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 .
Specification
The disclosure is objected to because of the following informalities:
Paragraphs [000119], [000120] contain struck through terms. Appropriate correction is required.
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.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over Guo [9535960] in view of Rudzicz [CA3048436]
With regard to claim 1 Guo teaches A computer-implemented method for ranking (Guo, Abstract “The search engine operates by ranking a plurality of documents”) digital documents (Id) at a digital platform (Guo, Fig. 1, 120 “Ranking Framework”), the method comprising:
receiving a search query as the query (Guo, Column 10, lines 33-35 “Here, a first transformation module 504 receives query information associated with the query inputted by the user”) submitted by a user as the user (Id) to the digital platform as 504 (Id) is part of the ranking framework 120 (Guo, Column 10, lines 32-33 “FIG. 5 shows another implementation 502 of the ranking framework 120.”);
generating a first vector embedding representative of the search query as query concept vector yQ (Gou, Column 10, lines 35-36 “The first transformation module 504 uses a first instance of the model 106 to project the query information into a query concept vector yQ in a semantic space.”);
identifying, in a search index of the digital platform, a plurality of digital document candidates that are responsive to the search query as the set of k documents (Gou, Column 10 lines 22-25 “In that variation, the set of k documents in the data store 416 correspond to the set of k entities which match the user's query, without taking into consideration the context information.”; Gou, Column 4, lines 45-48 “The search engine 112 can optionally rely on an indexing mechanism (not shown) to retrieve documents, given specified search terms.”);
retrieving, for each digital document candidate of the plurality of digital document candidates as the set of k documents (Gou, Column 10 lines 22-25 “In that variation, the set of k documents in the data store 416 correspond to the set of k entities which match the user's query, without taking into consideration the context information.”), a second vector embedding representative thereof as document concept vector yD(Gou, Column 10, lines 42-48 “A third transformation module 508 receives document information which describes a candidate document, among a collection of such documents to be considered, stored in a data store 510. The third transformation module 508 uses a third instance of the model 106 to transform the document information into a document concept vector, y D ·”), the second vector having been generated prior to the receiving of the search query (Gou, Column 19, lines 25-29 “Alternatively, each document concept vector can be computed in advance as part of an offline process, and then stored along with the document to which it pertains, and properly indexed to allow for later retrieval.”);
identifying, for a given one of the plurality of digital document candidates as the set of k documents (Gou, Column 10 lines 22-25 “In that variation, the set of k documents in the data store 416 correspond to the set of k entities which match the user's query, without taking into consideration the context information.”), at least one phrase candidate as part of the document (Gou, Column 18, lines 29-30 “the document D can have multiple parts.”) that is [[ as the context of the query term in the document (Column 18 “that is configured to compare the text that surrounds a user's query (in a source document) with the title of each candidate document”);
generating a third vector embedding representative of the at least one phrase candidate as context query vector yC (Goa, Column 10, lines 40-43 “The second transformation module 506 uses a second instance of the model 106 to transform the context information into a context query vector yc·”) ;
based on the first, second, and third vector embeddings (Gou, Column 10, lines 58-60 “FIG. 5, alternatively, or in addition, a comparison module can also form a relevance measure based on any joint analysis of all three concept vectors, yQ, yD , and yc·”), determining, for the given one of the plurality of digital document candidates (Gou, Column 10, lines 60-64 “A ranking module 518 can then rank the plurality of candidate documents based on a plurality of features, including the relevance measures, for each document, fed to it by the comparison module 512.”), a respective value of a ranking parameter as the ranking module ranking (Id),
the ranking parameter being indicative of relevancy of the given of the plurality of digital document candidates to the search query as ranking based on the relevance measure (Id); and
ranking the plurality of digital document candidates based on respective values of the ranking parameter associated therewith as to rank (Id).
Gou does not explicitly teach that the context of the query term is specifically a lexically relation.
Rudzicz teaches identifying, for a given one of the plurality of digital document candidates as the documents (Rudzicz, ¶49 “The NB and SVM classifiers make use of lexical features that encode the probability of various unigrams and bigrams occurring in documents of a specific class”), at least one phrase candidate as unigrams and bigrams (Id) that is lexically related as lexical features (Id; ¶55 “(f) Cosine distance between pairs of vectorized sentences within a document”) to at least one term as between pairs of sentences (Id);
It would have been obvious to one of ordinary skill to which said subject matter pertains at the time the invention was filed to have implemented the context analysis performed by Guo using the text classifiers taught by Rudzicz as it yields the predictable results of generating embeddings representing the text (Rudzicz, ¶48) and classifying the text (Rudzicz, ¶46). One of ordinary skill in the art would recognize that the language analytics taught by Rudzicz may be used to identify context of the terms within the document (Rudzicz, ¶55; Gou, Column 18 lines 29-30) which may be used to perform query disambiguation (Rudzicz, ¶2; Gou, Column 1, lines 5-25).
