EP2183686A2 - Identification de relations sémantiques dans un discours rapporté - Google Patents

Identification de relations sémantiques dans un discours rapporté

Info

Publication number
EP2183686A2
EP2183686A2 EP08828391A EP08828391A EP2183686A2 EP 2183686 A2 EP2183686 A2 EP 2183686A2 EP 08828391 A EP08828391 A EP 08828391A EP 08828391 A EP08828391 A EP 08828391A EP 2183686 A2 EP2183686 A2 EP 2183686A2
Authority
EP
European Patent Office
Prior art keywords
semantic
association
elements
reporting act
identified elements
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
EP08828391A
Other languages
German (de)
English (en)
Other versions
EP2183686A4 (fr
Inventor
Richard Crouch
Martin Van Den Berg
David Ahn
Olga Gurevich
Barney Pell
Livia Polanyi
Scott Prevost
Giovanni Lorenzo Thione
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhigu Holdings Ltd
Original Assignee
Microsoft Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Microsoft Corp filed Critical Microsoft Corp
Priority claimed from US12/201,675 external-priority patent/US8868562B2/en
Publication of EP2183686A2 publication Critical patent/EP2183686A2/fr
Publication of EP2183686A4 publication Critical patent/EP2183686A4/fr
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/205—Parsing
    • G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars

