WO2017198031A1 - 解析语义的方法和装置 - Google Patents

解析语义的方法和装置 Download PDF

Info

Publication number
WO2017198031A1
WO2017198031A1 PCT/CN2017/081197 CN2017081197W WO2017198031A1 WO 2017198031 A1 WO2017198031 A1 WO 2017198031A1 CN 2017081197 W CN2017081197 W CN 2017081197W WO 2017198031 A1 WO2017198031 A1 WO 2017198031A1
Authority
WO
WIPO (PCT)
Prior art keywords
feature
text
service
parsing
business
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
PCT/CN2017/081197
Other languages
English (en)
French (fr)
Inventor
张喜
周鑫
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.)
Beijing Jingdong Century Trading Co Ltd
Beijing Jingdong Shangke Information Technology Co Ltd
Original Assignee
Beijing Jingdong Century Trading Co Ltd
Beijing Jingdong Shangke Information Technology Co Ltd
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 Beijing Jingdong Century Trading Co Ltd, Beijing Jingdong Shangke Information Technology Co Ltd filed Critical Beijing Jingdong Century Trading Co Ltd
Priority to RU2018142588A priority Critical patent/RU2704531C1/ru
Priority to US16/303,000 priority patent/US10824816B2/en
Publication of WO2017198031A1 publication Critical patent/WO2017198031A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/242—Query formulation
    • G06F16/243—Natural language query formulation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/205—Parsing
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis

Definitions

  • the present application relates to the field of computer technologies, and in particular, to the field of Internet technologies, and in particular, to a method and apparatus for parsing semantics.
  • the application of intelligent robot technology in Internet products is relatively concentrated, such as: banking business consulting, e-commerce pre-sales and after-sales business Q & A.
  • the method of robotic parsing semantics usually includes: firstly sorting the business knowledge manually, classifying the business knowledge, configuring the questions and answers, and then in the case of artificial “supervision”, “church” the robot, and finally replace it by the robot. Labor to provide services.
  • the above-mentioned method of parsing user semantics by the robot can only parse the general intention of the user semantics, and the accuracy of the parsing semantics is low.
  • the purpose of the present application is to propose an improved method and apparatus for parsing semantics to solve the technical problems mentioned in the background art above.
  • the present application provides a method for parsing semantics, the method comprising: Receiving semantic information input by the user; parsing the semantic information to obtain a text feature of the associated service; calculating a similarity between the one-time text feature and the corpus feature in the corpus; and the corpus feature having the highest similarity with the one-time text feature
  • the associated corpus is used as a parsing text; the feature of the parsed text is extracted to obtain a quadratic text feature; the service feature matching the secondary text feature in the business feature database and the service domain to which the acquired service feature belongs are obtained; Obtaining a weight value associated with the acquired service feature in the weight library of the service feature; the service domain to which the service feature having the highest weight value belongs is the service domain to which the parsed text belongs; and the service domain to which the parsed text belongs Parsing the parsed text once and obtaining the parsed text.
  • the present application provides an apparatus for parsing semantics, where the apparatus includes: a semantic receiving unit, configured to receive semantic information input by a user; and a parsing unit configured to parse the semantic information to obtain an associated service.
  • a text feature configured to receive semantic information input by a user; and a parsing unit configured to parse the semantic information to obtain an associated service.
  • a text feature configured to receive semantic information input by a user; and a parsing unit configured to parse the semantic information to obtain an associated service.
  • a text feature configured to calculate a similarity between the one-time text feature and the corpus feature in the corpus
  • the corpus determining unit is configured to use the corpus to which the corpus feature having the highest similarity with the one-time text feature belongs
  • a feature extraction unit configured to extract a feature of the one-time parsed text, to obtain a secondary text feature
  • a feature acquiring unit configured to acquire a service feature in the service feature database that matches the secondary text feature and obtain the
  • the method and device for parsing semantics obtaineds a text feature by extracting the text feature of the associated service from the semantic information input by the user, and then calculates the similarity between the text feature and the corpus feature in the corpus, and then the text feature
  • the corpus to which the corpus feature with the highest similarity belongs is used as the parsing text, and then the feature of the parsed text is extracted once, and the secondary text feature is obtained.
  • the business feature matching the secondary text feature in the business feature database and the acquired business feature are obtained.
  • the business domain and then obtain the weight value associated with the acquired service feature in the weight library of the service feature, and then the business domain to which the service feature having the highest weight value belongs is the business domain to which the parsed text belongs. After parsing the parsed text once in the business domain to which the parsed text belongs, the parsed text is obtained twice, thereby improving the accuracy of parsing the semantic result.
  • FIG. 1 is an exemplary system architecture diagram to which the present application can be applied;
  • FIG. 2 is a flow diagram of one embodiment of a method of parsing semantics in accordance with the present application
  • FIG. 3 is a schematic flowchart of an application scenario of a method for parsing semantics according to the present application
  • FIG. 4 is a schematic structural diagram of an embodiment of an apparatus for parsing semantics according to the present application.
  • FIG. 5 is a schematic structural diagram of a computer system suitable for implementing a terminal device or a server of an embodiment of the present application.
  • FIG. 1 illustrates an exemplary system architecture 100 of an embodiment of an apparatus that can apply the method of parsing semantics or parsing semantics of the present application.
  • system architecture 100 can include terminal devices 101, 102, 103, network 104, and servers 105, 106.
  • the network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the servers 105, 106.
  • Network 104 may include various types of connections, such as wired, wireless communication links, fiber optic cables, and the like.
  • User 110 can use terminal device 101, 102, 103 to communicate with network 104 and The devices 105, 106 interact to receive or send messages and the like.
  • Various communication client applications such as a web browser application, a shopping application, a search application, an instant communication tool, a mailbox client, a social platform software, and the like, may be installed on the terminal devices 101, 102, and 103.
  • the terminal devices 101, 102, 103 may be various electronic devices having a display screen and supporting user input of semantic information, including but not limited to smart phones, tablets, MP3 players (Moving Picture Experts Group Audio Layer III, motion picture expert compression) Standard audio level 3), MP4 (Moving Picture Experts Group Audio Layer IV) player, laptop portable computer and desktop computer, etc.