With regard to claim 2 the proposed combination further teaches wherein the identifying the plurality of digital document candidates comprises applying a ranking function as generating the ranking score via the ranking module 912 (Gou, Column 19, lines 40-43 “In block 1116, the ranking framework 120 provides a search result based on the ranking score (e.g., after all other candidate documents have been processed in a similar manner to that described above).”; Column 14, lines 40-44 “A ranking module 912 may receive the relevance measures produced by the comparison modules (908, ... , 910). The ranking module 916 may then assign a ranking score to each candidate entity document based on the relevance measures, together with any other features.”).
With regard to claim 3, the proposed combination further teaches wherein the ranking function as ranking module 912 (Gou, Column 14, line 40) is an Okapi BM25 ranking function.
Gou does not explicitly teach that the ranking function is an Okapao BM25 ranking function. Rudzicz teaches an Okapi BM25 ranking function (Rudzicz, ¶64 “Search queries are conducted against the Solr instance, and the set of search results consist of the top n relevant documents, ranked according to their Okapi BM25 scores”).
It would have been obvious to one of ordinary skill to which said subject matter pertains at the time the invention was filed to have implemented the proposed combination using the known ranking function (e.g. Okapi BM25) taught by Rudzicz as a means of implementing the ranking architecture taught by Gou as it yields the predictable results of providing the model necessary for ranking the documents (Gou, Column 11, lines 3-9 “For example, some multi-level or cascading ranking architectures may use the deep learning model 106 at any level of the ranking operation, such as a secondary stage (e.g., as exemplified in FIG. 4). Other ranking architectures may apply the deep learning model 106 in a single inclusive ranking operation (e.g., as exemplified in FIG. 5).”).
With regard to claim 4 the proposed combination further teaches wherein the identifying the at least one phrase candidate comprises:
generating, for each phrase of a given one the plurality of digital document candidates, a respective phrase vector embedding as separate concept vectors for each part of the document (Column 18, lines 30-32 “The ranking framework 120 can generate separate concept vectors for these parts, and then form features based on any pairwise comparison of those concept vectors”);
determining, in an embedding space (Gou, Column 14, line 4 “in a semantic space”), a distance value (Gou, Column 14, line 39 “Manhattan distance determination technique.”) between the first vector embedding as C (Gou, Column 14, lines 21-23 “A first comparison module 908 determines the extent of a semantic relationship (e.g., similarity) between the context C and the document D1 in the semantic space”), representative of the search query (Gou, Column 14, lines 2-4 “first transformation module 902 produces a context concept vector yc which expresses a context (C) associated with a query (Q), in a semantic space.”), and the respective phrase vector embedding as the document D1 (Gou, Column 14, lines 21-23);
ranking phrases of the plurality of digital document candidates in accordance with respective distance values associated with respective phrase embeddings (Gou, Column 14, lines 40-44 “A ranking module 912 may receive the relevance measures produced by the comparison modules (908, ... , 910). The ranking module 916 may then assign a ranking score to each candidate entity document based on the relevance measures, together with any other features.”), thereby generating a ranked list as ranked candidate entity document (Id) of phrases for the search query as the parts of the document (Column 18, lines 30-32); and
selecting, from the ranked list of phrases, a top predetermined number of phrases (Gou, Column 11, lines 12-15 “Further, the term ranking, as broadly used herein, encompasses a filtering operation. In a filtering operation, a ranking architecture may use the deep learning model 106 to eliminate candidate documents from further consideration”).
With regard to claim 5 the proposed combination further teaches wherein the generating the respective phrase vector embedding comprises determining, for each term of the search query, a term frequency–inverse document frequency (TF-IDF) value, within a given one of the plurality of digital document candidates (Rudzicz, ¶49 “In embodiments, the labelling engine (104) uses tf-idf (term frequency-inverse document frequency) weighted unigram and bigram counts to determine the importance of unigrams and bigrams to the document under classification”) .
