one or more semantic context vectors by the set of weights; and adding a bias”
Dependent claim 14 recites additional observation, evaluation, and judgement “normalising the output of the intelligent filter”
Dependent claim 15 recites additional instructions to apply the judicial exception using generic computer components “training the intelligent filter based on an initial training set, wherein the initial training set comprises a plurality of input descriptors relating to published content, wherein at least a subset of the published content relating to the input descriptors are reviewed manually” (as previously mentioned a multilayer perceptron is a generic computer component which is necessarily trained on arbitrary data (training set))
Dependent claim 16 recites additional observation, evaluation, and judgement “training the intelligent filter based on the output value of the intelligent filter” (as previously mentioned a multilayer perceptron is a generic computer component which is necessarily trained on arbitrary data (training set) and based on the MLP models output)
Dependent claim 17 recites additional observation, evaluation, and judgement “the training compensates for drift”
Dependent claim 18 recites additional instructions to apply the judicial exception using generic computer components “A computer program product comprising computer readable executable code configured to implement the method of claim 1”.
Dependent claim 19 recites additional instructions to apply the judicial exception using generic computer components “A system, comprising: an intelligent filter module; and a processor, wherein the processor is configured to implement the method of claim 1”.
Dependent claim 20 recites additional instructions to apply the judicial exception using generic computer components “a context extractor module and a content identifier module”
Dependent claim 21 recites additional instructions to apply the judicial exception using generic computer components “a serverless environment, wherein the modules are executed on the serverless environment” (a generic personal computer without internet connection could reasonably be interpreted as a serverless environment)
Therefore, when considering the elements separately and in combination, they do not add significantly more to the inventive concept. Accordingly, claims 1-21 are rejected under 35 U.S.C. § 101.
Claim Rejections - 35 USC § 102
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 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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1, 2, 3, 5, 7, 10, 18, 19, 20, and 21 are rejected under U.S.C. §102(a)(1) as being anticipated by Rose (WO1995029452A1).
Regarding claim 1, Rose teaches A computer-implemented method of identifying content, the method implemented by a processing resource, the method comprising:([p. 5 l. 14-27] "The illustrated architecture comprises a client-server arrangement, in which a database of information is stored at a server computer 10, and is accessible through various client computers 12, 14. The server 10 can be any suitable micro, mini or mainframe computer having sufficient storage capacity to accommodate all of the items of information to be presented to users. The client computers can be suitable desktop computers 12 or portable computers 14, e.g. notebook computers, having the ability to access the server computer 10")
receiving one or more semantic context vectors, wherein the one or more semantic context vectors relate to published content;([p. 7 l. 15-23] "columns can contain other types of information that may assist the user in determining whether to retrieve a particular message, such as the date on which the message was posted to the system" [p. 10 l. 12-18] "the elements of a message, such as words in a document, are used to compute vectors for the messages and user profiles. It will be appreciated, of course, that the vector need not be based solely on such elements. Rather, any suitable attribute of a message can be employed to determine its relevance vector")
applying an intelligent filter to the one or more semantic context vectors to determine an output value, wherein the output value is indicative of the likelihood the content matches one or more predefined filter parameters;([p. 10 l. 24-30] "A prediction of a user's likely interest in a particular document is based on the similarity between the document's vector and the user's profile vector. For example, as shown in Figure 5B, a score of the document's relevance can be indicated by the cosine of the angle between that document's vector and the user's profile vector. A document having a vector which is close to that of the user's profile, such as Document 4, will be highly ranked, whereas those which are significantly different will have a lower ranking, for example Document 1" Rose's server retrieves a user's profile and uses it to rank messages. The system compares each document message vector to the user's profile vector, producing a relevance score explicitly described as a prediction (output) of "likely interest" and a "score of [...] relevance" (cosine similarity).)
comparing the output value to a predetermined threshold; and([p. 7 l. 24-31] "the client program or the server program can employ a suitable selection threshold, so that messages having a ranking below a certain threshold are not displayed")
transmitting a notification to a user device if the output value exceeds the first predetermined threshold, wherein the notification identifies the content for which the output value exceeds the first predetermined threshold.([p. 10 l. 24-30] "A prediction of a user's likely interest in a particular document is based on the similarity between the document's vector and the user's profile vector." [p. 7 l. 24-31] "the client program or the server program can employ a suitable selection threshold, so that messages having a ranking below a certain threshold are not displayed" [p. 8 l. 1] "only those messages whose ranking value exceeds a certain limit can be displayed." [Abstract] "the system delivers to that user an identification of only those items of information which are believed to be relevant to the user's interest." In Rose relevance is mathematically defined by similarity score, explicitly gated with a threshold, and the notification/transmission is explicitly conditioned on the relevance threshold.).