  • MP3 players Motion Picture Experts Group Audio Layer III, motion picture expert compression
  • MP4 Motion Picture Experts Group Audio Layer IV
  • laptop portable computer and desktop computer etc.
  • the servers 105, 106 may be servers that provide various services, such as a background server that parses semantic information entered by the user on the terminal devices 101, 102, 103 and provides answer information.
  • the background server may analyze and process data such as semantic information input by the user, and feed back the processing result (for example, an answer) to the terminal device.
  • the method for parsing semantics provided by the embodiments of the present application is generally performed by the servers 105 and 106. Accordingly, the device for parsing semantics is generally disposed in the servers 105 and 106.
  • terminal devices, networks, and servers in Figure 1 is merely illustrative. Depending on the implementation needs, there can be any number of terminal devices, networks, and servers.
  • the method for parsing semantics includes the following steps:
  • Step 201 Receive semantic information input by a user.
  • the electronic device for example, the server shown in FIG. 1 on which the method for parsing semantics can receive semantic information from a terminal with which the user inputs using a wired connection or a wireless connection.
  • the above wireless connection manner may include but is not limited to 3G/4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods that are now known or developed in the future. .
  • the user browses the webpage by using a web browser application, a shopping application, a search application, an instant communication tool, and the like installed on the terminal, and the user can pass one or more of the voice information, the text information, and the lip motion information.
  • the semantic information may include a text format, an audio format, a picture format, or other semantic information of a format to be developed in the future.
  • Step 202 Parse the semantic information to obtain a text feature of the associated service.
  • the semantic information may be parsed to obtain a textual feature of the associated service.
  • the semantic information is text information
  • the feature of the text information can be directly extracted, and one text feature of the associated service is filtered out in the extracted feature;
  • the semantic information is voice information and/or lip motion information, it needs to be first
  • the voice information and/or the lip motion information are parsed to obtain corresponding text information, and then the feature information is extracted from the text information, and one text feature of the associated service is selected from the extracted features.
  • the feature-based feature extraction method in the prior art or in the future developed technology may be used to extract features of the text information.
  • a word frequency-reverse file frequency method, a word frequency method, a file frequency method, or the like can be used to extract features of text information.
  • These statistical-based feature extraction methods can evaluate each feature in the feature set by constructing an evaluation function, and score each feature, so that each word obtains an evaluation value, also called a weight. All features are then sorted by weight, and a predetermined number of optimal features are extracted as feature subsets of the extraction results.
  • parsing the semantic information to obtain the primary text feature of the associated service may include: first obtaining the text information of the semantic information; then cleaning and reducing the dimensionality of the text information; and then cleaning and reducing the dimension After the text information is semantically and syntactically analyzed, the text information features are obtained. Finally, the features of the associated business in the text information feature are extracted to obtain a text feature.
  • cleaning text information refers to the last procedure for discovering and correcting identifiable errors in data files, including checking data consistency, handling invalid values, and missing values.
  • the task of data cleaning is to filter data that does not meet the requirements.
  • the data that does not meet the requirements are mainly three types: incomplete data, erroneous data, and repeated data.
  • converting Chinese and English symbols in text information converting Simplified and Traditional Chinese characters, and cleaning duplicate data, conflicting data, and meaningless characters, etc., meaningless characters here refer to features that do not affect classification, such as "Hello.” , "Excuse me” and swearing.
  • Dimensioning text information is to map sample points from the input space to a low-dimensional space through linear or nonlinear transformations. This results in a low dimensional representation of the compactness of the original data set.
  • the following methods are used to reduce the dimensionality of text information: missing value ratio method, high correlation filtering method, principal component analysis method, and the like.
  • the syntactic analysis here refers to the analysis of various grammatical relations between words and words in the syntactic structure: subjective, replenishment, narration, partiality, union, and so on.
  • the semantic relationship here refers to the relationship between objective objects represented by linguistic components in semantics.
  • the semantic relations we refer to are actions and actors, actions and actors, actions and tools, actions and premises, The relationship between things and things, etc.
  • the pre-established word segmentation model can be used to complete the analysis.
  • the data of the training segmentation model can be historical question and answer data containing business knowledge, or manually added data.
  • extracting the feature of the associated service in the text information feature means that on the basis of the text information feature, a part of the text feature word without the service feature is discarded and the weight of the other part of the text feature word is increased.
  • Step 203 Calculate the similarity between the text feature and the corpus feature in the corpus.
  • the corpus refers to an aggregate of written texts or spoken words that are previously marked by humans, and can be used for language analysis and language description.
  • a corpus can also be thought of as a finite collection of electronic machine-readable text, to the greatest extent representing a language or language variant. Calculating the similarity between the text feature and the corpus features in the corpus, it is possible to determine the corpus to which the corpus feature with the highest similarity to the primary text feature belongs, that is, to parse the text once.
  • Step 204 The corpus to which the corpus feature having the highest degree of similarity with the text feature belongs is used as the parsing text.
  • the corpus to which the corpus feature having the highest similarity with the primary text feature belongs can be selected, and the selected corpus is used as one parsing. text.
  • step 205 the feature of the parsed text is extracted once to obtain a secondary text feature.
  • the feature of the parsed text can be extracted once, and the extracted feature of the parsed text is used as the quadratic text feature.
  • the method of extracting the features of the parsed text once, with the above extracted text may be the same or different, and the feature of the analytic text may be extracted by using the statistic-based feature extraction method in the prior art or the technology developed in the future.
  • the word frequency-reverse file frequency method, the word frequency method, the file frequency method, and the like can be used to extract the features of the parsed text once.
  • These statistical-based feature extraction methods can evaluate each feature in the feature set by constructing an evaluation function, and score each feature, so that each word obtains an evaluation value, also called a weight. All features are then sorted by weight, and a predetermined number of optimal features are extracted as feature subsets of the extraction results.