With regard to claim 6 the proposed combination further teaches wherein the generating the respective phrase vector embedding comprises applying thereto a text embedding algorithm (Rudzicz, ¶48 “fastText, which is an efficient model for sentence classification that first learns an embedded representation of documents which are then fed into a multinomial logistic classifier.”).
With regard to claim 7 the proposed combination further teaches wherein the text embedding algorithm is a FastText word embedding algorithm (Rudzicz, Page 9, lines 29-32 “Exemplary classifiers include, for example, naive Bayes (NB), support vector machines (SVMs, …, artificial neural networks (ANNs) and fastText, which is an efficient model for sentence classification that first learns an embedded representation of documents which are then fed into a multinomial logistic classifier.”).
With regard to claim 8 the proposed combination further teaches wherein the determining comprises feeding the first, second, and third vector embeddings to a consolidated ML model (Gou, column 14, lines 45-49 “A.3. The Training System FIG. 10 shows one implementation of the training system 104 of FIG. 1. In one illustrative and non-limiting case, the training system 104 processes a corpus of click-through data (provided in a data store 108), to generate the model 106.”),
the consolidated ML model having been trained (Gou, column 14, lines 45-49) to determine the respective values of the ranking parameter for each one of a given plurality of digital documents (Gou, Column 10, lines 60-64) based on vector embeddings (Gou, Column 10, lines 58-60) of (i) a respective search query used (Gou, Column 14, lines 54-55 “(1) queries submitted by actual users over some span of time;”) for identifying the given plurality of digital documents (Gou, Column 14, lines 55-58 “(2) an indication of documents that the users clicked on and the documents that the users failed to click on after submitting those queries;”); (ii) each one of the given plurality of digital documents responsive to the respective search query (Id); and (iii) at least one phrase candidate identified in the given plurality of digital documents as being lexically as lexical features (Rudzicz, ¶49; ¶55) related to at least one term of the respective search query (Gou, Column 14, lines 58-59 “(3) information describing the contexts associated with the respective queries”).
With regard to claim 9 the proposed combination further teaches training the consolidated ML model by:
generating a training set of data comprising a plurality of training digital objects (Gou, column 14, lines 45-49 “A.3. The Training System FIG. 10 shows one implementation of the training system 104 of FIG. 1. In one illustrative and non-limiting case, the training system 104 processes a corpus of click-through data (provided in a data store 108), to generate the model 106.”), a given one of which comprises: (i) a training vector (Gou, Column 12, lines 54-57 “Generally, it may be desirable to reduce the dimensionality of the input vectors for the purpose of performing training of the model 106 (in an offline phase of operation) in a more time-efficient and resource-efficient manner.”) embedding of a training search query (Gou, Column 15, lines 1-4 “Each example includes a context (C) associated with a particular submitted query (Q), a document (D+) that the user selected in response to the query (and its associated context)”); (ii) training vector embeddings of a plurality of training digital documents candidates responsive to the training search query (Id); and (ii) training phrase vector embedding as part of the document (Gou, Column 18, lines 29-30) of at least one training phrase candidate as the documents responsive to the query (Gou, Column 15, lines 1-4), identified in the plurality of training digital documents, that is lexically as lexical features (Rudzicz, ¶49; ¶55) related to at least one term of the training search query(Column 18 “that is configured to compare the text that surrounds a user's query (in a source document) with the title of each candidate document”); and (iv) a respective label for a given one of the plurality of training digital documents, the respective label being indicative of how relevant the given one of the plurality of training digital documents is to the training search query as the document being clicked or not clicked by the user, e.g. a + or - document (Gou, Column 15, lines 2-6 “a document (D+) that the user selected in response to the query (and its associated context), and at least one document (D-) that the user did not select in response to the query (and its associated context).”);
feeding the plurality of training digital objects to the consolidated ML model (Gou, Column 14, lines 46-49 “FIG. 10 shows one implementation of the training system 104 of FIG. 1. In one illustrative and non-limiting case, the training system 104 processes a corpus of click-through data (provided in a data store 108), to generate the model 106.”); and
minimizing (Gou, Column 16, lines 15-18 “In other words, the objective is to derive a set of parameter values that minimizes the above equation, with respect to all of the examples in the empirical click-through data, or some subset thereof.”), at each training iteration (Gou, Column 15, lines 24-26 “The training system 104 operates by using an iterative solving mechanism 1002 to iteratively achieve an objective defined an objective function 1004, by iteratively changing the parameter values of the model A”), a difference as the loss calculation (Gou, Column 16, lines 4-15; Equation 4) between a current training prediction of the consolidated ML model as P(D+ | C) and the respective label as (C, D+).
