DETAILED ACTION
This communication is in response to the Applicant Arguments/Remarks dated 7/1/2025. Claims 1-5, 7-14, 16-18, 20-23 are pending in the application.
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 . 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.
Response to Arguments
Applicant's arguments filed 7/1/2025 have been fully considered.
Specification, para. 5: the LAT in this example is "capital," and this lexical form describes the category or type that should be exhibited by any answer satisfying this request; para. 36 teaches knowledge bases and other word network including lexical databases, semantic web content databases etc.
In the technological art, the limitation “lexical answer type query expansion” is the process of improving information retrieval by adding new terms to a user's original query. This can involve adding synonyms, hypernyms (broader terms), or hyponyms (more specific terms). And the limitation “word network” is a structured lexical database that represents relationships between words, such as WordNet or ConceptNet. It organizes words into sets of synonyms (synsets) and links them through various semantic relationships.
Regarding the arguments on pages 8-9 that “Copperman does not weight individual lexical answer type query expansions, but, according to para. 159, "After searching over the indexes, ranking is employed to merge knowledge container lists returned by the search stage to produce a single list ordered by relevance." Copperman discloses a step of query expansion, followed by a search and "return by the search stage," then "ranking is employed to merge knowledge container lists." This operation is quite distinct from the claimed processing of individual lexical answer type query expansions based on relevance to the context and subsequent using the highest or higher ranking individual lexical answer type query expansion(s) in the search stage.
Liddy discloses the presence of mandatory terms in a document, which are given greater weight than others, the documents containing the mandatory terms are identified and preferably segregated from documents not containing the mandatory terms. Where mandatory terms are match, the document contents are assumed to make the CN better when referenced., examiner respectfully disagrees.
The Liddy reference is no longer applied in this current Office action. The newly cited reference Zhou et al. has been applied to further teaches the amended limitations including “generating an expanded structured query that is optimized to include one or more top weighted individual lexical answer type query expansions”.
Copperman teaches in fig. 3: taxonomy tags are ordered by weight and grouped by taxonomy; para. 8-9: in a query-based retrieval, a user specifies a natural language query with one or more taxonomy tags, one or more taxonomic restrictions, and any knowledge container restrictions deemed necessary; para. 45: creating new types of knowledge containers, which represent new types of content or resources, by creating and augmenting subtypes of the existing types; para. 84: a set of transformational inference rules can be applied to refine the taxonomy tags produced by the previous steps. These rules are conditional on taxonomy tags, entity and technical term tags, and potentially other aspects of the content, and can either adjust the weights (confidence measure) of taxonomy tags, remove taxonomy tags, or add new taxonomy tags to the content; para. 53: the result is a list of taxonomies with associated taxonomy tags, ordered by the weight of the highest-weighted tag associated with that taxonomy; para. 59-61: lexical taxonomies are useful for identifying and grouping concepts that occur using specific words and phrases within knowledge containers. For example, using a lexical taxonomy of companies organized hierarchically by industry type, in conjunction with a topic taxonomy of legal issues; para. 113-114: if content is tagged to Government Agencies: Federal: Executive: IRS with weight above 0.60 and content is tagged to any node under Government Agencies: Government Issues: Legislation with weight X where X is greater than 0.35, add tag Government Issues: Legislation: Tax Legislation to the content with weight X. Finally, the system stores the results as a knowledge container in its data store.
Please see also the newly cited reference, with columns and lines below.
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.
Claim(s) 1-5, 7-14, 16-18, 20-23 are rejected under 35 U.S.C. 103 as being unpatentable over Copperman et al. (US 20070033221) in view of Zhou et al. (US 20140006012) and further in view of Shear (US 20160034305).
Specification, para. 5: the LAT is conventionally used in the context of question-answering, especially simple fact use-cases, where the user's input is in the form of a natural language question, such as, "What is the capital of lndia?" The LAT in this example is "capital," and this lexical form describes the category or type that should be exhibited by any answer satisfying this request;
para. 37 teaches “LAT-based expansions may be weighted for contextual relevance using context-sensitive weighting, such as with knowledge graph activation”.