Definitions

  • Online search engines have become an increasingly important tool for conducting research or navigating documents accessible via the Internet. Often, the online search engines perform a matching process for detecting possible documents, or text within those documents, that utilizes a query submitted by a user. Initially, the matching process, offered by conventional online search engines such as those maintained by Google or Yahoo, allow the user to specify one or more keywords in the query to describe information that s/he is looking for. Next, the conventional online search engine proceeds to find all documents that contain exact matches of the keywords, although these documents typically do not provide relevant or meaningful results in response to the query.
  • Embodiments of the present invention relate to computer-implemented methods and computer-readable media for developing associations between various words found in content of documents retrieved from the web or some other repository, as well as query search terms.
  • Content that may be semantically represented may be reported speech and other attitude reports, so that the semantic representation of the content may be compared against received natural language queries to provide a user with meaningful and highly relevant results.
  • Semantic relationships such as "about" relationships, may be identified between certain elements or search terms to allow for specific word associations to be formed.
  • a semantic representation may be generated for content in a document, and a proposition may be generated for a search query, both of which allow for rapid comparison of the proposition to one or more semantic relationships to determine the most relevant search results.
  • FIG. 1 is a block diagram of an exemplary computing environment suitable for use in implementing embodiments of the present invention
  • FIG. 2 is a schematic diagram of an exemplary system architecture suitable for use in implementing embodiments of the present invention
  • FIG. 3 is a diagram of a semantic representation generated from a text portion within a document, in accordance with an embodiment of the present invention
  • FIG. 4 is a diagram of a semantic representation generated from a text portion within a document, in accordance with an embodiment of the present invention
  • FIG. 5 is a diagram of a semantic representation generated from a text portion within a document, in accordance with an embodiment of the present invention
  • FIG. 6 is a diagram of a semantic representation generated from a text portion within a document, in accordance with an embodiment of the present invention
  • FIG. 7 is a diagram of a proposition generated from a search query, in accordance with an embodiment of the present invention
  • FIG. 8 is a diagram of a semantic representation generated from a text portion within a document, the text portion comprising two sentences, in accordance with an embodiment of the present invention;
  • FIG. 9 is a flow diagram illustrating a method for developing semantic relationships between elements distilled from content of a document, in accordance with an embodiment of the present invention.
  • FIG. 10 is a flow diagram illustrating a method for, in response to receiving a query, creating associations between various terms distilled from the query to generate a proposition, in accordance with an embodiment of the present invention.
  • FIG. 11 is a flow diagram illustrating a method for developing semantic relationships between elements distilled from content of a document, in accordance with an embodiment of the present invention.
  • a computer-implemented method for developing semantic relationships between elements distilled from content of a document to generate a semantic representation of the content for indexing includes identifying a text portion of the document to be indexed and determining semantic information for a plurality of elements identified in the text portion.
  • the semantic information may include one or both of the meanings of the identified elements or grammatical and/or semantic relations between the identified elements.
  • At least one of the identified elements may be identified as a reporting act corresponding to a speech report or an attitude report.
  • the method further includes associating the identified elements so that each association of identified elements represents a certain semantic relationship based on the determined semantic information of the identified elements.
  • a computer-implemented method for, in response to receiving a natural language query, creating associations between various terms distilled from the query to generate a proposition.
  • the proposition may be used to interrogate semantic representations of content from documents stored in a semantic index to provide relevant search results.
  • the method also includes determining associated semantic information for one or more search terms found within the query.
  • a first reporting act may be determined within the query, and a semantic relationship may be formed between the first reporting act and at least one of the search terms based on the determined semantic information for that search term.
  • the created association between the first reporting act and the search term is made by way of a relational element that describes the semantic relationship.
  • a proposition that includes the formed associations may be generated and may be further compared to semantic representations to determine highly relevant search results.
  • one or more computer-readable media having computer-useable instructions embodied thereon for performing a method of developing semantic relationships between elements distilled from content of a document to generate a semantic representation of the content to be indexed.
  • the method includes identifying at least a portion of the document, or a text portion, to be indexed.
  • the text portion may then be parsed to identify elements that are to be semantically represented. Potential meanings and grammatical or semantic relations between the identified elements are determined, in addition to determining one or more levels of association within the text portion.
  • the method also includes identifying a reporting act within the text portion for each of the one or more determined levels of association so that the first reporting act may be associated with a first set of identified elements.
  • the first reporting act may be associated with a first level of association.
  • a second reporting act may be associated with a second set of identified elements, the second reporting act being associated with a second level of association.
  • a semantic representation may be generated that includes associations by way of a relational element that describes the associations between the first set of identified elements to the first reporting act and the second set of identified elements to the second reporting act.
  • computing device 100 an exemplary operating environment for implementing embodiments of the present invention is shown and designated generally as computing device 100.
  • Computing device 100 is but one example of a suitable computing environment and is not intended to suggest any limitation as to the scope of use or functionality of the invention. Neither should the computing device 100 be interpreted as having any dependency or requirement relating to any one or combination of components illustrated.
  • the invention may be described in the general context of computer code or machine-useable instructions, including computer-executable instructions such as program components, being executed by a computer or other machine, such as a personal data assistant or other handheld device.
  • program components including routines, programs, objects, components, data structures, and the like, refer to code that performs particular tasks or implements particular abstract data types.
  • Embodiments of the present invention may be practiced in a variety of system configurations, including handheld devices, consumer electronics, general-purpose computers, specialty computing devices, etc.
  • Embodiments of the invention may also be practiced in distributed computing environments where tasks are performed by remote-processing devices that are linked through a communications network.
  • computing device 100 includes a bus
  • Bus 110 that directly or indirectly couples the following devices: memory 112, one or more processors 114, one or more presentation components 116, input/output (I/O) ports 118, I/O components 120, and an illustrative power supply 122.
  • Bus 110 represents what may be one or more busses (such as an address bus, data bus, or combination thereof).
  • FIG. 1 is merely illustrative of an exemplary computing device that can be used in connection with one or more embodiments of the present invention. Distinction is not made between such categories as “workstation,” “server,” “laptop,” “handheld device,” etc., as all are contemplated to be within the scope of FIG. 1 in reference to “computer” or “computing device.”
  • Computing device 100 typically includes a variety of computer-readable media.
  • computer-readable media may comprise Random Access Memory (RAM); Read Only Memory (ROM); Electronically Erasable Programmable Read Only Memory (EEPROM); flash memory or other memory technologies; CDROM, digital versatile disks (DVDs) or other optical or holographic media; magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices; or any other medium that can be used to encode desired information and be accessed by computing device 100.
  • RAM Random Access Memory
  • ROM Read Only Memory
  • EEPROM Electronically Erasable Programmable Read Only Memory
  • flash memory or other memory technologies
  • CDROM compact discs
  • DVDs digital versatile disks
  • Memory 112 includes computer- storage media in the form of volatile and/or nonvolatile memory.
  • the memory may be removable, nonremovable, or a combination thereof.
  • Exemplary hardware devices include solid-state memory, hard drives, optical-disc drives, etc.
  • Computing device 100 includes one or more processors that read data from various entities such as memory 112 or I/O components 120.
  • Presentation component(s) 116 present data indications to a user or other device.
  • Exemplary presentation components include a display device, speaker, printing component, vibrating component, etc.
  • I/O ports 118 allow computing device 100 to be logically coupled to other devices including I/O components 120, some of which may be built in.
  • Illustrative components include a microphone, joystick, game pad, satellite dish, scanner, printer, wireless device, etc.
  • FIG. 2 a schematic diagram of an exemplary system architecture 200 suitable for use in implementing embodiments of the present invention is shown, in accordance with an embodiment of the present invention.
  • the exemplary system architecture 200 shown in FIG. 2 is merely an example of one suitable computing environment and is not intended to suggest any limitation as to the scope of use or functionality of the present invention. Neither should the exemplary system architecture 200 be interpreted as having any dependency or requirement related to any single component or combination of components illustrated therein.
  • the system architecture 200 may include a distributed computing environment, where a client device 215 is operably coupled to a natural language engine 290, which, in turn, is operably coupled to a data store 220.
  • the operable coupling refers to linking the client device 215 and the data store 220 to the natural language engine 290, and other online components through appropriate connections.
  • These connections may be wired or wireless.