With regard to claim 10 the proposed combination further teaches wherein the respective label is generated by an ML model that has been pre-trained (Gou, column 14, lines 45-49 “A.3. The Training System FIG. 10 shows one implementation of the training system 104 of FIG. 1. In one illustrative and non-limiting case, the training system 104 processes a corpus of click-through data (provided in a data store 108), to generate the model 106.”), based on human assessor-generated labels (Gou, Column 14, lines 54-65), to determine a degree of relevancy of the given digital document to the respective search query (Gou, Column 10, lines 58-60 “FIG. 5, alternatively, or in addition, a comparison module can also form a relevance measure based on any joint analysis of all three concept vectors, yQ, yD , and yc·”).
With regard to claim 11 the proposed combination further teaches wherein the consolidated ML model comprises a Deep Semantic Similarity ML model (Gou, Column 3, lines 55-59 “FIG. 1 shows an environment 102 that includes a training system 104 for producing a deep learning model 106. A deep learning model 106 (henceforth, simply "model") refers to any model that expresses the underlying semantic content of an input linguistic item”).
With regard to claim 12 the proposed combination further teaches wherein the retrieving the second vector embedding comprises receiving the second vector embedding as document concept vector yD (Gou, Column 10, lines 42-48 “A third transformation module 508 receives document information which describes a candidate document, among a collection of such documents to be considered, stored in a data store 510. The third transformation module 508 uses a third instance of the model 106 to transform the document information into a document concept vector, y D ·”) from a second ML model (Fig 5, 106 is part of Semantic Transformation Module 508) that has been trained (Gou, Column 4, lines 2-4 “Subsection A.3 (below) provides further details regarding one implementation of the training system 104.”) to generate vector embeddings of input digital documents (Gou, Column 10, lines 42-48).
With regard to claim 13 the proposed combination further teaches training the second ML (Fig 5, 106 is part of Semantic Transformation Module 508) model by feeding thereto the plurality of digital documents (Gou, Column 14, lines 55-58) of the search index of the digital platform (Gou, Column 4, lines 45-48).
With regard to claim 14 the proposed combination further teaches wherein, prior to the determining, the method further comprises reducing a number of embeddings of each one of the first, second, and third vector embeddings (Gou, Column 11, lines 55-59 “The first transformation module 202 includes a dimensionality-reduction module (DRM) 702 and a deep structured semantic module (DSSM) 704. Similarly, the second transformation module 204 includes a DRM 706 and a DSSM 708”).
With regard to claim 15 the proposed combination further teaches the generating the first, second, and third vector embeddings comprises applying a Transformer-based machine-learning (ML) model (Gou, Column 11, lines 55-59 “The first transformation module 202 includes a dimensionality-reduction module (DRM) 702 and a deep structured semantic module (DSSM) 704. Similarly, the second transformation module 204 includes a DRM 706 and a DSSM 708”); and
the reducing the number of embeddings (Gou, Column 11, lines 55-59 “The first transformation module 202 includes a dimensionality-reduction module (DRM) 702 and a deep structured semantic module (DSSM) 704. Similarly, the second transformation module 204 includes a DRM 706 and a DSSM 708”) comprises progressively truncating outputs as reducing the dimension from 30k to 300 to 300 to 128 (Gou, Column 13, lines 38-45 “For instance, if a trigram hashing technique is used to produce the reduced-dimension vector, then the dimensionality layer 804 means that there are 30K unique trigrams within the original corpus of SOOK words (with the possible exclusion of some uninteresting trigrams). The layer 806 and the layer 808 each have a dimensionality of 300 elements. And the layer 810 has a dimensionality of 128 elements.”) of each intermediate layer of the Transformer-based ML model (Gou, Column 11, lines 3-7 “For example, some multi-level or cascading ranking architectures may use the deep learning model 106 at any level of the ranking operation, such as a secondary stage (e.g., as exemplified in FIG. 4).”) to a respective predetermined length of a given one of the first, second, and third vector embeddings as 128 elements (Gou, Column 13, lines 38-45 “For instance, if a trigram hashing technique is used to produce the reduced-dimension vector, then the dimensionality layer 804 means that there are 30K unique trigrams within the original corpus of SOOK words (with the possible exclusion of some uninteresting trigrams). The layer 806 and the layer 808 each have a dimensionality of 300 elements. And the layer 810 has a dimensionality of 128 elements.”).