As per claims 1, 11, 20, Copperman et al. teaches
a computer-implemented method for optimizing expansion of user query input in natural language processing applications (para. 9, 154: the text of the query may be augmented or expanded. This query expansion may be based upon a thesaurus, to include synonyms or other related terms in the text; para. 173: suppose a user asks a question/query about Jaguars. Auto-contextualization may produce tags related to both automobiles and animals, and these may be expanded by the retrieval process into different regions. The system may determine based on the taxonomic structure that these are likely to be mutually exclusive regions. Thus, the user may be presented with the question "Is your question more relevant to automobiles or to animals?" Just as for taxonomy selection, the user's responses to this type of question are added to the taxonomic restrictions of the user's question, resulting in a more precise response in the next round of the dialog);
the computer-implemented method comprising: identifying a lexical answer type for a user query input based on a corresponding domain of the user query input (para. 58-61: lexical taxonomies are useful for identifying and grouping concepts that occur using specific words and phrases within knowledge containers. E.g., a query: "Show documents which (a) mention software companies and (b) talk about intellectual property protection." Here, (a) would be fulfilled by limiting the search to knowledge containers tagged to any concept-node under the "software companies" concept-node of a lexical "Companies" taxonomy (e.g., knowledge containers that mention IBM, Microsoft, etc.); and (b) would be fulfilled by looking at or near the topic of "intellectual property protection" in the legal issues topic taxonomy. Semantically, concept nodes in each taxonomy represent classifications in a single "dimension" or area of concern. For example, one taxonomy might represent a company's complete product line, and another might represent geography-different parts of the world; para. 155: performs region designation to identify additional areas of the taxonomy to improve the results of the query; para. 175: Parameterized Questions (PQs): the system may have additional information about specific types of clarifying questions that are useful in the domain);
leveraging a word network to constrain lexical answer type query expansions to target an ambiguous term of the user query input, based on context of the user query input (para. 9: aid a researcher or user in quickly identifying relevant documents, in response to an inputted query, retrieving documents through the use of taxonomies and knowledge containers seeks to identify matches between the query and the concept nodes in a taxonomy, to provide a faster and more relevant response than a content-based retrieval, which is driven by the actual words in the document; para. 154: the text of the query may be augmented or expanded. This query expansion may be based upon a thesaurus, to include synonyms or other related terms in the text; para. 155-156: the imperfection can be ameliorated by augmenting the query taxonomy tags, which results in augmenting the set of knowledge containers that are considered by the subsequent search stage. A disjunctive interpretation is generally appropriate for a lexically ambiguous query that is tagged to one concept node because of some query term, and is tagged to another concept node. The term "jaguar" occurring in a query, for example, may result in query taxonomy tags to concept nodes "Jungle Cat" and "Automobile", but the query is about one or the other, not both);
feeding the lexical answer type into a query expansion method to expand the user query input with one or more precise terms based on the lexical answer type; wherein the query expansion method uses the word network to generate a plurality of lexical answer type query expansions; weighting the lexical answer type query expansions based on relevance to a context of the user query input (fig. 3: taxonomy tags are ordered by weight and grouped by taxonomy; para. 53: the result is a list of taxonomies with associated taxonomy tags, ordered by the weight of the highest-weighted tag associated with that taxonomy; para. 8, 59-61: lexical taxonomies are useful for identifying and grouping concepts that occur using specific words and phrases within knowledge containers. For example, using a lexical taxonomy of companies organized hierarchically by industry type, in conjunction with a topic taxonomy of legal issues; para. 113-114: if content is tagged to Government Agencies: Federal: Executive: IRS with weight above 0.60 and content is tagged to any node under Government Agencies: Government Issues: Legislation with weight X where X is greater than 0.35, add tag Government Issues: Legislation: Tax Legislation to the content with weight X. Finally, the system stores the results as a knowledge container in its data store; para. 146-148: retrieving an appropriate answer from a corporate knowledge base of populated taxonomies in response to a query from a customer or from a knowledge worker; para. 155: the text of the query may be augmented or expanded. This query expansion may be based upon a thesaurus, to include synonyms or other related terms in the text; para. 159: the resulting rank of knowledge container represents the knowledge container's relevance to the query; para. 165: taxonomy tags and weights (perhaps segmented for ease of entry; e.g. "Very relevant", "Some what relevant", "Not relevant") to be associated with the question; para. 167: a) a list of result knowledge containers that are possible "answers" to the question, each with a relevance score between 0 and 1; para. 173: resulting in a more precise response in the next round of the dialog);