  • extracting the feature of the parsed text once, and obtaining the quadratic text feature may include: first obtaining the parsed text; then cleaning and reducing the dimension of the parsed text; then cleaning and reducing the dimension After the analysis of the text, the semantic and syntactic analysis is performed to obtain the secondary text feature.
  • the cleaning and dimension reduction here are the same as the above method of cleaning and reducing the dimensionality of the text information, and will not be described here.
  • Step 206 Obtain a service feature in the service feature database that matches the secondary text feature and a service domain to which the acquired service feature belongs.
  • the service feature database includes a service feature extracted from a user problem in advance and a service domain to which the feature belongs. Based on the secondary text feature obtained in step 205, a service feature matching the secondary text feature can be obtained in the service feature database, and the service domain to which the service feature belongs is obtained. It should be understood that there may be one or more business domain business features in the business feature library that match the secondary text features.
  • the acquired service feature may be referred to after the secondary text feature is matched with the feature in the service feature database, that is, the service feature is obtained.
  • the service features matching the secondary text features in the library may include: firstly acquiring the service features in the business feature database that match the secondary text features; and then performing the referencing on the service features matching the secondary text features, Refers to the business characteristics after the digestion; finally refers to the business characteristics after the digestion as the acquired business characteristics.
  • pronouns such as: "this”, “this”, “this paragraph”, “this paragraph”, and pronouns modify nouns, gerunds, dispel pronouns;
  • business words are nouns, gerunds, pre-adjectives , Vocabulary, dispelling adjectives and gerunds;
  • Business words are nouns, gerunds, pre-verbs, followed by verbs, then the preceding verbs are eliminated; 4.
  • Business words are verbs, and the sentence contains only one noun, one Verbs, other words are eliminated, such as: "system / n is compatible / v (business words) is very good” is "system compatible”; 5, business words are nouns, and there is only one noun in the sentence, then dissolve Other words of speech, such as: "How much is this size / n (business words)” is digested into "size”; 6, business words are adjectives, if they are business words before and after, compare weights, and retain important business words.
  • the method for parsing semantics further includes: determining, by using the following steps, a service area in which the service feature in the service feature database and the service feature belong to: obtaining the service information of the marked service domain. Samples; perform statistical and machine learning-based segmentation of the sample and determine the part-of-speech of the word segmentation; determine the business characteristics in the business feature database and the business domain to which the business feature belongs according to the frequency and part of speech of the segmentation.
  • Step 207 Acquire a weight value associated with the acquired service feature in the weight library of the service feature.
  • the service feature corresponding to the step 206 is one, and the weight corresponding to the service feature can be obtained in the weight library of the service feature, and the service feature corresponding to the step 206 is multiple, which can be in the service.
  • Each weight of the feature in the feature library obtains its corresponding weight respectively.
  • the weight pool here is a weight preset in advance based on the business characteristics and the business domain to which the business characteristics belong. When prescribing weights, it is usually considered to give higher weight to those business features that can express the theme. When giving business feature weights, factors such as word frequency, part of speech, word length, location, and synonym phenomenon can be fully considered. For example, words with the same or similar meaning are treated as one, and their word frequencies are added together; a noun or a combined word containing a noun is given a higher weight; a word with 4 or more words is given a higher weight.
  • the foregoing method for parsing semantics further includes: The weight of the service feature in the weight library of the service feature is set according to the granularity of the service feature and the corresponding relationship between the granularity and the weight.
  • the granularity refers to the refinement and comprehensiveness of business characteristics.
  • the weights in the weight library of the service feature may be divided into three levels: detailed data, light summary, and high summary.
  • One principle should be noted when determining the granularity of data: the higher the degree of refinement, the smaller the granularity; the lower the degree of refinement, the larger the granularity.
  • the weight of the service feature in the weight library of the service feature may be set according to the preset correspondence between the granularity and the weight.
  • Step 208 The service domain to which the service feature with the highest weight value belongs is regarded as the service domain to which the parsed text belongs.
  • the weight of the service feature obtained in step 207 is a single, the service domain to which the service feature belongs is used as the service domain to which the text is parsed; if the weight of the service feature obtained in step 207 is multiple Then, the plurality of weights are sorted from large to small, and the service domain to which the service feature corresponding to the first weight is selected is selected as the service domain to which the parsed text belongs, thereby more accurately determining the domain to which the user intends belongs.
  • step 209 the parsed text is parsed once in the business domain to which the parsed text belongs, and the parsed text is obtained twice.
  • the parsed text is subjected to statistical-based word segmentation, and then the segmentation word is compared with the feature in the service domain to which the parsed text belongs, and is obtained once. Parse the secondary parsed text of the text within the business domain to which it belongs.
  • the answer to the second parsed text may also be obtained according to the second parsed text: first, according to the service domain to which the parsed text belongs, the answer set is correspondingly determined. A subset of answers in the business domain, and then querying the question corresponding to the secondary parsed text in the subset of answers, and then obtaining an answer to the question, and using the answer as the answer corresponding to the second parsed text, that is, the parsing semantics described above
  • the method further includes: querying an answer corresponding to the second parsed text in an answer set corresponding to the business domain to which the parsed text belongs; and presenting an answer corresponding to the second parsed text.
  • the answer set here may generally include a predetermined question and answer pair with a field mark, the answer to the question including the question and the corresponding question, the question and answer may be from the common problem items and corresponding questions of the official website, or may come from
  • the question and answer pair selected from the chat record of the manual customer service can also be a question and answer pair written manually according to the question.
  • FIG. 3 is a schematic flowchart of an application scenario of a method for parsing semantics according to the embodiment.