  • Examples of particular wired embodiments, within the scope of the present invention include USB connections and cable connections over a network (not shown), or a bus or other channel that interconnects components within a single machine.
  • Examples of particular wireless embodiments, within the scope of the present invention include a near-range wireless network and radio-frequency technology.
  • near-range wireless network is not meant to be limiting and should be interpreted broadly to include at least the following technologies: negotiated wireless peripheral (NWP) devices; short- range wireless air interference networks (e.g., wireless personal area network (wPAN), wireless local area network (wLAN), wireless wide area network (wWAN), BluetoothTM, and the like); wireless peer-to-peer communication (e.g., Ultra Wideband); and any protocol that supports wireless communication of data between devices. Additionally, persons familiar with the field of the invention will realize that a near-range wireless network may be practiced by various data-transfer methods (e.g., satellite transmission, telecommunications network, etc.).
  • Exemplary system architecture 200 includes the client device 215 for, in part, supporting operation of the presentation device 275.
  • the presentation device e.g., a touchscreen display
  • the client device 215 may take the form of various types of computing devices.
  • the client device 215 may be a personal computing device (e.g., computing device 100 of FIG. 1), handheld device (e.g., personal digital assistant), a mobile device (e.g., laptop computer, cell phone, media player), consumer electronic device, various servers, and the like. Additionally, the computing device may comprise two or more electronic devices configured to share information therebetween.
  • a personal computing device e.g., computing device 100 of FIG. 1
  • handheld device e.g., personal digital assistant
  • a mobile device e.g., laptop computer, cell phone, media player
  • consumer electronic device e.g., various servers, and the like.
  • the computing device may comprise two or more electronic devices configured to share information therebetween.
  • the client device 215 includes, or is operably coupled to the presentation device 275, which is configured to present a UI display 295 on the presentation device 275.
  • the presentation device 275 may be configured as any display device that is capable of presenting information to a user, such as a monitor, electronic display panel, touch-screen, liquid crystal display (LCD), plasma screen, one or more light-emitting diodes (LED), incandescent bulbs, a laser, an electroluminescent light source, a chemical light, a flexible light wire, and/or fluorescent light, or any other display type, or may comprise a reflective surface upon which the visual information is projected.
  • LCD liquid crystal display
  • LED light-emitting diodes
  • presentation device 275 Although several differing configurations of the presentation device 275 have been described above, it should be understood and appreciated by those of ordinary skill in the art that various types of presentation devices that present information may be employed as the presentation device 275, and that embodiments of the present invention are not limited to those presentation devices 275 that are shown and described.
  • the UI display 295 rendered by the presentation device 275 is configured to present a web page (not shown) that is associated with natural language engine 290 and/or a content publisher.
  • the web page may reveal a search-entry area that receives a query and search results that are discovered by searching the semantic index with the query.
  • the query may be manually provided by a user at the search-entry area, or may be automatically generated by software.
  • the query may include one or more keywords that, when submitted, invokes the natural language engine 290 to identify appropriate search results that are most responsive to the keywords in a query.
  • the natural language engine 290 shown in FIG.
  • the natural language engine 290 may be a personal computer, desktop computer, laptop computer, consumer electronic device, handheld device (e.g., personal digital assistant), various remote servers (e.g., online server cloud), processing equipment, and the like. It should be noted, however, that the invention is not limited to implementation on such computing devices but may be implemented on any of a variety of different types of computing devices within the scope of embodiments of the present invention.
  • the natural language engine 290 is configured as a search engine designed for searching for information on the Internet and/or the data store 220, and for gathering search results from the information, within the scope of the search, in response to submission of the query via the client device 215.
  • the search engine includes one or more web crawlers that mine available data (e.g., newsgroups, databases, open directories, the data store 220, and the like) accessible via the Internet and build a semantic index 260 containing web addresses along with the subject matter of web pages or other documents stored in a meaningful format.
  • the search engine is operable to facilitate identifying and retrieving the search results (e.g., listing, table, ranked order of web addresses, and the like) from the semantic index that are relevant to search terms within the submitted query.
  • the search engine may be accessed by Internet users through a web-browser application disposed on the client device 215. Accordingly, the users may conduct an Internet search by submitting search terms at the search-entry area (e.g., surfaced on the UI display 295 generated by the web-browser application associated with the search engine).
  • a search may be conducted whereby a query is submitted to one or more system indexes in order to retrieve contents from a local information store, such as a user's hard-disk.
  • the data store 220 is generally configured to store information associated with online items and/or materials that have searchable content associated therewith (e.g., documents that comprise the Wikipedia website).
  • information may include, without limitation, documents, content of a web page/site, electronic materials accessible via the Internet, a local intranet, or the memory or hard-disk of the user's machine, and other typical resources available to a search engine.
  • the data store 220 may be configured to be searchable for suitable access of the stored information. In one instance, allowing for suitable access includes selecting or filtering a subset of the documents in the data store according to criteria supplied thereto.
  • the data store 220 may be searchable for one or more documents selected for processing by the natural language engine 290.
  • the natural language engine 290 is allowed to freely inspect the data store for documents that have been recently added or amended in order to update the semantic index. The process of inspection may be carried out continuously, in predefined intervals, or upon an indication that a change has occurred to one or more documents aggregated at the data store 220.
  • the information stored in the data store 220 may be configurable and may include any information within a scope of an online search. The content and volume of such information are not intended to limit the scope of embodiments of the present invention in any way.
  • the data store 220 may, in fact, be a plurality of databases, for instance, a database cluster, portions of which may reside on the client device 215, the natural language engine 290, another external computing device (not shown), and/or any combination thereof.
  • the natural language engine 290 provides a tool to assist users aspiring to explore and find information online.
  • this tool operates by applying natural language processing technology to compute the meanings of passages in sets of documents, such as documents drawn from the data store 220. These meanings are stored in the semantic index 260 that is referenced upon executing a search.
  • a query search pipeline 205 analyzes the user's query (e.g., a character string, complete words, phrases, alphanumeric compositions, symbols, or questions) and translates the query into a structural representation utilizing semantic relationships.
  • This representation referred to hereinafter as a "proposition,” may be utilized to interrogate information stored in the semantic index 260 to arrive upon relevant search results.
  • the information stored in the semantic index 260 includes representations extracted from the documents maintained at the data store 220, or any other materials encompassed within the scope of an online search.
  • This representation referred to hereinafter as a "semantic representation,” relates to the intuitive meaning of content distilled from common text and may be stored in the semantic index 260.
  • the semantic representation is derived from a semantic structure utilizing an ordered sequence of term-rewriting rules, or any other heuristics known in the relevant field.
  • the "semantic structure" is generated at an intermediate stage of an analysis pipeline by a document parsing component that converts the content of the document to the semantic structure utilizing, in part, lexical semantic grammar rules.
  • the architecture of the semantic index 260 allows for rapid comparison of the stored semantic representations against the derived propositions in order to find semantic representations that match the propositions and to retrieve documents mapped to the semantic representations that are relevant to the submitted query.
  • the natural language engine 290 can determine the meaning of a user's query requirements from the query submitted into a search interface (e.g., the search-entry area surfaced on the UI display 295), and then to sift through a large amount of information to find corresponding search results that satisfy those needs.
  • a search interface e.g., the search-entry area surfaced on the UI display 295
  • the process above may be implemented by various functional elements that carry out one or more steps for discovering relevant search results.
  • These functional elements include a query parsing component 235, a document parsing component 240, a semantic interpretation component 245, a semantic interpretation component 250, a grammar specification component 255, the semantic index 260, a matching component 265, and a ranking component 270.
  • These functional components 235, 240, 245, 250, 255, 260, 265, and 270 generally refer to individual modular software routines, and their associated hardware that are dynamically linked and ready to use with other components or devices.
  • the data store 220, the document parsing component 240, and the semantic interpretation component 250 comprise an indexing pipeline 210.
  • the indexing pipeline 210 serves to distill the semantic representations from content within documents 230 accessed at the data store 220, and to construct the semantic index 260 upon gathering the semantic representations.
  • the semantic representations may retain a mapping to the documents 230, and/or location of content within the documents 230, from which they were derived.
  • the semantic index 260 encodes the semantic representations (being derived from the semantic structures created at the document parsing component 240) generated and conveyed by the semantic interpretation component 250.
  • the document parsing component 240 and semantic interpretation component 250 may be configured as a single element that does not divide the natural language processing into two stages (i.e., LFG parsing and semantic interpretation), but instead, produces semantic representations in a single step, without having a separate stage in which semantic structures are produced.