Regarding claim 2, Rose teaches The method of claim 1, further comprising: receiving one or more input descriptors, wherein each input descriptor identifies the published content; and(Rose [p. 6 l. 13-16] "the message database has associated therewith an index 24, which provides a representation of each of the stored messages 22, for example its title. The index can contain other information pertinent to the stored messages as well")
performing content extraction on the published content to determine the one or more semantic context vectors.(Rose [p. 9 l. 28-29] "the results of this procedure is a vector of weights, which represents the content of the document" [p. 10 l. 24-30] "A prediction of a user's likely interest in a particular document is based on the similarity between the document's vector and the user's profile vector. For example, as shown in Figure 5B, a score of the document's relevance can be indicated by the cosine of the angle between that document's vector and the user's profile vector. A document having a vector which is close to that of the user's profile, such as Document 4, will be highly ranked, whereas those which are significantly different will have a lower ranking, for example Document 1").
Regarding claim 3, Rose teaches The method of claim 2, further comprising: performing content identification on one or more sources (Rose [p. 6 l. 5-15] "The message server carries out communications with each of the clients, for example over a network, and retrieves information from two databases, a user database 18 and a message database 20" Database interpreted as source)
to identify the one or more input descriptors, (Rose [p. 7 l. 10-23] "the interface comprises a window 26 containing a number of columns of information. The left hand column 28 indicates the relative ranking score of each message, for example in the form of a horizontal thermometer-type bar 30. The remaining columns can contain other types of information that may assist the user in determining whether to retrieve a particular message, such as the date on which the message was posted to the system, the message's author, and the title or subject of the message. The information that is displayed within the window can be stored as part of the index 24")
wherein the content identification is performed by one or more Miner processes.(Rose [p. 6 l. 5-16] "the server program contains a message server 16. The message server carries out communications with each of the clients […] the message database has associated therewith an index 24" Message server interpreted as Miner process).
Regarding claim 5, Rose teaches The method of claim 1, further comprising: receiving the one or more predefined filter parameters from a user device.(Rose [p. 8 l. 2-3] "Preferably, the selection threshold can be changed by the user").
Regarding claim 7, Rose teaches The method of claim 1, in which the content extraction is based on a natural language processing model; preferably wherein the natural language processing model includes a Bidirectional Encoder Representations from Transformers.(Rose [p. 9 l. 19-29] "words which frequently occur in a particular language are given a low weight value, while those which are rarely used have a high weight value. The weight value for each term is multiplied by the number of times that term occurs in the document").
Regarding claim 10, Rose teaches The method of claim 2, further comprising: performing the content extraction on text of the published content to transform the text to the one or more semantic context vectors; (Rose [p. 10 l. 24-30] "A prediction of a user's likely interest in a particular document is based on the similarity between the document's vector and the user's profile vector. For example, as shown in Figure 5B, a score of the document's relevance can be indicated by the cosine of the angle between that document's vector and the user's profile vector. A document having a vector which is close to that of the user's profile, such as Document 4, will be highly ranked, whereas those which are significantly different will have a lower ranking, for example Document 1" Rose's server retrieves a user's profile and uses it to rank messages. The system compares each document message vector to the user's profile vector, producing a relevance score explicitly described as a prediction (output) of "likely interest" and a "score of [...] relevance" (cosine similarity).)
wherein the one or more semantic context vectors include a numerical representation of a meaning of the text of the published content.(Rose [p. 24-25] "each word, in a document can be assigned a weight, based on its statistical importance" [p. 9 l. 27-29] "the results of this procedure is a vector of weights, which represents the content of the document" [pp. 9-10] "the vectors for document content would likely have hundreds or thousands of dimensions, depending on the number of terms that are monitored").