  • the method of parsing semantics includes the following steps:
  • step 301 receiving a user question, and then performing step 302;
  • step 302 the data information of the received user problem is cleaned and reduced in dimension, and then step 303 is performed;
  • step 303 the intent recognition is performed on the cleaned and reduced dimensional text information: semantic and syntactic analysis of the cleaned and reduced dimensional text information, obtaining text information features, extracting features of the associated business in the text information, and obtaining a text. Feature, then calculate the similarity between the text feature and the corpus feature in the corpus, and the corpus to which the corpus feature with the highest similarity of the text feature belongs is the result of the intention recognition, and then step 307 is performed;
  • step 304 the result of the above intent recognition is taken as one parsing text
  • step 305 the service feature library including the service feature and the service domain and the weight library including the service feature and the weighted service feature are offlinely calculated;
  • step 306 the weights of the service features and the service features are extracted from the service feature database
  • step 307 statistically-based word segmentation is performed on the parsed text, and then the segmentation word and the extracted business feature are matched. If the match is recorded, the position of the word segment is recorded, and at the same time, whether the part-of-speech type of the participle and the preceding adjacent participle is searched and judged is Consistent or solvable, such as pronouns, adjectives, gerunds, etc., if so, from the position of the word, to the end of the search merged into a phrase, digested into key business words, with corresponding weights; End of the text, re-merging the merged text, the business domain to which the business feature with the highest weight value belongs is regarded as the business domain to which the parsed text belongs, and the business domain is marked for the merged text, and then steps 309 and 310 are performed;
  • an answer set is predetermined, and the answer set includes a question and answer pair with a field mark;
  • step 309 the business text of the parsed text will be marked once in the merged text.
  • the domain is re-parsed to obtain the second parsed text, and then step 311 is performed;
  • step 310 the answer set determined in step 308 is extracted to meet the answer subset of the business domain marked by the merged text, and then step 311 is performed;
  • step 311 the answer feature corresponding to the feature of the secondary parsed text is queried from the answer subset, and the answer feature is assembled into an answer, and then step 312 is performed;
  • step 312 the assembled answer is returned.
  • the weight of “dark yellow” is higher than “skin quality”, then the “dark yellow skin texture” is dissolved into “dark yellow”, and the final sentence is “mask”.
  • the method provided by the above embodiment of the present application improves the accuracy of the result of parsing semantics by determining the business domain to which the parsed text belongs once and parsing the parsed text once in the subdivided business domain to obtain the secondary parsed text.
  • the present application provides an embodiment of an apparatus for parsing semantics, and the apparatus embodiment corresponds to the method embodiment shown in FIG. Can be applied to a variety of electronic devices.
  • the apparatus 400 for parsing semantics in this embodiment includes: a semantic receiving unit 410, a primary parsing unit 420, a similarity calculating unit 430, a corpus determining unit 440, a feature extracting unit 450, a feature acquiring unit 460, The weight acquisition unit 470, the domain determination unit 480, and the secondary analysis unit 490.
  • the semantic receiving unit 410 is configured to receive semantic information input by the user;
  • the parsing unit 420 is configured to parse the semantic information to obtain a primary text feature of the associated service;
  • the similarity calculating unit 430 is configured to calculate a similarity between the primary text feature and the corpus feature in the corpus;
  • the corpus determining unit 440 configured The corpus to which the corpus feature having the highest degree of similarity with the text feature belongs is used as the analytic text;
  • the feature extracting unit 450 is configured to extract the feature of the parsed text once to obtain the secondary text feature;
  • the feature acquiring unit 460 is configured to obtain a service feature that matches the secondary text feature in the service feature database and a service domain to which the acquired service feature belongs;
  • the weight obtaining unit 470 is configured to obtain a weight value associated with the acquired service feature in the weight library of the service feature;
  • the domain determining unit 480 is configured to use the service domain to which the service feature having the highest weight value belongs as the service domain to
  • the device for parsing semantics further includes (not shown in the figure): an answer query unit configured to query and answer the answer in the answer field corresponding to the service domain to which the parsed text belongs once An answer corresponding to the text is parsed; an answer presentation unit configured to present an answer corresponding to the second parsed text.
  • the primary parsing unit is further configured to: obtain text information of the semantic information; perform cleaning and dimension reduction on the text information; and perform semantic and syntactic analysis on the cleaned and reduced dimensional text information.
  • the feature obtaining unit is further configured to: obtain a service feature in the service feature database that matches the secondary text feature; and refer to the service feature that matches the secondary text feature.
  • the solution is decomposed to obtain the service characteristics after the digestion; the service features after the digestion are referred to as the acquired service features.
  • the apparatus for parsing semantics further includes: (not shown in the figure): a sample obtaining unit configured to acquire a sample of service information of the marked service domain; a part of speech determining unit, configured Used to perform sample-based statistics and machine studies The participle of the idiom determines the part of speech of the participle; the feature determining unit is configured to determine the business feature and the business domain to which the business feature belongs according to the frequency and part of speech of the participle.
  • the apparatus for parsing semantics further includes (not shown in the figure): a weight setting unit, configured to perform, according to a granularity of the service feature, a preset correspondence between the granularity and the weight, Set the weight of the business features in the weight library of the business feature.
  • a weight setting unit configured to perform, according to a granularity of the service feature, a preset correspondence between the granularity and the weight, Set the weight of the business features in the weight library of the business feature.
  • the primary text feature and the secondary text feature only represent that the source data of the obtained text feature is different from the obtained text feature, and the method for obtaining the text feature may be the same or different.
  • the method for obtaining the text feature may be the same or different.
  • the one-time parsed text and the second parsed text represent only the parsed source data, the method, and the resulting parsed text.
  • the method and the resulting parsed text.
  • FIG. 5 there is shown a block diagram of a computer system 500 suitable for use in implementing a terminal device or server of an embodiment of the present application.
  • computer system 500 includes a central processing unit (CPU) 501 that can be loaded into a program in random access memory (RAM) 503 according to a program stored in read only memory (ROM) 502 or from storage portion 508. And perform various appropriate actions and processes.
  • RAM random access memory
  • ROM read only memory
  • RAM 503 various programs and data required for the operation of the system 500 are also stored.
  • the CPU 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504.
  • An input/output (I/O) interface 505 is also coupled to bus 504.