  • the document parsing component 240 is configured to gather data that is available to the natural language engine 290. In one instance, gathering data includes inspecting the data store 220 to scan content of documents 230, or other information, stored therein. Because, the information within the data store 220 may be constantly updated, the process of gathering data may be executed at a regular interval, continuously, or upon notification that an update is made to one or more of the documents 230.
  • the document parsing component 240 Upon gathering the content from the documents 230 and other available sources, the document parsing component 240 performs various procedures to prepare the content for semantic analysis thereof. These procedures may include text extraction, entity recognition, and parsing.
  • the text extraction procedure substantially involves extracting tables, images, templates, and textual sections of data from the content of the documents 230 and to converting them from a raw online format to a usable format (e.g., HyperText Markup Language (HTML)), while saving links to documents 230 from which they are extracted in order to facilitate mapping.
  • the usable format of the content may then be split up into sentences.
  • breaking content into sentences involves assembling a string of characters as an input, applying a set of rules to test the character string for specific properties, and, based on the specific properties, dividing the content into sentences.
  • the specific properties of the content being tested may include punctuation and capitalization in order to determine the beginning and end of a sentence.
  • each individual sentence is examined to detect words therein and to potentially recognize each word as an object (e.g., "The Hindenburg"), an event (e.g., "World War II”), a time (e.g., "September”), a verb, or any other category of word that may be utilized for promoting distinctions between words or for understanding the meaning of the subject sentence.
  • the entity recognition procedure assists in recognizing which words are names, as they provide specific answers to question-related keywords of a query (e.g., who, where, when).
  • recognizing words includes identifying words as names and annotating the word with a tag to facilitate retrieval when interrogating the semantic index 260.
  • identifying words as names includes looking up the words in predefined lists of names to determine if there is a match. If no match exists, statistical information may be used to guess whether the word is a name. For example, statistical information may assist in recognizing a variation of a complex name, such as "USS Enterprise," which may have several common variations in spelling.
  • the parsing procedure when implemented, provides insights into the structure of the sentences identified above.
  • these insights are provided by applying rules maintained in a framework of the grammar specification component 255.
  • these rules, or grammars expedite analyzing the sentences to distill representations of the relationships among the words in the sentences.
  • these representations are referred to as semantic structures, and allow the semantic interpretation component 250 to capture critical information about the grammatical structure of the sentence (e.g., verb, subject, object, and the like).
  • the semantic interpretation component 250 is generally configured to diagnose the role of each word in the semantic structure(s), generated by the document parsing component 240, by recognizing a semantic relationship between the words. Initially, diagnosing may include analyzing the grammatical organization of the semantic structure and separating it into logical assertions that each expresses a discrete idea and particular facts. These logical assertions may be further analyzed to determine a function of each of a sequence of words that comprises the assertion. In one instance, determining the function of the sequence of words includes utilizing an ordered sequence of term- rewriting rules, or any other heuristics known in the relevant field.
  • one or more of the sequence of words may be expanded to include synonyms (i.e., linking to other words that correspond to the expanded word's specific meaning) or hypernyms (i.e., linking to other words that generally relate to the expanded word's general meaning).
  • the semantic index 260 serves to store the semantic representation derived by one or many components of the indexing pipeline 210 and may be configured in any manner known in the relevant field.
  • the semantic index may be configured as an inverted index that is structurally similar to conventional search engine indexes.
  • the inverted index is a rapidly searchable database whose entries are words with pointers to the documents 230, and locations therein, on which those words occur. Accordingly, when writing the semantic structures to the semantic index 260, each word and associated function is indexed along with the pointers to the sentences in documents in which the semantic word appeared.
  • This framework of the semantic index 260 allows the matching component 265 to efficiently access, navigate, and match stored information to recover meaningful search results that correspond with the submitted query.
  • the client device 215, the query parsing component 235, and the semantic interpretation component 245 comprise a query conditioning pipeline 205. Similar to the indexing pipeline 210, the query conditioning pipeline 205 distills meaningful information from a sequence of words. However, in contrast to processing passages within documents 230, the query conditioning pipeline 205 processes words submitted within a query 225. For instance, the query parsing component 235 receives the query 225 and performs various procedures to prepare the words for semantic analysis thereof. These procedures may be similar to the procedures employed by the document parsing component 240 such as text extraction, entity recognition, and parsing.
  • the structure of the query 225 may be identified by applying rules maintained in a framework of the grammar specification component 255 and in the semantic interpretation component 245, thus, deriving a meaningful representation, or proposition, of the query 225.
  • the semantic interpretation component 245 may process the query semantic representation in a substantially comparable manner as the semantic interpretation component 250 interprets the semantic structure derived from a passage of text in a document 230.
  • the semantic interpretation component 245 may identify a grammatical and/or semantic relationship of keywords within a string of keywords (e.g., a question or a phrase) that comprise the query 225.
  • identifying the grammatical and/or semantic relationship includes identifying whether a word or phrase functions as the subject (agent of an action), object, predicate, indirect object, or temporal location of the proposition of the query 225.
  • the proposition is evaluated to identify a logical language structure associated with each of the keywords.
  • evaluation may include one or more of the following steps: determining a function of at least one of the keywords; based on the function, replacing the keywords with a logical variable that encompasses a plurality of meanings (e.g., associating with the function a plurality of meanings); and writing those meanings to the proposition of the query.
  • This proposition of the query 225, the keywords, and the information distilled from the proposition and/or keywords are then sent to the matching component 265 for comparison against the semantic representations extracted from the documents 230 and stored at the semantic index 260.
  • the matching component 265 compares the propositions of the queries 225 against the semantic representations at the semantic index 260 to ascertain matching semantic representations. These matching semantic representations may be mapped back to the documents 230 from which they were extracted by associating the documents 230, and the locations therein, from which the semantic representations were derived. These documents 230, targeted by the associated locations, are collected and sorted by the ranking component 270. Sorting may be performed in any known method within the relevant field, and may include without limitation, ranking according to closeness of match, listing based on popularity of the returned documents 230, or sorting based on attributes of the user submitting the query 225.
  • this exemplary system architecture 200 is but one example of a suitable environment that may be implemented to carry out aspects of the present invention and is not intended to suggest any limitation as to the scope of use or functionality of the invention. Neither should the illustrated exemplary system architecture 200, or the natural language engine 290, be interpreted as having any dependency or requirement relating to any one or combination of the components 235, 240, 245, 250, 255, 260, 265, and 270 as illustrated.
  • one or more of the components 235, 240, 245, 250, 255, 260, 265, and 270 may be implemented as stand-alone devices. In other embodiments, one or more of the components 235, 240, 245, 250, 255, 260, 265, and 270 may be integrated directly into the client device 215. It will be understood by those of ordinary skill in the art that the components 235, 240, 245, 250, 255, 260, 265, and 270 illustrated in FIG. 2 are exemplary in nature and in number and should not be construed as limiting.
  • FIG. 2 diagram 300 of a semantic representation generated from a text portion within a document is illustrated, in accordance with an embodiment of the present invention.
  • Text portions may be extracted from content of one or more documents, for example, which may be stored in a data store for easy access during indexing.
  • the document from which text portions are extracted is a web document, but the document in other embodiments may be any kind of text-based document from any type of collection of documents. It will become clear to one of ordinary skill in the art that any type of document may be retrieved, such as documents retrieved from any document collection or even for the analysis of a particular document with a collection.
  • Text portions may include reported speech and other attitude reports, which may be identified by a plurality of words found in the text portion, such as, but certainly not limited to, denounce, say, believe, desire, deny, etc.
  • attitude reports because they describe a person's attitude toward a certain topic.
  • Reported speech may take the form of direct quotations from a person, or may be second hand reported speech. As the following examples are shown and described, a variety of forms of reported speech and other attitude reports, including those mentioned above, will become apparent.
  • Semantic representations generally encompass three main objectives, including, but not limited to, meanings of various words, relationships between the words, and contexts. Semantic representations allow for a more thorough understanding of text than merely depending on keywords from a query matched with words in documents (e.g., web documents), for example. Here, relationships are determined to allow for a deeper analysis of text.
  • the diagram 300 includes a text portion 305, a first level of association 310, a second level of association 320, and a third level of association 330. Each level of association 310, 320, and 330 contains one or more elements, and one or more relational elements. The relational elements are represented by items 312, 314, 316, 322, 332, and 334 in the embodiment of FIG. 3.