Regarding claim 18, Rose teaches A computer program product comprising computer readable executable code configured to implement the method of claim 1.(Rose [p. 5 l. 14-27] "The illustrated architecture comprises a client-server arrangement, in which a database of information is stored at a server computer 10, and is accessible through various client computers 12, 14. The server 10 can be any suitable micro, mini or mainframe computer having sufficient storage capacity to accommodate all of the items of information to be presented to users. The client computers can be suitable desktop computers 12 or portable computers 14, e.g. notebook computers, having the ability to access the server computer 10").
Regarding claim 19, Rose teaches A system, comprising: an intelligent filter module; and(Rose [p. 10 l. 24-30] "A prediction of a user's likely interest in a particular document is based on the similarity between the document's vector and the user's profile vector. For example, as shown in Figure 5B, a score of the document's relevance can be indicated by the cosine of the angle between that document's vector and the user's profile vector. A document having a vector which is close to that of the user's profile, such as Document 4, will be highly ranked, whereas those which are significantly different will have a lower ranking, for example Document 1" Rose's server retrieves a user's profile and uses it to rank messages. The system compares each document message vector to the user's profile vector, producing a relevance score explicitly described as a prediction (output) of "likely interest" and a "score of [...] relevance" (cosine similarity).)
a processor, wherein the processor is configured to implement any one of method claim 1.(Rose [p. 5 l. 14-27] "The illustrated architecture comprises a client-server arrangement, in which a database of information is stored at a server computer 10, and is accessible through various client computers 12, 14. The server 10 can be any suitable micro, mini or mainframe computer having sufficient storage capacity to accommodate all of the items of information to be presented to users. The client computers can be suitable desktop computers 12 or portable computers 14, e.g. notebook computers, having the ability to access the server computer 10").
Regarding claim 20, Rose teaches The system of claim 19, further comprising: a context extractor module (Rose [p. 9 l. 25-29] "the results of this procedure is a vector of weights, which represents the content of the document" content vector computation process interpreted as content extractor module)
and a content identifier module.(Rose [p. 6 l. 10-16] "the message database has associated therewith an index 24, which provides a representation of each of the stored messages 22, for example its title. The index can contain other information pertinent to the stored messages as well." Software architecture including an index that provides a representation/identifier for each message interpreted as content identifier module).
Regarding claim 21, Rose teaches The system of claim 19, in further comprising a serverless environment, wherein the modules are executed on the serverless environment.(Rose [p. 4 l. 19-28] "To facilitate an understanding of the principles of the present invention, they are described hereinafter with reference to the implementation of the invention in a system having multiple personal computers that are connected via a network. It will be appreciated, however, that the practical applications of the invention are not limited to this particular environment. Rather, the invention can find utility in any situation which facilitates communication between users and provides for access to information. For example, it is equally applicable to other types of multiuser computer systems, such as mainframe and minicomputer systems in which many users can have simultaneous access to the same computer." Mainframe interpreted as serverless environment).
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 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.
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.
Claims 4, 8, 9, 14, 15, 16, and 17 are rejected under U.S.C. §103 as being unpatentable over the combination of Rose and Batra (US20200184017A1).
Regarding claim 4, Rose teaches The method of claim 2.
However, Rose doesn't explicitly teach in which each input descriptor is a Uniform Resource Locator (URL).
Batra, in the same field of endeavor, teaches each input descriptor is a Uniform Resource Locator (URL). ([¶0021] "data subscriber 103 may be configured to store a web link (e.g., URL) associated with the data. Storing only the web link may decrease storage needs in subscriber database 105. Subscriber database 105 may comprise any database, data structure, or the like capable of storing and maintaining data and/or web links" See also FIG. 3 which shows machine learning system obtaining data (like URLs) from subscriber database 103).
Rose as well as Batra are directed towards natural language processing. Therefore, Rose as well as Batra are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Rose with the teachings of Batra by using the machine learning natural language processing ensemble model and additional input data types disclosed in Barta for identifying relevant content and generating alerts as done in Rose. Rose explicitly invites alternate ranking techniques and notifications ([p. 3 l. 20] “A variety of techniques can be used to rank the information” [p. 13] “This type of predictor acts in the manner similar to a neural network”) and Batra provides as additional motivation for combination ([¶0017] “accuracy in the identification of data of interest may be improved and the system may be easily adaptable and scalable to different types of datasets and domains that may typically require a subject matter expert's review”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 8, Rose teaches The method of claim 1.
However, Rose doesn't explicitly teach in which the content extraction is based on an ensembled model.