generating an expanded structured query that is optimized to include one or more top weighted individual lexical answer type query expansions; and outputting the expanded structured query for use as input into a natural language processing application (fig. 3: taxonomy tags are ordered by weight and grouped by taxonomy; para. 9: in a query-based retrieval, a user specifies a natural language query with one or more taxonomy tags, one or more taxonomic restrictions, and any knowledge container restrictions deemed necessary; para. 45: creating new types of knowledge containers, which represent new types of content or resources, by creating and augmenting subtypes of the existing types; para. 84: a set of transformational inference rules can be applied to refine the taxonomy tags produced by the previous steps. These rules are conditional on taxonomy tags, entity and technical term tags, and potentially other aspects of the content, and can either adjust the weights (confidence measure) of taxonomy tags, remove taxonomy tags, or add new taxonomy tags to the content – See para. 111-113; para. 154-155: the text of the query may be augmented or expanded. This query expansion may be based upon a thesaurus, to include synonyms or other related terms in the text. The query undergoes at least some of the stages of auto-contextualization as described above. The system now performs region designation to identify additional areas of the taxonomy to improve the results of the query; para. 161: users type a question, perhaps augmented by initial taxonomy tags, interest taxonomy tags, and/or taxonomic restrictions (filters), and a single list of knowledge containers is returned; para. 148: the user's inputs are then passed to the query-based retrieval system for resolution. Query-based Retrieval includes five stages: preparation; auto-contextualization of query; region designation; search; and ranking …).
Even if Copperman does not explicitly disclose wherein the query expansion method uses the word network, generating an expanded structured query that is optimized to include one or more top weighted individual lexical answer type query expansions.
Zhou teaches
wherein the query expansion method uses the word network to generate a plurality of lexical answer type query expansions to generate a plurality of lexical answer type query expansions (para. 3-4: on receiving a natural language question entered by a user, an analysis is performed to determine a question type, answer type, and/or lexical answer type (LAT) for the question. The extracted query units, answer type, question type, and/or LAT may then be applied to one or more query generation templates to generate a plurality of queries to be used to gather evidence to determine the answer to the natural language question. The queries may then be ranked using a ranker that is trained offline using machine learning, and the top N ranked queries may be sent to a search engine. Results (e.g., addresses and/or snippets of web documents) may then be filtered and/or ranked using another machine learning trained ranker, and candidate answers are extracted from the results based on the answer type and/or LAT; para. 79-83: LAT context, the nearest N number of words before and after the LAT in the natural language question (e.g., N=3); [0080] Title tag, whether the LAT is contained in a title dictionary (e.g., as in an external knowledge base 212, or commercial available online dictionary such as WordNet.RTM); [0081] Synonym words of the LAT, e.g. as determined through a dictionary; [0082] Hypernym words of the LAT, e.g. as determined through a dictionary; and/or [0083] Specific unigram, e.g. whether the question includes particular words such as where, who, what, etc.);
weighting individual lexical answer type query expansions based on relevance to a context of the user query input; generating an expanded structured query that is optimized to include one or more top weighted individual lexical answer type query expansions (para. 19: the information gained from the Question Understanding phase may be used to generate one or more search queries for gathering evidence to determine an answer to the natural language question. The extracted query units as well as the question type, answer type, and/or LAT are applied to one or more query generation templates to generate a set of candidate queries. The candidate queries may be ranked using a ranker trained offline using an unsupervised or supervised machine learning technique such as support vector machine; para. 22: the top-ranked candidate answer is provided as the answer to the user's question when the top-ranked candidate answer has a confidence level that exceeds a predetermined threshold confidence level. In the example of FIG. 1, the answer 118 is Franz Schubert with a confidence level of 0.85).
Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Copperman et al. and Zhou to effectively analyze and improve the initial query to additional terms that are highly relevant to the query and the user which help retrieve best possible search results for the users and thus, improve performance.