  • the elements include the words “denounce,” “Bush,” “Washington,” “calls,” “withdraw,” “US,” and “Iraq.” Also illustrated for each level of association is a reporting act, which, here, are the words “denounce,” “calls,” and “withdraw.” As such, in some embodiments, there may be some words that are elements, but are also categorized as reporting acts, such as “denounce,” “calls,” and “withdraw.” [0056] To clearly illustrate the embodiment of FIG. 3, a semantic representation is shown for text portion 305, which is as follows: “In Washington, George Bush denounced calls for the US to withdraw from Iraq.” It should be noted that FIG. 3 is a diagram of the semantic representation reproduced below.
  • semantic representations are generated and stored in a semantic index, such as semantic index 260 of FIG. 2, but diagrams are not generated. In these embodiments, diagrams are reproduced for illustration and exemplary purposes only.
  • Context(top) word B [George Bush, person] Context(top) word: DNC [denounce, criticize, say] Context(top) word: W [Washington D", city, location] Context(top) word: CL [call, say] Context(3) word: WTHD [withdraw, move] Context(5) word: U [United States of America, country, location] Context(5) word: I [Iraq, country, location] Context(5)
  • Levels of associations also referred to herein as contexts, that have been identified in the text portion (i.e., item 305 in FIG. 3).
  • the levels of association, or contexts are Context(top), Context(3), and Context(5).
  • the levels of association are identified as being the topic of the reporting acts, which are generally
  • action words and in some embodiments, are verbs.
  • "denounce” is the reporting act associated with the first level of association 310.
  • the second level of association 320 may be considered to be the topic of the reporting act, "denounce,” identified in the first level of association 310.
  • the third level of association 330 may be the topic of the reporting act, "calls,” identified in the second level of association 320.
  • Levels of association are formed to gather together a bundle of relations that all hold true in the same way.
  • a top level of association such as Context(top) may be one that holds true according to every question of the sentence. For example, in the embodiment of FIG. 3, it may be true that Bush made the statement in Washington no
  • reporting acts may be determined based on a number of factors, and a reporting act may be identified for each level of association.
  • a reporting act in some instances is an action word, such as, in the embodiment of FIG.
  • Reporting acts may be, for instance, verbs, nouns, and the like, and are typically determined by the surrounding text, or how the word is used in the sentence. This type of grammatical information may be determined, for example, by applying a set of rules, which may be maintained in a framework of the grammar specification component 255 of FIG. 2, for example. By applying a set of rules, or grammars, relationships of words are determined, which leads to the identification of reporting acts. [0060] As shown in FIG. 3, reporting acts are linked to elements, such as words or phrases, or may be linked to a different level of association. Reporting acts are identified as roles of an event, which in this example, may be termed a denunciation event.
  • agent 312 is a relational element linking the two words together thus forming a semantic relationship.
  • location 314 is a relational element linking two words together, which include “denounce” and "Washington.” In order to link a relational element with a word(s) within a different layer of context, a topic may be found that links the two together.
  • logical variable may represent a plurality of synonyms having meanings similar to the element, categories into which the element fits, and may also represent a number of meanings that the element may have. Some elements are easier than others to determine the correct meaning. Meanings may be determined, in one instant, based on how the element is used within the context of the text portion. As shown above, “Bush” is identified as “George Bush,” which is identified as a person.
  • “Denounce,” the reporting act, is associated with both “criticize” and “say,” provided here for exemplary purposes only. There may be a plurality of other words having a similar meaning to “denounce,” and may also be determined to be associated with it. Also, “Washington” is associated with categories including city, and location. “Withdraw” is associated with “move,” and both “US” and “Iraq” are categorized as a country and a location.
  • Semantic representations such as that illustrated in FIG. 3, allow for better, more accurate and more relevant search results to be returned to a user after the user's query is received and analyzed.
  • this text may be returned to a user upon receiving a natural language query such as, "What did Bush say about Iraq,” but not "What did Bush say about Washington.”
  • a conventional keyword search on the assumption that it could identify "denouncing” as a form of "saying," would give the terms “Washington,” “US,” and “Iraq” equal prominence in the target sentence, leading to its retrieval by a keyword query such as "say Bush Washington.”
  • a more advanced indexing scheme that linked the term “denounced” to its direct argument, "calls,” but which went no further, would fail to detect that the denunciation was about Iraq.
  • the term “Washington” is excluded from being linked to "
  • FIG. 4 illustrates a diagram 400 of a semantic representation generated from a text portion within a document, in accordance with an embodiment of the present invention.
  • text portions may include reported speech and other attitude reports, which may be identified by a plurality of words found in the text portion, such as, but certainly not limited to, denounce, say, believe, desire, deny, etc.
  • FIG. 3 provides a semantic representation as a result of analyzing semantic relationships between words, this representation may be supplemented with information about which arguments to the verb "denounce" convey the content of the denunciation. Additional lexical information may be added to indicate what the denunciation is about.
  • FIG. 3 illustrates a diagram 400 of a semantic representation generated from a text portion within a document, in accordance with an embodiment of the present invention.
  • text portions may include reported speech and other attitude reports, which may be identified by a plurality of words found in the text portion, such as, but certainly not limited to, denounce, say, believe, desire, deny, etc.
  • FIG. 3 provides
  • Context(5) WTHD agent: U Context(5): WTHD location: I word: B [George Bush, person] Context(top) word: DNC [denounce, criticize, say] Context(top) word: W [Washington_ DC, city, location] Context(top) word: CL [call, say] Context(3) word: WTHD [withdraw, move] Context(5) word: U [United States of America, country, location] Context(5) word: I [Iraq, country, location] Context(5)
  • FIG. 4 illustrates a text portion 405, of which is semantically represented.
  • FIG. 4 illustrates three levels of association, those being a first level of association 410, a second level of association 430, and a third level of association 440.
  • FIG. 5 a diagram 500 of a semantic representation generated from a text portion within a document is shown, in accordance with an embodiment of the present invention.
  • the semantic representation of this embodiment is shown below for the following text portion 505: "In Washington, George Bush denounced calls for the US to withdraw from Iraq.”
  • Context(5) WTHD spoken: DNC word: B [George Bush, person] Context(top) word: DNC [denounce, criticize, say] Context(top) word: W [Washington_ DC, city, location] Context(top) word: CL [call, say] Context(3) word: WTHD [withdraw, move] Context(5) word: U [United States of America, country, location] Context(5) word: I [Iraq, country, location] Context(5) [0068] As shown above, instead of marking each element that the report is about, as was the case in FIG. 4, the index marks only the head of each reported fact. For instance, FIG. 4, the index marks only the head of each reported fact. For instance, FIG.
  • FIG. 5 appears to be simpler, as it does not explicitly contain any "about” relationships using "about” relational elements.
  • FIG. 5 contains fewer identified relationships, it occupies a smaller amount of space within an index, and therefore a data store where the index is stored. While less memory and storage space may be used for the embodiment of FIG. 5, more time may be required to match a query proposition with semantic representations within the index, as the "about” relationships have not already been identified. In other words, "about” relationships have not been explicitly coded within the index to allow for a quick comparison. This requires much more computing on the back end, which may result in increased wait times for users.
  • FIG. 4 explicitly computes "about" relationships up front and stores those relationships in the index, which allows for a quick comparison of query propositions to semantic representations, resulting in less time required for computations after the user has entered a query.
  • FIG. 6 illustrates a diagram 600 of a semantic representation generated from a text portion within a document, in accordance with an embodiment of the present invention.
  • the embodiment of FIG. 6 illustrates a similar, yet simpler example than the embodiment of FIGS. 3 and 4.
  • the text portion 605 states the following: "John believes that Mary went to Washington.” Below is the semantic representation of text portion 605.
  • Context(top) word J [John, person] Context(top) word: BEL [believe] Context(top) word: W [Washington_ DC, city, location] Context(2) word: G [go, move] Context(2) word: M [Mary, person] Context(2)
  • the elements parsed from raw content include "John,” “believe,” “go,”
  • the second level of association 630 is the topic of "believe,” while there are three words that are “about” the reporting act of "believe.” "Believe” is about “Mary,” where Mary “went” (e.g., go), and where Mary went (e.g., "Washington”).
  • a logical variable in some embodiments, may be replaced with an element, and the logical variable may be associated with a plurality of synonyms, various meanings of that element or word, or the like.
  • FIG. 7 a diagram 700 of a proposition generated from search query is shown, in accordance with an embodiment of the present invention.
  • a proposition is generated in a similar fashion as a semantic representation (e.g., representation of content derived from a web document).
  • the query 705 is as follows: "Who said something about Iraq?" The proposition is shown below.
  • queries are generally shorter in length, and may contain only one level of association, as illustrated in
  • the level of association 710 contains several elements that have been parsed or identified, including "Person,” “say,” and “Iraq.” There is an additional element, but it is similar to a wild card, as it can be many things, not just one word. This additional element represents the word "something" from the query. In one embodiment, elements such as "something" that can match anything may be extracted from the query when the query is being parsed, so as to not pose a restriction when the proposition is being matched to the semantic representation.
  • the proposition illustrated in FIG. 7 also includes a reporting act, "say,” in addition to several relational elements.
  • Agent 712 links “person” to “say.”
  • About 714 links “say” to "Iraq,” and topic 716 links “say” to the wild card element, which as mentioned above, can be anything. As shown, the word “who” is replaced with "person” in the proposition.
  • a semantic representation generated from content of a document e.g., web document
  • a proposition generated from a query such as that shown above in relation to FIG. 6, may be matched or linked in order to determine the most relevant search results from the received query.
  • the semantic representation below illustrates a matching of a semantic representation and a proposition. The matches are shown adjacent to each other.