With regard to claim 16 the proposed combination further teaches wherein the generating the first and third vector embeddings is conducted independently of generating the second vector embedding (Gou, Column 11, line 66 “The first DRM 702 produces a first reduced-dimension vector, while the second DRM 706 produces a second reduced-dimension vector.”).
With regard to claim 17 the proposed combination further teaches wherein the generating the first vector embedding representative of the search query as query concept vector yQ (Gou, Column 10, lines 35-36 “The first transformation module 504 uses a first instance of the model 106 to project the query information into a query concept vector yQ in a semantic space.”) and the generating the third vector embedding as context query vector yC (Goa, Column 10, lines 40-43 “The second transformation module 506 uses a second instance of the model 106 to transform the context information into a context query vector yc·”) representative of the at least one phrase candidate as part of the document (Gou, Column 18, lines 29-30 “the document D can have multiple parts.”) comprises applying an ML model (Column 6, lines 59-60 “the first transformation module 202 uses a first instance of the model 106”; Colum 7 lines 26-27 “the second transformation module 204 may use a second instance of the model 106”) that has been trained (Gou, column 14, lines 45-49 “A.3. The Training System FIG. 10 shows one implementation of the training system 104 of FIG. 1. In one illustrative and non-limiting case, the training system 104 processes a corpus of click-through data (provided in a data store 108), to generate the model 106.”) to generate vector embeddings (Gou, Column 10, lines 40-43) of input phrases (Gou, Column 18, lines 29-30).
With regard to claim 18 the proposed combination further teaches training the ML model by:
generating a training set of data comprising a plurality of training digital objects(Gou, column 14, lines 45-49 “A.3. The Training System FIG. 10 shows one implementation of the training system 104 of FIG. 1. In one illustrative and non-limiting case, the training system 104 processes a corpus of click-through data (provided in a data store 108), to generate the model 106.”), a given one of which comprises: (i) a training search query (Gou, Column 15, lines 1-4 “Each example includes a context (C) associated with a particular submitted query (Q), a document (D+) that the user selected in response to the query (and its associated context)”); and (ii) a plurality of training phrase candidates (Gou, Column 18, lines 29-30), identified in a respective plurality of training digital documents responsive to the training search query (Gou, Column 15, lines 1-4 “Each example includes a context (C) associated with a particular submitted query (Q), a document (D+) that the user selected in response to the query (and its associated context)”), that are lexically as lexical features (Rudzicz, ¶49; ¶55) related to at least one term of the training search query as responsive to the query (Gou, Column 15, lines 1-4); and
feeding the plurality of training digital objects to the ML model (Gou, Column 14, lines 46-49 “FIG. 10 shows one implementation of the training system 104 of FIG. 1. In one illustrative and non-limiting case, the training system 104 processes a corpus of click-through data (provided in a data store 108), to generate the model 106.”).
With regard to claim 19 the proposed combination further teaches during the identifying the at least one phrase candidate (Gou, Column 18, lines 29-30) that is lexically as lexical features (Rudzicz, ¶49; ¶55) related to at least one term of the search query for the given one of the plurality of digital document candidates (Column 18 “that is configured to compare the text that surrounds a user's query (in a source document) with the title of each candidate document”), the method further comprises using an unprocessed version of the given one of the plurality of digital document candidates (Gou, Column 15, lines 1-4 “Each example includes a context (C) associated with a particular submitted query (Q), a document (D+) that the user selected in response to the query (and its associated context)”).