Batra, in the same field of endeavor, teaches The method of claim 1, in which the content extraction is based on an ensembled model. ([¶0003] "The system may input the unstructured data into a first machine learning model, a second machine learning model, a named entity recognition (NER) model, and a semantic role labeling (SRL) model. " [¶0004] "the system inputs the output of at least one of the first machine learning model, the second machine learning model, the NER model, the SRL model, or the sentiment score into a gradient boosted regression tree (GBRT) machine learning model").
Rose as well as Batra are directed towards natural language processing. Therefore, Rose as well as Batra are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Rose with the teachings of Batra by using the machine learning natural language processing ensemble model and additional input data types disclosed in Barta for identifying relevant content and generating alerts as done in Rose. Rose explicitly invites alternate ranking techniques and notifications ([p. 3 l. 20] “A variety of techniques can be used to rank the information” [p. 13] “This type of predictor acts in the manner similar to a neural network”) and Batra provides as additional motivation for combination ([¶0017] “accuracy in the identification of data of interest may be improved and the system may be easily adaptable and scalable to different types of datasets and domains that may typically require a subject matter expert's review”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 9, the combination of Rose, and Batra teaches The method of claim 8, in which the ensembled model comprises a first model of a text classification model; (Batra [¶0051] " The first machine learning may comprise a Naïve Bayes machine learning model (NB 372). For example, and in accordance with various embodiments, the first machine learning model may comprise the machine learning model trained in method 401, with brief reference to FIG. 4. The first machine learning model may process the data based on a bag of words technique used to identify a given set of topics and a set of terms associated with each topic. The data classification may be polymorphic and may be associated with multiple topics. The first machine learning model may process the data to determine whether the data is of interest in the system")
and a second model of an industry classification model.(Batra [¶0054] "Machine learning system 370 inputs the preprocessed data into semantic role labeling (SRL) model 378 (step 518). SRL model 378 may comprise one or more semantic role labeling algorithms, relational machine learning algorithms, or the like. For example, SRL model 378 may be configured to perform various classification, identification, and predictions tasks such as, for example, collective classification (e.g., prediction of the class of several words or phrases in the data, based on the attribute and relationships between words or phrases), object link prediction (e.g., predicting whether two or more words or phrases are related), object link-based clustering (e.g., the grouping of similar words and phrases, the filtering of data that is relevant to a located entity, etc.), social network modelling, entity resolution (e.g., the identification of equivalent words and/or phrases for a common entity), and/or the like. In various embodiments, an output from SRL model 378 may identify actions, events, etc. that are happening about or to the named entity, whether the name entity is the source or target of the action, event, etc., and/or the like." SRL model interpreted as industry classification model).
Regarding claim 14, Rose teaches The method of claim 1.
However, Rose doesn't explicitly teach, further comprising: normalising the output of the intelligent filter.
Batra, in the same field of endeavor, teaches The method of claim 1, further comprising: normalising the output of the intelligent filter.([¶0007] "The one or more training keywords may be identified by analyzing prefiltered training data using at least one of a latent Dirichlet allocation (LDA) model, a correlated topic model, a word2vec processing algorithm, a word frequency analysis, or a phrase frequency analysis. The generated training dataset may be prefiltered by at least one of a parts-of-speech tagging process, a lemmatization process, removing stop words, generating n-grams, normalizing or filtering email IDs, numbers, and URLs, or replacing proper nouns with common nouns." Barta explicitly teaches that the output of an intelligent filter is a training dataset which is then explicitly normalized).
Rose as well as Batra are directed towards natural language processing. Therefore, Rose as well as Batra are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Rose with the teachings of Batra by using the machine learning natural language processing ensemble model and additional input data types disclosed in Barta for identifying relevant content and generating alerts as done in Rose. Rose explicitly invites alternate ranking techniques and notifications ([p. 3 l. 20] “A variety of techniques can be used to rank the information” [p. 13] “This type of predictor acts in the manner similar to a neural network”) and Batra provides as additional motivation for combination ([¶0017] “accuracy in the identification of data of interest may be improved and the system may be easily adaptable and scalable to different types of datasets and domains that may typically require a subject matter expert's review”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 15, Rose teaches The method of claim 1.
However, Rose doesn't explicitly teach, further comprising: training the intelligent filter based on an initial training set.