Even if Copperman and Zhou do not explicitly teach the term leveraging the word network,
Shear teaches
leveraging a word network to constrain lexical answer type query expansions to target an ambiguous term of the user query input based on context of the user query input (para. 4939, 5480: for example, for disambiguation, it may leverage WordNet® (a trademark of Princeton University), which is a large English lexical database. WordNet groups nouns, verbs, adjectives and adverbs into sets of cognitive synonyms (synsets), each expressing a distinct concept. Synsets are interlinked by methods of conceptual-semantic and lexical relations. The resulting network of meaningfully related words and concepts may be navigated with a browser. Word Net's structure makes it a useful tool for computational linguistics and natural language processing; para. 3765, 5348: detect natural language words and phrases that may be ambiguous or otherwise unclear.) Shear also teaches query expansion: para. 2268, 2811).
Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Copperman, Zhou with the teaching of Shear to effectively utilize knowledge graphs/word networks to analyze and improve the initial query to additional terms that are highly relevant to the query which help retrieve best possible search results for the users and thus, improve performance.
As per claims 2, 12, Copperman et al. teaches
wherein identifying a lexical answer type for the user query input is based on a domain-appropriate lexical answer type extraction for the corresponding domain of the user query input (para. 8-9, 63: the purpose of auto-contextualization is to provide a mechanism for transforming a document (e.g., a document created by a word processor, or an e-mail) into a structured record and to automatically (without human review) construct indexes usable by a content-based retrieval engine to help identify when the structured record is an appropriate response to a particular query; para. 81: creates the taxonomy tags appropriate to the content of a knowledge container for taxonomies of the "topic taxonomy" type described above. Based on the entities, technical terms, and other words contained in the content, a text classifier is employed to identify concept nodes from a topic taxonomy. Each knowledge-container /concept-node association comprises a taxonomy tag; para. 119, 173).
para. 58-61: lexical taxonomies are useful for identifying and grouping concepts that occur using specific words and phrases within knowledge containers. E.g., a query: "Show documents which (a) mention software companies and (b) talk about intellectual property protection." Here, (a) would be fulfilled by limiting the search to knowledge containers tagged to any concept-node under the "software companies" concept-node of a lexical "Companies" taxonomy (e.g., knowledge containers that mention IBM, Microsoft, etc.); and (b) would be fulfilled by looking at or near the topic of "intellectual property protection" in the legal issues topic taxonomy. Semantically, concept nodes in each taxonomy represent classifications in a single "dimension" or area of concern. For example, one taxonomy might represent a company's complete product line, and another might represent geography-different parts of the world; para. 119,155: performs region designation to identify additional areas of the taxonomy to improve the results of the query; para. 173,175: Parameterized Questions (PQs): the system may have additional information about specific types of clarifying questions that are useful in the domain).
Zhou also teaches at para. 92-93: the rank of the result page within the set of results generated from the search query, as ranked by the search engine; the domain of snippet of the result, e.g. a quality of the domain; para. 106: if the natural language question has a predicted answer type of "person," the "person" type named entities are extracted from the search results. At 804 the extracted named entities are normalized to expand contractions, correct spelling errors in the search results, expand proper names (e.g., Bill to William), and so forth.
As per claims 3, 13, Copperman et al. teaches
ranking the plurality of lexical answer type query expansions by evaluating the individual lexical answer type query expansions based on relevance to the context of the user query input; and selecting the one or more top weighted individual lexical answer type query expansions based on the ranking (fig. 19: the knowledge containers ordered by their adjusted ranks; para. 53, 155: performs region designation to identify additional areas of the taxonomy to improve the results of the query. The imperfection can be ameliorated by augmenting the query taxonomy tags, which results in augmenting the set of knowledge containers that are considered by the subsequent search stage; para. 158-160: the resulting rank of knowledge container represents the knowledge container's relevance to the query).