With regard to claim 20 Gou teaches A server for ranking (Guo, Abstract “The search engine operates by ranking a plurality of documents”) digital documents (Id) at a digital platform (Guo, Fig. 1, 120 “Ranking Framework”), the server comprising at least processor (Gou, Column 11, lines 38-41 “The remote processing system 608 can be implemented as one or more server computing devices in conjunction with one or more data stores.”) and at least one non-transitory computer-readable memory comprising executable instructions (Id), which, when executed by the at least one processor, cause the server to:
receive a search query as the query (Guo, Column 10, lines 33-35 “Here, a first transformation module 504 receives query information associated with the query inputted by the user”) submitted by a user as the user (Id) to the digital platform as 504 (Id) is part of the ranking framework 120 (Guo, Column 10, lines 32-33 “FIG. 5 shows another implementation 502 of the ranking framework 120.”);
generate a first vector embedding representative of the search query as query concept vector yQ (Gou, Column 10, lines 35-36 “The first transformation module 504 uses a first instance of the model 106 to project the query information into a query concept vector yQ in a semantic space.”);
identify, in a search index of the digital platform, a plurality of digital document candidates that are responsive to the search query as the set of k documents (Gou, Column 10 lines 22-25 “In that variation, the set of k documents in the data store 416 correspond to the set of k entities which match the user's query, without taking into consideration the context information.”; Gou, Column 4, lines 45-48 “The search engine 112 can optionally rely on an indexing mechanism (not shown) to retrieve documents, given specified search terms.”);
retrieve, for each digital document candidate of the plurality of digital document candidates as the set of k documents (Gou, Column 10 lines 22-25 “In that variation, the set of k documents in the data store 416 correspond to the set of k entities which match the user's query, without taking into consideration the context information.”), a second vector embedding representative thereof as document concept vector yD(Gou, Column 10, lines 42-48 “A third transformation module 508 receives document information which describes a candidate document, among a collection of such documents to be considered, stored in a data store 510. The third transformation module 508 uses a third instance of the model 106 to transform the document information into a document concept vector, y D ·”), the second vector having been generated prior to the receiving of the search query (Gou, Column 19, lines 25-29 “Alternatively, each document concept vector can be computed in advance as part of an offline process, and then stored along with the document to which it pertains, and properly indexed to allow for later retrieval.”);
identify, for a given one of the plurality of digital document candidates as the set of k documents (Gou, Column 10 lines 22-25 “In that variation, the set of k documents in the data store 416 correspond to the set of k entities which match the user's query, without taking into consideration the context information.”), at least one phrase candidate as part of the document (Gou, Column 18, lines 29-30 “the document D can have multiple parts.”) that is [[ as the context of the query term in the document (Column 18 “that is configured to compare the text that surrounds a user's query (in a source document) with the title of each candidate document”);
generate a third vector embedding representative of the at least one phrase candidate as context query vector yC (Goa, Column 10, lines 40-43 “The second transformation module 506 uses a second instance of the model 106 to transform the context information into a context query vector yc·”) ;
based on the first, second, and third vector embeddings (Gou, Column 10, lines 58-60 “FIG. 5, alternatively, or in addition, a comparison module can also form a relevance measure based on any joint analysis of all three concept vectors, yQ, yD , and yc·”), determine, for the given one of the plurality of digital document candidates (Gou, Column 10, lines 60-64 “A ranking module 518 can then rank the plurality of candidate documents based on a plurality of features, including the relevance measures, for each document, fed to it by the comparison module 512.”), a respective value of a ranking parameter as the ranking module ranking (Id),
the ranking parameter being indicative of relevancy of the given of the plurality of digital document candidates to the search query as ranking based on the relevance measure (Id); and
rank the plurality of digital document candidates based on respective values of the ranking parameter associated therewith as to rank (Id).
Gou does not explicitly teach that the context of the query term is specifically a lexically relation.
Rudzicz teaches identify, for a given one of the plurality of digital document candidates as the documents (Rudzicz, ¶49 “The NB and SVM classifiers make use of lexical features that encode the probability of various unigrams and bigrams occurring in documents of a specific class”), at least one phrase candidate as unigrams and bigrams (Id) that is lexically related as lexical features (Id; ¶55 “(f) Cosine distance between pairs of vectorized sentences within a document”) to at least one term as between pairs of sentences (Id);
It would have been obvious to one of ordinary skill to which said subject matter pertains at the time the invention was filed to have implemented the context analysis performed by Guo using the text classifiers taught by Rudzicz as it yields the predictable results of generating embeddings representing the text (Rudzicz, ¶48) and classifying the text (Rudzicz, ¶46). One of ordinary skill in the art would recognize that the language analytics taught by Rudzicz may be used to identify context of the terms within the document (Rudzicz, ¶55; Gou, Column 18 lines 29-30) which may be used to perform query disambiguation (Rudzicz, ¶2; Gou, Column 1, lines 5-25).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Zhang [2017/0228361 teaches generating context-dependent vectors which encode both semantic and lexical information into the encoding.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AMANDA WILLIS whose telephone number is (571)270-7691. The examiner can normally be reached Monday-Friday 8am-2pm.
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, Ajay Bhatia can be reached at 571-272-3906. 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.
/AMANDA L WILLIS/ Primary Examiner, Art Unit 2156