  • the following components are connected to the I/O interface 505: an input portion 506 including a keyboard, a mouse, etc.; An output portion 507 such as a cathode ray tube (CRT), a liquid crystal display (LCD), or the like, a speaker, and the like; a storage portion 508 including a hard disk or the like; and a communication portion 509 including a network interface card such as a LAN card, a modem, and the like. The communication section 509 performs communication processing via a network such as the Internet.
  • Driver 510 is also coupled to I/O interface 505 as needed.
  • a removable medium 511 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory or the like is mounted on the drive 510 as needed so that a computer program read therefrom is installed into the storage portion 508 as needed.
  • an embodiment of the present disclosure includes a computer program product comprising a computer program tangibly embodied on a machine readable medium, the computer program comprising program code for executing the method illustrated in the flowchart.
  • the computer program can be downloaded and installed from the network via the communication portion 509, and/or installed from the removable medium 511.
  • each block of the flowchart or block diagrams can represent a unit, a program segment, or a portion of code that includes one or more logic for implementing the specified.
  • Functional executable instructions can also occur in a different order than that illustrated in the drawings. For example, two successively represented blocks may in fact be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowcharts, and combinations of blocks in the block diagrams and/or flowcharts can be implemented in a dedicated hardware-based system that performs the specified function or operation. Or it can be implemented by a combination of dedicated hardware and computer instructions.
  • the units involved in the embodiments of the present application may be implemented by software or by hardware.
  • the described unit may also be disposed in the processor.
  • a processor includes a semantic receiving unit, a primary parsing unit, a similarity calculating unit, a corpus determining unit, a feature extracting unit, a feature acquiring unit, and a weight acquisition.
  • Unit, field determination unit, and secondary analysis unit The name of these units does not constitute a limitation on the unit itself under certain circumstances, for example, semantic reception.
  • a unit may also be described as "a unit that receives semantic information entered by a user.”
  • the present application further provides a non-volatile computer storage medium, which may be a non-volatile computer storage medium included in the apparatus described in the foregoing embodiments; It may be a non-volatile computer storage medium that exists alone and is not assembled into the terminal.