  • Context(5) WTHD location: I word: B [George Bush, person] Context(top) word: P-2 [person]
  • Context(top) word DNC [denounce, criticize,say]
  • Context(top) word SY-2 [say]
  • Context(top) word W [Washington_ DC, city, location]
  • Context(top) word CL [call, say]
  • Context(top) word WTHD [withdraw, move] Context(5) word: U [United States of America, country, location] Context(5) word: I [Iraq,country,location]Context(5) word:I-2[Iraq,country,location] Context(top) [0077] If there is a match of relational elements, such as the match of agent to agent, the elements associated with the relational elements are then inspected to determine if the words are the same, or even similar. Above, it was mentioned that "denounce” was associated with “say” in order to broaden the search, and "Bush” was associated with "person” for the same reason. Therefore, a match is found between Context(top): DNC agent: B and Context(top): SY-2 agent: P-2.
  • FIG. 8 illustrates a diagram 800 of a semantic representation generated from a text portion within a document, the text portion comprising two sentences, in accordance with an embodiment of the present invention.
  • the embodiment of FIG. 8 illustrates that more than one sentence may be represented in a single semantic representation, especially if the sentences are related.
  • both sentences are authored by the same person, Bush. Therefore, it makes sense and is relevant to put both sentences in a single representation.
  • any number of sentences or even phrases may be grouped together to generate a semantic representation.
  • the process of parsing content extracted from a document e.g., web document
  • This process may be performed by a document parsing component, such as component 240 of FIG. 2.
  • the first level of association 820 e.g., Top context (t)
  • the second level of association 840 e.g., Context (ctx-7)
  • the reporting act "say” located within the first level of association 820.
  • "Say” and the second level of association 840 are associated by a topic relational element 824, such that the sentence “calls to withdraw are bad” is the topic of what Bush said.
  • a plurality of "about” relationships are also formed, and as mentioned above, allow for a greater accuracy of search results.
  • "say” is linked to "Bush” through an agent relational element 822, as Bush is the person or agent who spoke or said those words.
  • relational elements 826, 828, and 830 are linked from “say” to “withdraw,” “calls,” and “bad,” respectively. These "about” relationships or associations allow for efficient and effective matching of these relationships to similar relationships found in query propositions. Additional, within the second layer of association 840, reporting act "calls” is directly linked to "withdraw” by relational element topic 842, and to "bad” by relational element modifier 844.
  • a modal e.g., should
  • the two text portions 805 and 810 may now be intertwined as to determine “aboutness” relations between the first text portion 805 and the second text portion 810.
  • FIG. 8 illustrates that "about” relationships are formed between “say” and “US,” “stay,” and “Iraq” through relational elements about 852, about 854, and about 856, respectively.
  • FIG. 9 a flow diagram 900 illustrating a method for developing semantic relationships between elements distilled from content of a document to generate a semantic representation of the content is shown, in accordance with an embodiment of the present invention.
  • a text portion of a document is identified at step 910, which allows for the identified text portion to be indexed and stored in semantic index 260 of FIG. 2, for example.
  • Text portions may be derived from content of one or more documents, such as web pages, which may be stored in a data store, such as data store 220 of FIG. 2.
  • the format of the content may be a raw online format that requires conversion.
  • the content is converted from a raw online format to a HyperText Markup Language (HTML) to generate the text portion.
  • HTML HyperText Markup Language
  • Content may be extracted in the form of one or more sentences or phrases, a table, a template, or a plurality of data.
  • Text portions may include reported speech and other attitude reports, which may be identified by a plurality of words found in the text portion, such as, but certainly not limited to, denounce, say, believe, desire, deny, etc. These words are identified in attitude reports because they describe a person's attitude toward a certain topic.
  • Reported speech may take the form of direct quotations from a person, or may be second-hand reported speech.
  • the text portion may be parsed in order to identify one or more elements that are to be semantically represented for further indexing. Parsing may also include text extraction and entity recognition, wherein an entity is recognized by searching a predefined list of words stored in data store 220, for example. This procedure assists in that it recognizes words that may be names of a person or thing.
  • semantic information for each of the identified elements is determined.
  • the semantic information may include one or more meanings and/or grammatical functions of the identified elements therein.
  • synonyms or hyponyms may also be determined and included as semantic information.
  • one or more words may have similar meanings, and those words and meanings may be represented in a semantic representation by a logical variable by replacing a certain element with the logical variable.
  • This logical variable may represent a plurality of synonyms having meanings similar to the element, categories of which the element fits, and may also represent a number of meanings that the element may have, which allows for a broadened, but more accurate search.
  • Logical variables may be stored in a data store.
  • one or more levels of association, or contexts may be determined, and each level of association may include one or more of the identified elements. Elements within different levels of association may be associated with each other, and may be associated by way of a reporting act. Reporting acts may be, for instance, verbs, nouns, or the like, and are typically determined by the surrounding text, or how the word is used in the sentence. This type of grammatical information may be determined, for example, by applying a set of rules, which may be maintained in a framework of grammar specification component 255 of FIG. 2, for example. For exemplary purposes only, suppose a text portion recites: "In Washington, Bush denounced calls for the US to withdraw from Iraq.” Here, three levels of association may be identified, each containing a reporting act.
  • relational elements may also be determined, which describe the relationship between a reporting act and an element or a level of association. For example, with continued reference to the example above, “Bush” may be associated or linked to "denounced” by way of a relational element agent, as Bush is the agent doing the denouncing. Relational elements may take various forms of relationships, but may be words such as, but not limited to, an agent, a location, a topic, or about. "About" relationships indicate what the reporting act is referring to, or what it is about.
  • FIG. 10 shows a flow diagram 1000 that illustrates a method for, in response to receiving a query, creating associations between various terms distilled from the query to generate a proposition, the proposition being used to interrogate information stored in an index to provide relevant search results, in accordance with an embodiment of the present invention.
  • a proposition is a logical representation of a conceptual meaning of the query that is used to interrogate semantic relationships contained within semantic representations of content from the documents. The process of generating a proposition from a query is very similar to the process described herein for generating a semantic representation of content of a document.
  • a query is received as input from a user, and in one embodiment, the received query is parsed to determine one or more search terms within the query. Search terms are similar to elements identified in a text portion.
  • semantic information for each of one or more search terms is determined, and this semantic information may include one or more meanings and/ or grammatical functions of the search terms therein.
  • a logical variable may be identified and may be associated with, or may even replace one or more of the search terms.
  • a logical variable may be a number, letter, or a series or combination of both, and may represent a plurality of words having similar meanings to the search terms. This allows for a broadened, yet more relevant return of search results to the user.
  • a first reporting act is identified within the query at step 1020.
  • the reporting act may be a verb, noun, or any other part of speech, and may include an action, such as "say,” “call,” “denounce,” “believe,” etc.
  • more than one reporting act may be identified within a query, such as a second reporting act.
  • a semantic relationship may be determined between each reporting act and another search term to create an association between the words, and is shown at step 1030. Semantic relationships may be based on the determined semantic information, as described above. Associations are linked by relational elements, which describe the association, such as, but not limited to, agent, a location, a topic, or about. Other relational elements are certainly contemplated to be within the scope of the present invention.
  • a proposition is generated that includes the formed associations between each reporting act and one or more of the search terms parsed from the query.
  • the proposition e.g., associations within the proposition
  • the proposition may be compared or matched against one or more semantic representations stored in semantic index 260, for example, to determine the most relevant matches for the proposition.
  • a query may contain more than one level of association, as described above, and thus a reporting act may be identified for each level of association.
  • FIG. 11 a flow diagram 1100 illustrating a method for developing semantic relationships between elements distilled from content of a document to generate a semantic representation of the content, further allowing for indexing of the content, is shown, in accordance with an embodiment of the present invention.
  • At step 1110 at least a portion of a document (e.g., web document) to be indexed is identified.
  • the text portion of the document is parsed to identify elements that are to be semantically represented, shown at step 1120.
  • a data store is accessed to determine potential meanings and grammatical functions of the identified elements.
  • one or more levels of association within the text portion are determined.
  • a reporting act within the text portion is identified for each of the one or more determined levels of association, shown at step 1150.
  • a first reporting act is associated with a first set of identified elements, which are determined by analyzing semantic relationships between the elements determined at step 1120 above and the determined reporting act.
  • the first reporting act is associated with a first level of association.
  • a second reporting act is associated with a second set of identified elements at step 1170, and the second reporting act is associated with a second level of association.
  • a semantic representation of the associations may then be generated at step 1180 so that it may be stored in semantic index 260, for example, for further analysis, including a comparison to query propositions, as described above.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Machine Translation (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