Batra, in the same field of endeavor, teaches training the intelligent filter based on an initial training set, ([¶0007] "at least one of the first machine learning model or the second machine learning model are trained using a generated training dataset. The generated training dataset may be generated by filtering public business data based on one or more training keywords. The one or more training keywords may be identified by analyzing prefiltered training data using at least one of a latent Dirichlet allocation (LDA) model, a correlated topic model, a word2vec processing algorithm, a word frequency analysis, or a phrase frequency analysis.").
Rose as well as Batra are directed towards natural language processing. Therefore, Rose as well as Batra are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Rose with the teachings of Batra by using the machine learning natural language processing ensemble model and additional input data types disclosed in Barta for identifying relevant content and generating alerts as done in Rose. Rose explicitly invites alternate ranking techniques and notifications ([p. 3 l. 20] “A variety of techniques can be used to rank the information” [p. 13] “This type of predictor acts in the manner similar to a neural network”) and Batra provides as additional motivation for combination ([¶0017] “accuracy in the identification of data of interest may be improved and the system may be easily adaptable and scalable to different types of datasets and domains that may typically require a subject matter expert's review”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 16, Rose teaches The method of claim 1.
However, Rose doesn't explicitly teach further comprising: training the intelligent filter based on the output value of the intelligent filter.
wherein at least a subset of the published content relating to the input descriptors are reviewed manually.
Batra, in the same field of endeavor, teaches training the intelligent filter based on the output value of the intelligent filter. ([¶0007] "at least one of the first machine learning model or the second machine learning model are trained using a generated training dataset. The generated training dataset may be generated by filtering public business data based on one or more training keywords. The one or more training keywords may be identified by analyzing prefiltered training data using at least one of a latent Dirichlet allocation (LDA) model, a correlated topic model, a word2vec processing algorithm, a word frequency analysis, or a phrase frequency analysis.")
wherein at least a subset of the published content relating to the input descriptors are reviewed manually.([¶0017] "Typically, manual browsing and review of news articles by a subject matter expert is needed to identify news, articles, posts, and the like of interest for a particular entity" [¶0017] "by using a plurality of scores from various machine learning models, accuracy in the identification of data of interest may be improved and the system may be easily adaptable and scalable to different types of datasets and domains that may typically require a subject matter expert's review.").
Rose as well as Cui are directed towards natural language processing and specifically about filtering unwanted text. Therefore, Rose as well as Cui are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Rose with the teachings of Cui by using the model in Cui as the text/message filtering model in Rose. Rose explicitly invites alternate ranking techniques and notifications ([p. 3 l. 20] “A variety of techniques can be used to rank the information” [p. 13] “This type of predictor acts in the manner similar to a neural network”) and Cui provides as additional motivation for combination ([p. 2] “To solve the problem of semantically filtering objectionable short text on the Internet and to realize automatic and rapid filtering, this paper proposes a biterm topic modelling- (BTM-) based adaptive objectionable short text filtering framework”).
Regarding claim 17, the combination of Rose, and Batra teaches The method of claim 15, in which the training compensates for drift.(Batra [¶0041] "feedback may be provided to model building system 110 to update future training datasets in response to identifying false positives" When the deployed system starts generating false positives due to changing real-world data (drift), Batra explicitly updates future training datasets based on those errors).
Claim 6 is rejected under U.S.C. §103 as being unpatentable over the combination of Rose and Cui (“A BTM-Based Adaptive Objectionable Short Text Filtering Framework”, 2022).
Regarding claim 6, Rose teaches The method of claim 1.
However, Rose doesn't explicitly teach in which the first predetermined threshold is based on a maximised Area Under Curve.
Cui, in the same field of endeavor, teaches the first predetermined threshold is based on a maximised Area Under Curve. ([p. 2] "Once the number of matches reaches a predefined threshold, the text is determined to be objectionable text." [p. 8] "We use the receiver operating characteristic (ROC) curve and area under the ROC curve (AUC) for evaluation. Regarding the ROC curve, the ordinate represents the detection rate (DR), and the abscissa represents the false alarm rate (FR). They also represent detection precision (true positive rate) and false positive rate, respectively. The AUC refers to the size of the area under the ROC curve in the coordinate system. The larger the AUC value is, the better the detection effect").