  • the non-volatile computer storage medium stores one or more programs, when the one or more programs are executed by a device, causing the device to: receive semantic information input by a user; parse the semantic information, and obtain an associated service One-time text feature; calculating the similarity between the text feature and the corpus feature in the corpus; taking the corpus with the highest similarity of the text feature as the primary analytic text; extracting the feature of the parsed text once, and obtaining the secondary text feature; Obtaining a service feature in the service feature database that matches the secondary text feature and a service domain to which the acquired service feature belongs; obtaining a weight value associated with the acquired service feature in the weight library of the service feature; and the service feature having the highest weight value
  • the business domain belongs to the business domain to which the parsed text belongs; the parsed text is parsed once in the business domain to which the parsed text belongs, and the parsed text is obtained twice.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Machine Translation (AREA)

Abstract

一种解析语义的方法和装置。所述方法的一具体实施方式包括：接收用户输入的语义信息（201）；解析语义信息，得到关联业务的一次文本特征（202）；计算一次文本特征与语料库中的语料特征的相似度（203）；将与一次文本特征相似度最高的语料特征所属的语料作为一次解析文本（204）；提取一次解析文本的特征，得到二次文本特征（205）；获取业务特征库中与二次文本特征相匹配的业务特征及获取的业务特征所属的业务领域（206）；获取业务特征的权重库中与获取的业务特征相关联的权重值（207）；将权重值最高的业务特征所属的业务领域作为一次解析文本所属的业务领域（208）；在一次解析文本所属的业务领域中解析一次解析文本，得到二次解析文本（209）。该方法提高了解析语义的结果的准确性。

Description

解析语义的方法和装置
相关申请的交叉引用
本申请要求于2016年5月19日提交的中国专利申请号为“201610338834.2”的优先权，其全部内容作为整体并入本申请中。
技术领域
本申请涉及计算机技术领域，具体涉及互联网技术领域，尤其涉及解析语义的方法和装置。
背景技术
近几年，作为科学领域和人工智能领域一个重要方向的自然语言处理技术发展迅速，随之带动了智能机器人理论研究和应用的发展，越来越多的电子商务网站或涉及咨询业务的商业领域开始应用这一技术来代替传统的人工服务。
当前，智能机器人技术在互联网产品的应用相对比较集中，如：银行业务咨询、电子商务的售前售后业务答疑等。在这些业务中，机器人解析语义的方法通常包括：首先由人工整理业务知识，对业务知识分类，配置好问题和答案，之后在人工“监督”的情况下，“教会”机器人，最后由机器人替换人工来提供服务。
然而，在某一服务的业务点较多的情况下，采用上述机器人解析用户语义的方法，只能解析用户语义的大致意图，解析语义的精确度较低。
发明内容
本申请的目的在于提出一种改进的解析语义的方法和装置，来解决以上背景技术部分提到的技术问题。
第一方面，本申请提供了一种解析语义的方法，所述方法包括： 接收用户输入的语义信息；解析所述语义信息，得到关联业务的一次文本特征；计算所述一次文本特征与语料库中的语料特征的相似度；将与所述一次文本特征相似度最高的语料特征所属的语料作为一次解析文本；提取所述一次解析文本的特征，得到二次文本特征；获取业务特征库中与所述二次文本特征相匹配的业务特征及获取的业务特征所属的业务领域；获取业务特征的权重库中与获取的业务特征相关联的权重值；将权重值最高的业务特征所属的业务领域作为所述一次解析文本所属的业务领域；在所述一次解析文本所属的业务领域中解析所述一次解析文本，得到二次解析文本。
第二方面，本申请提供了一种解析语义的装置，所述装置包括：语义接收单元，用于接收用户输入的语义信息；一次解析单元，用于解析所述语义信息，得到关联业务的一次文本特征；相似度计算单元，用于计算所述一次文本特征与语料库中的语料特征的相似度；语料确定单元，用于将与所述一次文本特征相似度最高的语料特征所属的语料作为一次解析文本；特征提取单元，用于提取所述一次解析文本的特征，得到二次文本特征；特征获取单元，用于获取业务特征库中与所述二次文本特征相匹配的业务特征及获取的业务特征所属的业务领域；权重获取单元，用于获取业务特征的权重库中与获取的业务特征相关联的权重值；领域确定单元，用于将权重值最高的业务特征所属的业务领域作为所述一次解析文本所属的业务领域；二次解析单元，用于在所述一次解析文本所属的业务领域中解析所述一次解析文本，得到二次解析文本。
本申请提供的解析语义的方法和装置，通过对用户输入的语义信息提取关联业务的文本特征得到一次文本特征，而后计算一次文本特征与语料库中的语料特征的相似度，之后将与一次文本特征相似度最高的语料特征所属的语料作为一次解析文本，之后提取一次解析文本的特征，得到二次文本特征，之后获取业务特征库中与二次文本特征相匹配的业务特征及获取的业务特征所属的业务领域，之后获取业务特征的权重库中与获取的业务特征相关联的权重值，之后将权重值最高的业务特征所属的业务领域作为一次解析文本所属的业务领域，之 后在一次解析文本所属的业务领域中解析一次解析文本，得到二次解析文本，从而提高了解析语义的结果的准确性。
附图说明
通过阅读参照以下附图所作的对非限制性实施例所作的详细描述，本申请的其它特征、目的和优点将会变得更明显：
图1是本申请可以应用于其中的示例性系统架构图；
图2是根据本申请的解析语义的方法的一个实施例的流程图；
图3是根据本申请的解析语义的方法的一个应用场景的示意性流程图；
图4是根据本申请的解析语义的装置的一个实施例的结构示意图；
图5是适于用来实现本申请实施例的终端设备或服务器的计算机系统的结构示意图。
具体实施方式
下面结合附图和实施例对本申请作进一步的详细说明。可以理解的是，此处所描述的具体实施例仅仅用于解释相关发明，而非对该发明的限定。另外还需要说明的是，为了便于描述，附图中仅示出了与有关发明相关的部分。
需要说明的是，在不冲突的情况下，本申请中的实施例及实施例中的特征可以相互组合。下面将参考附图并结合实施例来详细说明本申请。
图1示出了可以应用本申请的解析语义的方法或解析语义的装置的实施例的示例性系统架构100。