L'invention concerne des procédés et des supports lisibles par ordinateur associant des mots et des groupes de mots extraits du contenu (discours rapporté ou compte rendu d'attitude, par exemple) d'un document en vue de constituer des relations sémantiques utilisées collectivement pour générer une représentation sémantique du contenu. Des représentations sémantiques peuvent comprendre des éléments identifiés ou analysés à partir d'une partie texte du contenu dont les éléments peuvent être associés à d'autres éléments partageant une relation sémantique, telle qu'une relation de sujet, de lieu ou d'objet. Des relations peuvent également être élaborées par association d'un élément en relation avec ou concernant un autre élément, ce qui permet une comparaison rapide et effective d'associations trouvées dans une représentation sémantique avec des associations résultant d'interrogations. Les relations sémantiques peuvent être déterminées sur la base d'informations sémantiques, telles que des signfications potentielles et des fonctions grammaticales de chaque élément dans la partie texte du contenu.
EP08828391.6A 2007-08-31 2008-08-29 Identification de relations sémantiques dans un discours rapporté Ceased EP2183686A4 (fr)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
US96943407P 2007-08-31 2007-08-31
US12/201,675 US8868562B2 (en) 2007-08-31 2008-08-29 Identification of semantic relationships within reported speech
PCT/US2008/074938 WO2009029905A2 (fr) 2007-08-31 2008-08-29 Identification de relations sémantiques dans un discours rapporté