Even if Copperman does not explicitly teach selecting the one or more top weighted individual lexical answer type query expansions based on the ranking,
Zhou teaches at para. 19: the information gained from the Question Understanding phase may be used to generate one or more search queries for gathering evidence to determine an answer to the natural language question. The extracted query units as well as the question type, answer type, and/or LAT are applied to one or more query generation templates to generate a set of candidate queries. The candidate queries may be ranked using a ranker trained offline using an unsupervised or supervised machine learning technique such as support vector machine; para. 22: the top-ranked candidate answer is provided as the answer to the user's question when the top-ranked candidate answer has a confidence level that exceeds a predetermined threshold confidence level. In the example of FIG. 1, the answer 118 is Franz Schubert with a confidence level of 0.85).
Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Copperman et al. and Zhou to effectively analyze and improve the initial query to additional terms that are highly relevant to the query and the user which help retrieve best possible search results for the users and thus, improve performance.
As per claims 4, 14, Copperman et al. teaches
ranking the plurality of lexical answer type query expansions comprises: using one or more semantic concepts, including the lexical answer type, from the user query input as inputs to one or more active knowledge graphs of a query expansion method, with concepts associated with the lexical answer type amplified as inputs; and sorting individual lexical answer type query expansions in a final activated sub-graph by relevance to the user query input (para. 58-61: in lexical taxonomies, a knowledge container is tagged to a concept-node based on a simple lexical rule that matches against the content of the knowledge container… "software companies" concept-node of a lexical "Companies" taxonomy (e.g., knowledge containers that mention IBM, Microsoft, etc.); and (b) would be fulfilled by looking at or near the topic of "intellectual property protection" in the legal issues topic taxonomy; Semantically, concept nodes in each taxonomy represent classifications in a single "dimension" or area of concern. For example, one taxonomy might represent a company's complete product line, and another might represent geography-different parts of the world; para. 112-114, 147: interest taxonomy tags support personalization; it may be appreciated that an individual's interest profile affects the presentation of results of all of the user's information requests; para. 154, 159: ranking is employed to merge knowledge container lists returned by the search stage to produce a single list ordered by relevance; fig. 20).
Zhou also teaches using one or more semantic concepts, including the lexical answer type, from the user query input as inputs to one or more active knowledge graphs of a query expansion method, with concepts associated with the lexical answer type amplified as inputs; and sorting individual lexical answer type query expansions in a final activated sub-graph by relevance to the user query input at para. 3-4: on receiving a natural language question entered by a user, an analysis is performed to determine a question type, answer type, and/or lexical answer type (LAT) for the question; para. 85: the extracted query units may also include attributes of the natural language question found in the at least one knowledge base. Extraction of query units may include one or more of the following: sentence boundary detection 518, sentence pattern detection 520, parsing 522, named entity detection 524, part-of-speech tagging 526, tokenization 528, and chunking; para. 114-115, fig. 1: ranked search queries, ranked candidate answers.
As per claim 5, Copperman et al. teaches
wherein sorting individual lexical answer type query expansions includes using one or more techniques selected from the group consisting of: degree of separation from lexical answer type concepts; degree of separation from a most highly activated concept; and overall activation weight (para. 53: the process of creating a smart summary begins as follows: in step 100, taxonomy tags are grouped by taxonomy and then ordered by weight. The result is a list of taxonomies with associated taxonomy tags, ordered by the weight of the highest-weighted tag associated with that taxonomy; para. 120: the system then performs term separation. Terms are presented to a subject matter expert (SME) highly familiar with the knowledge domain associated with the generation corpus. The SME designates whether the term is relevant to each of the taxonomies in the input set. the term "jaguar" may be relevant to the taxonomy on "Mammals" and the taxonomy on "Automobiles". The result of this step is N lists of terms where N is equal to the number of root concept-nodes. In one embodiment, the SME generates a set of terms a priori, from his or her knowledge of the domain, for each root concept node; fig. 3: taxonomy tags are ordered by weight and grouped by taxonomy, if taxonomy weight, tag above threshold, emit the high confidence summary phrase associated with the concept node and the tag; para. 148: ranks the results by combining the ordered lists into a single list. The final result of executing these five stages is a single ordered list of knowledge containers; para. 158-160).