如图1所示，系统架构100可以包括终端设备101、102、103，网络104和服务器105、106。网络104用以在终端设备101、102、103和服务器105、106之间提供通信链路的介质。网络104可以包括各种连接类型，例如有线、无线通信链路或者光纤电缆等等。
用户110可以使用终端设备101、102、103通过网络104与服务 器105、106交互，以接收或发送消息等。终端设备101、102、103上可以安装有各种通讯客户端应用，例如网页浏览器应用、购物类应用、搜索类应用、即时通信工具、邮箱客户端、社交平台软件等。
终端设备101、102、103可以是具有显示屏并且支持用户输入语义信息的各种电子设备，包括但不限于智能手机、平板电脑、MP3播放器(Moving Picture Experts Group Audio Layer III，动态影像专家压缩标准音频层面3)、MP4(Moving Picture Experts Group Audio Layer IV，动态影像专家压缩标准音频层面4)播放器、膝上型便携计算机和台式计算机等等。
服务器105、106可以是提供各种服务的服务器，例如对终端设备101、102、103上用户输入的语义信息进行解析并提供答案信息的后台服务器。后台服务器可以对接收到的用户输入的语义信息等数据进行分析等处理，并将处理结果(例如答案)反馈给终端设备。
需要说明的是，本申请实施例所提供的解析语义的方法一般由服务器105、106执行，相应地，解析语义的装置一般设置于服务器105、106中。
应该理解，图1中的终端设备、网络和服务器的数目仅仅是示意性的。根据实现需要，可以具有任意数目的终端设备、网络和服务器。
继续参考图2，示出了根据本申请的解析语义的方法的一个实施例的流程200。所述的解析语义的方法，包括以下步骤：
步骤201，接收用户输入的语义信息。
在本实施例中，解析语义的方法运行于其上的电子设备(例如图1所示的服务器)可以通过有线连接方式或者无线连接方式从用户利用其进行输入的终端接收语义信息。需要指出的是，上述无线连接方式可以包括但不限于3G/4G连接、WiFi连接、蓝牙连接、WiMAX连接、Zigbee连接、UWB(ultra wideband)连接、以及其他现在已知或将来开发的无线连接方式。
通常，用户利用终端上安装的网页浏览器应用、购物类应用、搜索类应用、即时通信工具等来浏览网页，这时，用户可以通过语音信息、文本信息和唇动信息中的一种或多种来输入语义信息。在本实施 例中，上述语义信息可以包括文字格式、音频格式、图片格式、或者其它未来将开发的格式的语义信息。
步骤202，解析语义信息，得到关联业务的一次文本特征。
在本实施例中，基于步骤201中接收的语义信息，可以对语义信息进行解析，以得到关联业务的一次文本特征。在这里，若语义信息为文本信息，则可以直接提取文本信息的特征，并在提取的特征中筛选出关联业务的一次文本特征；若语义信息为语音信息和/或唇动信息，则需要先对语音信息和/或唇动信息进行解析以得到其对应的文本信息，之后对文本信息提取特征，并在提取的特征中筛选出关联业务的一次文本特征。
在这里，可以采用现有技术中或未来发展的技术中的基于统计的特征提取方法来提取文本信息的特征。例如，可以采用词频-逆向文件频率方法、词频方法、文件频次方法等来提取文本信息的特征。这些基于统计的特征提取方法，可以通过构造评估函数来对特征集合中的每个特征进行评估，并对每个特征打分，这样每个词语都获得一个评估值，又称为权值。然后将所有特征按权值大小排序，提取预定数目的最优特征作为提取结果的特征子集。
在本实施例的一些可选实现方式中，解析语义信息，得到关联业务的一次文本特征可以包括：首先获取语义信息的文本信息；之后对文本信息进行清洗和降维；然后对清洗和降维后的文本信息进行语义和句法分析，得到文本信息特征；最后提取文本信息特征中关联业务的特征，得到一次文本特征。
在这里，对文本信息进行清洗是指发现并纠正数据文件中可识别的错误的最后一道程序，包括检查数据一致性，处理无效值和缺失值等。数据清洗的任务是过滤那些不符合要求的数据，不符合要求的数据主要是有不完整的数据、错误的数据、重复的数据三大类。例如，对文本信息中的中英文符号进行转换、对简繁体进行转换以及清洗重复数据、冲突数据及无意义字符等，这里的无意义字符是指有不影响分类的特征，如“你好”、“请问”以及脏话等。对文本信息进行降维是将样本点从输入空间通过线性或非线性变换映射到一个低维空间， 从而获得一个关于原数据集紧致的低维表示。例如采用以下方法对文本信息进行降维：缺失值比率法、高相关滤波法、主成分分析法等。
在这里，对清洗和降维后的文本信息进行语义和句法分析，可以得到标记多维度的文本信息特征。这里的句法分析，是指对句法结构中词语与词语之间发生的种种语法关系进行分析：主谓、述补、述宾、偏正、联合等。这里的语义关系，在语义学中是指语言成分所表示的客观对象之间的关系，我们所说的语义关系是指动作和动作者、动作和受动者、动作和工具、动作和处所、事物和事物之间的关系等。在进行语义和句法分析时，可以采用预先建立的分词模型来完成分析，训练分词模型的数据可以为包含业务知识的历史问答数据，也可以是人工添加的数据。
在这里，提取文本信息特征中关联业务的特征，是指在文本信息特征的基础上，放弃一部分无业务特征的文本特征词并提高另一部分文本特征词的权重。
步骤203，计算一次文本特征与语料库中的语料特征的相似度。
在本实施例中，语料库是指预先由人工标记的书写文本或口语转写的集合体，可以用于语言分析和语言描写。语料库也可以认为是电子机读文本的有限集合，从最大程度上代表了某种语言或语言变体。计算一次文本特征与语料库中的语料特征的相似度，可以确定与一次文本特征相似度最高的语料特征所属的语料，也即一次解析文本。
步骤204，将与一次文本特征相似度最高的语料特征所属的语料作为一次解析文本。
在本实施例中，基于步骤203中计算得到的一次文本特征与语料库中的语料特征的相似度，可以选择与一次文本特征相似度最高的语料特征所属的语料，并将选择的语料作为一次解析文本。
步骤205，提取一次解析文本的特征，得到二次文本特征。
在本实施例中，在步骤204得到的一次解析文本的基础上，可以提取一次解析文本的特征，并将提取的一次解析文本的特征作为二次文本特征。
在这里，提取一次解析文本的特征的方法，与上述的提取文本特 征的方法可以相同，也可以不同，可以采用现有技术中或未来发展的技术中的基于统计的特征提取方法来提取一次解析文本的特征。例如，可以采用词频-逆向文件频率方法、词频方法、文件频次方法等来提取一次解析文本的特征。这些基于统计的特征提取方法，可以通过构造评估函数来对特征集合中的每个特征进行评估，并对每个特征打分，这样每个词语都获得一个评估值，又称为权值。然后将所有特征按权值大小排序，提取预定数目的最优特征作为提取结果的特征子集。
在本实施例的一些可选实现方式中，提取一次解析文本的特征，得到二次文本特征可以包括：首先获取一次解析文本；之后对一次解析文本进行清洗和降维；然后对清洗和降维后的一次解析文本进行语义和句法分析，得到二次文本特征。
这里的清洗和降维，同上述对文本信息进行清洗和降维的方法相同，在此不再赘述。
步骤206，获取业务特征库中与二次文本特征相匹配的业务特征及获取的业务特征所属的业务领域。
在本实施例中，业务特征库中包括预先从用户问题中提取的业务特征以及该特征所属的业务领域。基于步骤205中得到的二次文本特征，可以在业务特征库中获取与二次文本特征相匹配的业务特征，并获取该业务特征所属的业务领域。应当理解，业务特征库中可能有一个或多个业务领域的业务特征与二次文本特征相匹配。