Rose as well as Cui are directed towards natural language processing and specifically about filtering unwanted text. Therefore, Rose as well as Cui are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Rose with the teachings of Cui by using the model in Cui as the text/message filtering model in Rose. Rose explicitly invites alternate ranking techniques and notifications ([p. 3 l. 20] “A variety of techniques can be used to rank the information” [p. 13] “This type of predictor acts in the manner similar to a neural network”) and Cui provides as additional motivation for combination ([p. 2] “To solve the problem of semantically filtering objectionable short text on the Internet and to realize automatic and rapid filtering, this paper proposes a biterm topic modelling- (BTM-) based adaptive objectionable short text filtering framework”).
Claims 11, 12, and 13 are rejected under U.S.C. §103 as being unpatentable over the combination of Rose and He (“Neural Collaborative Filtering”, 2017).
PNG
media_image1.png
422
718
media_image1.png
Greyscale
FIG. 2 of He
Regarding claim 11, Rose teaches The method of claim 1.
However, Rose doesn't explicitly teach in which the intelligent filter is based on a Multilayer Perceptron.
He, in the same field of endeavor, teaches in which the intelligent filter is based on a Multilayer Perceptron. ([p. 3 §3] "We first present the general NCF framework, elaborating how to learn NCF with a probabilistic model that emphasizes the binary property of implicit data. We then show that MF can be expressed and generalized under NCF. To explore DNNs for collaborative filtering, we then pro pose an instantiation of NCF, using a multi-layer perceptron (MLP) to learn the user–item interaction function").
Rose as well as He are directed towards natural language processing. Therefore, Rose as well as He are reasonably pertinent analogous art. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Rose with the teachings of He by using the neural collaborative filtering model as the text processing model for message filtering in Rose. Rose explicitly invites alternate ranking techniques and notifications ([p. 3 l. 20] “A variety of techniques can be used to rank the information” [p. 13] “This type of predictor acts in the manner similar to a neural network”) and He provides as additional motivation for combination ([p. 1] “Empirical evidence shows that using deeper layers of neural networks offers better recommendation performance […] recommender systems play a pivotal role in alleviating information overload”).
Regarding claim 12, the combination of Rose, and He teaches The method of claim 11, in which the Multilayer Perceptron is formed of two or more perceptrons and comprises:(He [p. 3 §3] "We first present the general NCF framework, elaborating how to learn NCF with a probabilistic model that emphasizes the binary property of implicit data. We then show that MF can be expressed and generalized under NCF. To explore DNNs for collaborative filtering, we then pro pose an instantiation of NCF, using a multi-layer perceptron (MLP) to learn the user–item interaction function" A MLP has two or more perceptrons by definition (as suggested by the name))
an input layer to receive the input semantic context vectors;(He [p. 3] "Input Layer" See also FIG. 2)
one or more hidden layers to receive a set of weighted inputs and to determine the output value based on an activation function; and(He [p. 3] "Each layer of the neural CF layers can be customized to discover certain latent structures of user–item interactions. The dimension of the last hidden layer X determines the model’s capability" See also FIG. 2)
an output layer to predict the likelihood the content matches the one or more predefined filter parameters.(He [p. 3] "The final output layer is the predicted score ˆyui, and training is performed by minimizing the pointwise loss between ˆyui and its target value yui" See also FIG. 2).
Regarding claim 13, the combination of Rose, and He teaches The method of claim 12, further comprising: multiplying each of the one or more semantic context vectors by the set of weights; and adding a bias.(He [p. 4 §3.3] "we propose to add hidden layers on the concatenated vector, using a standard MLP to learn the interaction between user and item latent features. In this sense, we can endow the model a large level of flexibility and non-linearity to learn the interactions between pu and qi, rather than the way of GMF that uses only a fixed element-wise product on them. More precisely, the MLP model under our NCF framework is defined as [...] a2(WT2*z1+b2)" a2(WT2*z1+b2 is exactly multiplying one or more semantic context vectors by the set of weights and adding a bias).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Pugoy (“BERT-Based Neural Collaborative Filtering and Fixed-Length Contiguous Tokens Explanation”, 2020) is directed towards a text filtering method using a BERT / MLP based neural collaborative filtering architecture.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SIDNEY VINCENT BOSTWICK whose telephone number is (571)272-4720. The examiner can normally be reached M-F 7:30am-5:00pm EST.
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, Miranda Huang can be reached on (571)270-7092. 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.
/SIDNEY VINCENT BOSTWICK/Examiner, Art Unit 2124