Publications (2)

Publication Number Publication Date
EP2183686A2 true EP2183686A2 (fr) 2010-05-12
EP2183686A4 EP2183686A4 (fr) 2018-03-28

Family

ID=42041477

Family Applications (1)

Application Number Title Priority Date Filing Date
EP08828391.6A Ceased EP2183686A4 (fr) 2007-08-31 2008-08-29 Identification de relations sémantiques dans un discours rapporté

Country Status (11)

Country Link
EP (1) EP2183686A4 (fr)
JP (1) JP5501967B2 (fr)
KR (1) KR101524889B1 (fr)
CN (1) CN101796511B (fr)
AU (1) AU2008292781B2 (fr)
BR (1) BRPI0816088A2 (fr)
CA (1) CA2698105C (fr)
IL (1) IL204108A (fr)
MX (1) MX2010002350A (fr)
RU (1) RU2488877C2 (fr)
WO (1) WO2009029905A2 (fr)

Families Citing this family (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9842168B2 (en) 2011-03-31 2017-12-12 Microsoft Technology Licensing, Llc Task driven user intents
KR101922744B1 (ko) * 2011-03-31 2018-11-27 마이크로소프트 테크놀로지 라이센싱, 엘엘씨 위치-기반 대화 해석 기법
US9760566B2 (en) 2011-03-31 2017-09-12 Microsoft Technology Licensing, Llc Augmented conversational understanding agent to identify conversation context between two humans and taking an agent action thereof
US10642934B2 (en) 2011-03-31 2020-05-05 Microsoft Technology Licensing, Llc Augmented conversational understanding architecture
US9244984B2 (en) 2011-03-31 2016-01-26 Microsoft Technology Licensing, Llc Location based conversational understanding
US9064006B2 (en) 2012-08-23 2015-06-23 Microsoft Technology Licensing, Llc Translating natural language utterances to keyword search queries
JP6176017B2 (ja) * 2013-09-17 2017-08-09 富士通株式会社 検索装置、検索方法、およびプログラム
RU2544739C1 (ru) * 2014-03-25 2015-03-20 Игорь Петрович Рогачев Способ преобразования структурированного массива данных
CA2970168A1 (fr) * 2014-12-10 2016-06-16 Kyndi, Inc. Codage de donnees sous-symbolique pondere
US10503832B2 (en) * 2016-07-29 2019-12-10 Rovi Guides, Inc. Systems and methods for disambiguating a term based on static and temporal knowledge graphs
CN107818076B (zh) * 2016-09-12 2021-11-12 微软技术许可有限责任公司 针对自然语言的语义处理
US10798027B2 (en) * 2017-03-05 2020-10-06 Microsoft Technology Licensing, Llc Personalized communications using semantic memory
JP7176233B2 (ja) 2018-06-04 2022-11-22 富士通株式会社 検索方法、検索プログラムおよび検索装置
CN110895657B (zh) * 2018-09-11 2023-05-26 慧捷(上海)科技股份有限公司 一种基于口语对话特征的语义逻辑表达和分析方法
WO2020079752A1 (fr) * 2018-10-16 2020-04-23 株式会社島津製作所 Procédé de recherche de document et système de recherche de document
CN109871428B (zh) * 2019-01-30 2022-02-18 北京百度网讯科技有限公司 用于确定文本相关度的方法、装置、设备和介质

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6076051A (en) * 1997-03-07 2000-06-13 Microsoft Corporation Information retrieval utilizing semantic representation of text
EP0962873A1 (fr) * 1998-06-02 1999-12-08 International Business Machines Corporation Traitement des informations textuels et comprehension automatisée d'information
US6901402B1 (en) * 1999-06-18 2005-05-31 Microsoft Corporation System for improving the performance of information retrieval-type tasks by identifying the relations of constituents
US7120574B2 (en) * 2000-04-03 2006-10-10 Invention Machine Corporation Synonym extension of search queries with validation
US8799776B2 (en) * 2001-07-31 2014-08-05 Invention Machine Corporation Semantic processor for recognition of whole-part relations in natural language documents
WO2003100659A1 (fr) * 2002-05-28 2003-12-04 Vladimir Vladimirovich Nasypny Procede de synthese d'un systeme a auto-apprentissage d'extraction de connaissances a partir de documents textuels pour moteurs de recherche
RU2273879C2 (ru) * 2002-05-28 2006-04-10 Владимир Владимирович Насыпный Способ синтеза самообучающейся системы извлечения знаний из текстовых документов для поисковых систем
US7424420B2 (en) * 2003-02-11 2008-09-09 Fuji Xerox Co., Ltd. System and method for dynamically determining the function of a lexical item based on context
US7593845B2 (en) * 2003-10-06 2009-09-22 Microsoflt Corporation Method and apparatus for identifying semantic structures from text
US20050182617A1 (en) * 2004-02-17 2005-08-18 Microsoft Corporation Methods and systems for providing automated actions on recognized text strings in a computer-generated document
JP2005284723A (ja) * 2004-03-30 2005-10-13 Fuji Xerox Co Ltd 自然言語処理システム及び自然言語処理方法、並びにコンピュータ・プログラム
US20070073533A1 (en) * 2005-09-23 2007-03-29 Fuji Xerox Co., Ltd. Systems and methods for structural indexing of natural language text

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See references of WO2009029905A2 *

Also Published As

Publication number Publication date
CA2698105A1 (fr) 2009-03-05
BRPI0816088A2 (pt) 2015-03-03
KR20100075454A (ko) 2010-07-02
RU2488877C2 (ru) 2013-07-27
WO2009029905A3 (fr) 2009-05-14
AU2008292781B2 (en) 2012-08-09
KR101524889B1 (ko) 2015-06-01
IL204108A (en) 2013-09-30
MX2010002350A (es) 2010-07-30
JP2010538375A (ja) 2010-12-09
CN101796511B (zh) 2012-11-14
CN101796511A (zh) 2010-08-04
WO2009029905A2 (fr) 2009-03-05
RU2010107150A (ru) 2011-09-10
EP2183686A4 (fr) 2018-03-28
CA2698105C (fr) 2016-07-05
JP5501967B2 (ja) 2014-05-28
AU2008292781A1 (en) 2009-03-05

Similar Documents

Publication Publication Date Title
US9449081B2 (en) Identification of semantic relationships within reported speech
CA2698105C (fr) Identification de relations semantiques dans un discours rapporte
US8639708B2 (en) Fact-based indexing for natural language search
US8229730B2 (en) Indexing role hierarchies for words in a search index
US10698977B1 (en) System and methods for processing fuzzy expressions in search engines and for information extraction
US10713571B2 (en) Displaying quality of question being asked a question answering system
US9864808B2 (en) Knowledge-based entity detection and disambiguation
US8346795B2 (en) System and method for guiding entity-based searching
US8291319B2 (en) Intelligent self-enabled solution discovery
US20090070322A1 (en) Browsing knowledge on the basis of semantic relations
TW201314476A (zh) 基於本體之自動自助式使用者支援
US20180225374A1 (en) Automatic Corpus Selection and Halting Condition Detection for Semantic Asset Expansion
WO2009035871A1 (fr) Connaissances de navigation sur la base de relations sémantiques
CA2914398A1 (fr) Identification de relations semantiques dans un discours rapporte
EP2181403B1 (fr) Indexation d'arborescence des fonctions pour des mots dans un index de recherche
EP2185999A2 (fr) Mise en évidence de résultats de recherche en fonction d'une signification conceptuelle
WO2009029922A2 (fr) Indexation factuelle servant à une recherche en langage naturel
Panchasara et al. AI Based Job Recommedation System using BERT

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 20100223

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MT NL NO PL PT RO SE SI SK TR

AX Request for extension of the european patent

Extension state: AL BA MK RS

DAX Request for extension of the european patent (deleted)
RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: MICROSOFT TECHNOLOGY LICENSING, LLC

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: ZHIGU HOLDINGS LIMITED

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: ZHIGU HOLDINGS LIMITED

A4 Supplementary search report drawn up and despatched

Effective date: 20180226

RIC1 Information provided on ipc code assigned before grant

Ipc: G06F 17/28 20060101ALI20180220BHEP

Ipc: G06F 17/27 20060101AFI20180220BHEP

Ipc: G06F 17/30 20060101ALI20180220BHEP

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: EXAMINATION IS IN PROGRESS

17Q First examination report despatched

Effective date: 20190220

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: EXAMINATION IS IN PROGRESS

REG Reference to a national code

Ref country code: DE

Ref legal event code: R003

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION HAS BEEN REFUSED

18R Application refused

Effective date: 20210622