Zhou also teaches using one or more semantic concepts, including the lexical answer type, from the user query input as inputs to one or more active knowledge graphs of a query expansion method, with concepts associated with the lexical answer type amplified as inputs; and sorting individual lexical answer type query expansions in a final activated sub-graph by relevance to the user query input at para. 3-4: on receiving a natural language question entered by a user, an analysis is performed to determine a question type, answer type, and/or lexical answer type (LAT) for the question; para. 85: the extracted query units may also include attributes of the natural language question found in the at least one knowledge base. Extraction of query units may include one or more of the following: sentence boundary detection 518, sentence pattern detection 520, parsing 522, named entity detection 524, part-of-speech tagging 526, tokenization 528, and chunking; para. 114-115, fig. 1: ranked search queries, ranked candidate answers.
As per claims 7, 16, Copperman et al. teaches
wherein generating the expanded structured query includes: automatically selecting a number of top weighted individual lexical answer type query expansions (para. 8: documents stored in the organization and retrieval subsystem may be automatically classified into a predetermined number of taxonomies through a process called auto-contextualization; para. 158: it can be determined if there is a knowledge container that happens to not be in any of the smaller indexes searched, but which has a very good content match to the query. The result of this step is a ranked list of knowledge containers; para. 160: the taxonomic distance from the knowledge container's taxonomy tags to the user's interest taxonomy tags is a measure of a knowledge container's relevance to the user's interests. Upon completion of the ranking step, a ranked list of knowledge containers is presented to the user. This completes an instance of retrieving an appropriate answer from a corporate knowledge base of populated taxonomies in response to a query).
Even if Copperman does not explicitly teach top weighted individual lexical answer type query expansions,
Zhou teaches said limitation at para. 19: the information gained from the Question Understanding phase may be used to generate one or more search queries for gathering evidence to determine an answer to the natural language question. The extracted query units as well as the question type, answer type, and/or LAT are applied to one or more query generation templates to generate a set of candidate queries. The candidate queries may be ranked using a ranker trained offline using an unsupervised or supervised machine learning technique such as support vector machine; para. 22: the top-ranked candidate answer is provided as the answer to the user's question when the top-ranked candidate answer has a confidence level that exceeds a predetermined threshold confidence level. In the example of FIG. 1, the answer 118 is Franz Schubert with a confidence level of 0.85).
Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Copperman et al. and Zhou to effectively analyze and improve the initial query to additional terms that are highly relevant to the query and the user which help retrieve best possible search results for the users and thus, improve performance.
As per claims 8, 17, Copperman et al. teaches
wherein the one or more top weighted individual lexical answer type query expansions are selected by: receiving a user selection of a number of weighted individual lexical answer type query expansions; and feeding back the user selection to enhance future expansions and context weightings (figs. 20-21: the user can choose among the taxonomies, clusters for query refinement; para. 58-60: lexical taxonomies are useful for identifying and grouping concepts that occur using specific words and phrases within knowledge containers. For example, using a lexical taxonomy of companies organized hierarchically by industry type, in conjunction with a topic taxonomy of legal issues; fig. 3: taxonomy tags are ordered by weight and grouped by taxonomy; para. 159: the number of regions into which a knowledge container is tagged, the proportion of a knowledge container's taxonomy tags that are within designated regions, and the level of previous user satisfaction with the knowledge container (based upon implicit or explicit user feedback from previous queries).)
Even if Copperman does not explicitly teach the one or more top weighted individual lexical answer type query expansions are selected by: receiving a user selection of a number of weighted individual lexical answer type query expansions,
Zhou teaches said limitation at para. 19: the information gained from the Question Understanding phase may be used to generate one or more search queries for gathering evidence to determine an answer to the natural language question. The extracted query units as well as the question type, answer type, and/or LAT are applied to one or more query generation templates to generate a set of candidate queries. The candidate queries may be ranked using a ranker trained offline using an unsupervised or supervised machine learning technique such as support vector machine; para. 22: the top-ranked candidate answer is provided as the answer to the user's question when the top-ranked candidate answer has a confidence level that exceeds a predetermined threshold confidence level. In the example of FIG. 1, the answer 118 is Franz Schubert with a confidence level of 0.85; para. 89, 105).
Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Copperman et al. and Zhou to effectively analyze and improve the initial query to additional terms that are highly relevant to the query and the user which help retrieve best possible search results for the users and thus, improve performance.
As per claims 9, 18, Copperman et al. teaches
wherein feeding the lexical answer type into the query expansion method comprises: using a synonym lookup based on a categorical nature of the lexical answer type (para. 118: the root concept-nodes are synonyms, and taxonomies generated from them would cover substantially the same portion and aspect of the knowledge domain; para. 154: the text of the query may be augmented or expanded. This query expansion may be based upon a thesaurus, to include synonyms or other related terms in the text.)
Zhou teaches using a synonym lookup based on a categorical nature of the lexical answer type at para. 18: Question Understanding may also include the extraction of query units from the natural language question. Query units may include one or more of the following: words, base noun-phrases, sentences, named entities, quotations, paraphrases (e.g., reformulations based on synonyms, hypernyms, and the like; para. 79-83: LAT context, the nearest N number of words before and after the LAT in the natural language question (e.g., N=3); Title tag, whether the LAT is contained in a title dictionary (e.g., as in an external knowledge base 212, or commercial available online dictionary such as WordNet.RTM); Synonym words of the LAT, e.g. as determined through a dictionary; Hypernym words of the LAT, e.g. as determined through a dictionary; and/or Specific unigram, e.g. whether the question includes particular words such as where, who, what, etc.);
As per claims 10, 19, Copperman et al. teaches
using the lexical answer type for post-search filtering of results (para. 139: the subject matter expert classifies the taxonomy as either a topic, filter or lexical taxonomy-meaning that the either a search engine will be invoked on indexes built from them or the taxonomy will be used as a filter on retrieval; the user or application screen may specify a taxonomic restriction or filter to limit the knowledge containers that are presented to the user. The taxonomic restriction in turn, specifies a set of concept nodes using Boolean expressions and taxonomic relationships among the selected nodes. In the end, only knowledge containers tagged to a set of nodes that satisfy the relationships are presented to the user; para. 155: performs region designation to identify additional areas of the taxonomy to improve the results of the query.)
As per claims 21, Copperman et al. teaches
ranking the weighted individual lexical answer type query expansions (para. 8, 59-61: lexical taxonomies are useful for identifying and grouping concepts that occur using specific words and phrases within knowledge containers. For example, using a lexical taxonomy of companies organized hierarchically by industry type, in conjunction with a topic taxonomy of legal issues; para. 113-114: if content is tagged to Government Agencies: Federal: Executive: IRS with weight above 0.60 and content is tagged to any node under Government Agencies: Government Issues: Legislation with weight X where X is greater than 0.35, add tag Government Issues: Legislation: Tax Legislation to the content with weight X. Finally, the system stores the results as a knowledge container in its data store; para. 146-148: retrieving an appropriate answer from a corporate knowledge base of populated taxonomies in response to a query from a customer or from a knowledge worker; para. 155: the text of the query may be augmented or expanded. This query expansion may be based upon a thesaurus, to include synonyms or other related terms in the text; para. 159: the resulting rank of knowledge container represents the knowledge container's relevance to the query).
Zhou also teach said limitation at para. 3-4: on receiving a natural language question entered by a user, an analysis is performed to determine a question type, answer type, and/or lexical answer type (LAT) for the question. The extracted query units, answer type, question type, and/or LAT may then be applied to one or more query generation templates to generate a plurality of queries to be used to gather evidence to determine the answer to the natural language question. The queries may then be ranked using a ranker that is trained offline using machine learning, and the top N ranked queries may be sent to a search engine. Results (e.g., addresses and/or snippets of web documents) may then be filtered and/or ranked using another machine learning trained ranker, and candidate answers are extracted from the results based on the answer type and/or LAT; para. 79-83: LAT context, the nearest N number of words before and after the LAT in the natural language question (e.g., N=3); [0080] Title tag, whether the LAT is contained in a title dictionary (e.g., as in an external knowledge base 212, or commercial available online dictionary such as WordNet.RTM); [0081] Synonym words of the LAT, e.g. as determined through a dictionary; [0082] Hypernym words of the LAT, e.g. as determined through a dictionary; and/or [0083] Specific unigram, e.g. whether the question includes particular words such as where, who, what, etc.
As per claim 22, Copperman et al. teaches
detecting the domain of the user query input (para. 46-47: context tags or taxonomy tags represent a multidimensional classification of the knowledge container against a knowledge map, as depicted in fig. 1. Such a classification puts the knowledge container 20 in context within a knowledge domain; para. 74: some examples of technical terms in the network computing field are "distributed computing", "local area network" and "router". Within a particular knowledge domain, however, technical terms are generally well understood by experts in the field).
As per claim 23, Copperman et al. teaches
utilizing a categorical nature of the lexical answer type to improve the query expansion method (fig. 9B: concept nodes improvement is initiated, F-Measure > .65% para. 83-84: e.g. based on the presence of entity "XYZ Corp.", add markup that indicates a mapping to the concept-node "XYZ-CORP" in a lexical "Companies" taxonomy. One piece of content may contain entities and technical terms that are mapped to concept-nodes in one or many lexical taxonomies; para. 154-155: the text of the query may be augmented or expanded. This query expansion may be based upon a thesaurus, to include synonyms or other related terms in the text. Performs region designation to identify additional areas of the taxonomy to improve the results of the query).
Zhou also teaches at para. 3-4: on receiving a natural language question entered by a user, an analysis is performed to determine a question type, answer type, and/or lexical answer type (LAT) for the question. The extracted query units, answer type, question type, and/or LAT may then be applied to one or more query generation templates to generate a plurality of queries to be used to gather evidence to determine the answer to the natural language question. The queries may then be ranked using a ranker that is trained offline using machine learning, and the top N ranked queries may be sent to a search engine. Results (e.g., addresses and/or snippets of web documents) may then be filtered and/or ranked using another machine learning trained ranker, and candidate answers are extracted from the results based on the answer type and/or LAT; para. 79-83: LAT context, the nearest N number of words before and after the LAT in the natural language question (e.g., N=3); [0080] Title tag, whether the LAT is contained in a title dictionary (e.g., as in an external knowledge base 212, or commercial available online dictionary such as WordNet.RTM); [0081] Synonym words of the LAT, e.g. as determined through a dictionary; [0082] Hypernym words of the LAT, e.g. as determined through a dictionary; and/or [0083] Specific unigram, e.g. whether the question includes particular words such as where, who, what, etc.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Duboue et al. (US 20120078902) teaches at para. 83-85, 99.
Wurzer (US11934391) teaches at claim 1: wherein a request in the technical query language for retrieving data objects and not text parts, is created out of a determined meaning without a need of analyzing a grammatical structure, sentence trees or lexical answer types.
Gruber et al. (10276170) teaches at col. 73:19-67: In one embodiment, the precedence order among criteria may be tuned with domain-specific parameters, since the way criteria interact may depend on the selection class. For example, when selecting among hotels, availability and price may be dominant constraints, whereas for restaurants, cuisine and proximity may be more important.
Sweeney (US 20130060785) teaches at para. 111: FIG. 2A illustrates a small complex KR 200 (in this example, a taxonomy) that may be input to analysis engine 150, e.g., by a user or a software application using system 100. Complex KR 200 includes a set of concepts linked by various hierarchical relationships. For example, concept 210 labeled "Animal" is linked in parent-child relationships to concept 220 labeled "Pet" and concept 230 labeled "Mountain Animal".
Salokhe et al. (US 20200142978) teaches at para. 20: a knowledge graph enables such a system to generate additional keywords through attribute-to-attribute relationships.
Malik et al. (US 20190278777) teaches at para. 32: a natural language interface to ask questions of the knowledge graph; para. 180: determines a frequency and co-occurrence of each entity in each of the set of documents; para. 266: normalizations.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 LINH BLACK whose telephone number is (571)272-4106. The examiner can normally be reached 9AM-5PM EST M-F.
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, Tony Mahmoudi can be reached on 571-272-4078. 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.
/LINH BLACK/Examiner, Art Unit 2163 10/1/2025
/TONY MAHMOUDI/Supervisory Patent Examiner, Art Unit 2163
Prosecution Insights