在本实施例的一些可选实现方式中，为了获取精简的业务特征，可以在二次文本特征与业务特征库中的特征进行匹配后对获取的业务特征进行指代消解，也即获取业务特征库中与二次文本特征相匹配的业务特征可以包括：首先获取业务特征库中与二次文本特征相匹配的业务特征；之后对与二次文本特征相匹配的业务特征进行指代消解，得到指代消解后的业务特征；最后将指代消解后的业务特征作为获取的业务特征。
在这里，指代消解的规则可以包括以下几种情况：
1、包含代词，如：“这个”、“这款”、“该款”、“本款”，且代词修饰名词、动名词，消解代词；2、业务词是名词、动名词，前有形容词、 动名词，消解掉形容词、动名词；3、业务词是名词、动名词，前有动词，后面有动词，则前面动词消解掉；4、业务词是动词，且句子里面只包含一个名词，一个动词，则消解掉其他词性词，如：“系统/n是不是兼容/v(业务词)很好”消解为“系统兼容”；5、业务词是名词，且句子里面只有一个名词，则消解掉其他词性词，如：“这个尺寸/n(业务词)是多少”消解为“尺寸”；6、业务词是形容词，如果前后都是业务词，比较权重，留权重大的业务词。
在本实施例的一些可选实现方式中，上述解析语义的方法还包括：可以通过以下步骤获取确定业务特征库中的业务特征及业务特征所属的业务领域：获取已标记业务领域的业务信息的样本；对样本进行基于统计和机器学习的分词并确定分词的词性；根据分词出现的频率和词性，确定业务特征库中的业务特征及业务特征所属的业务领域。例如，在业务领域为尺寸的模型中，文本“商品/n的/uj尺寸/n是/v多大/a”和“宝贝/n的/uj厚度/n是/v多少/r”，其中，n为名词，nj为结构助词的，v是动词，a是形容词，因此这里的“尺寸”、“厚度”即为业务特征。然后，分析业务词之间的从属关系，如，{厚度}∈{尺寸}，因此业务特征所属的业务领域为尺寸。
步骤207，获取业务特征的权重库中与获取的业务特征相关联的权重值。
在本实施例中，对应于步骤206中获取的业务特征为一个，可以在业务特征的权重库中获取该业务特征对应的权重，对应于步骤206中获取的业务特征为多个，可以在业务特征的权重库中对应每一个业务特征分别获取其对应的权重。
这里的权重库，为预先根据业务特征及业务特征所属的业务领域预设的权重。在预设权重时，通常考虑给那些能表达主题的业务特征赋予较高的权重，在赋予业务特征权重时，可以充分考虑业务特征的词频、词性、词长、位置以及同义词现象等因素。例如，将意义相同或相近的词视为一个，并将它们的词频相加；赋予名词或者含有名词的组合词较高的权重；赋予4个或者4个字以上的词较高的权重。
在本实施例的一些可选实现方式中，上述解析语义的方法还包括： 根据业务特征的粒度和预设的粒度与权重的对应关系，设定业务特征的权重库中业务特征的权重。
这里的粒度，是指业务特征的细化和综合程度。例如，可以根据业务特征的划分标准，可以将业务特征的权重库中的权重划分为：详细数据、轻度总结、高度总结三级。在确定数据粒度时应注意一条原则：细化程度越高，粒度越小；细化程度越低，粒度越大。
在确定业务特征的粒度之后，可以根据预设的粒度与权重的对应关系，设定业务特征的权重库中业务特征的权重。
步骤208，将权重值最高的业务特征所属的业务领域作为一次解析文本所属的业务领域。
在本实施例中，若步骤207中获取的业务特征的权重为单个，则将该业务特征所属的业务领域作为一次解析文本所属的业务领域；若步骤207中获取的业务特征的权重为多个，则对多个权重从大到小进行排序，选择排序第一的权重对应的业务特征所属的业务领域作为一次解析文本所属的业务领域，从而更为精确的确定了用户意图所属的领域。
步骤209，在一次解析文本所属的业务领域中解析一次解析文本，得到二次解析文本。
在本实施例中，基于步骤208中获取的一次解析文本所属的业务领域，将一次解析文本进行基于统计的分词，然后将分词与一次解析文本所属的业务领域中的特征进行比对，得到一次解析文本在其所属的业务领域内的二次解析文本。
在本实施例的一些可选实现方式中，在得到二次解析文本之后，还可以根据二次解析文本获取答复二次解析文本的答案：首先根据一次解析文本所属的业务领域确定答案集中对应该业务领域的答案子集，之后在该答案子集中查询与二次解析文本对应的问题，之后获取对应该问题的答案，并将该答案作为二次解析文本对应的答案，也即上述的解析语义的方法还包括：在一次解析文本所属的业务领域对应的答案集中查询与二次解析文本对应的答案；呈现与二次解析文本对应的答案。
应当理解，这里的答案集通常可以包括预先确定的附有领域标记的问答对，问答对包括问题以及对应问题的答案，问答对可以来自于官网常见的问题项目与对应问题的解答，也可以来自于从人工客服的聊天记录筛选的问答对，还可以是人工根据问题编写的问答对。
继续参见图3，图3是根据本实施例的解析语义的方法的应用场景的一个示意性流程图。
如图3所示，解析语义的方法包括以下步骤：
在步骤301中，接收用户问题，之后执行步骤302；
在步骤302中，对接收的用户问题的文本信息进行数据清洗和降维，之后执行步骤303；
在步骤303中，对清洗和降维后的文本信息进行意图识别：对清洗和降维后的文本信息进行语义和句法分析，得到文本信息特征，提取文本信息中关联业务的特征，得到一次文本特征，之后计算一次文本特征与语料库中的语料特征的相似度，将与一次文本特征相似度最高的语料特征所属的语料作为意图识别的结果，之后执行步骤307；
在步骤304中，将上述意图识别的结果作为一次解析文本；
在步骤305中，离线计算包括业务特征和业务领域的业务特征库以及包括业务特征和权重的业务特征的权重库；
在步骤306中，从业务特征库中提取业务特征及业务特征的权重；
在步骤307中，对一次解析文本进行基于统计的分词，之后匹配分词和提取的业务特征，若匹配，则记录该分词的位置，同时，搜索和判断该分词与前面相邻分词的词性类型是否一致或者可消解，如都是代词、形容词、动名词等，如果是，则从该词位置起，到搜索结束的位置合并为一个短语，消解为关键业务词，并附对应权重；当消解到文本结尾，重新合并消解后的文本，将权重值最高的业务特征所属的业务领域作为一次解析文本所属的业务领域，并为合并的文本标记业务领域，之后执行步骤309和310；
在步骤308中，预先确定答案集，答案集包括附有领域标记的问答对；
在步骤309中，将一次解析文本在合并后的文本所标记的业务领 域中重新进行解析，得到二次解析文本，之后执行步骤311；
在步骤310中，从步骤308中确定的答案集中提取符合上述合并的文本所标记的业务领域的答案子集，之后执行步骤311；
在步骤311中，从答案子集中查询与二次解析文本的特征对应的答案特征，并将答案特征组装为答案，之后执行步骤312；
在步骤312中，返回组装的答案。
在与上述图3的流程图对应的一个应用场景中，首先接收用户输入的“你好，请问这款红石榴免洗睡眠面膜可以改善暗黄的肤质吗？”经过数据清洗、降维以及与语料库的匹配，输入变为一次解析文本“这款红石榴免洗睡眠面膜可以改善暗黄的肤质？”，之后分词，搜索到业务词，分别有“面膜{w＝0.48}”、“暗黄{w＝0.80}”和“肤质{w＝0.6}”，“红”是形容词，修饰“石榴”，则合并“红石榴”，前面是代词“这款”，则消解“这款红石榴”为“红石榴”，业务词“面膜”前有动词(免洗、睡眠)和名词(红石榴)，则消解掉变成“面膜”，形容词“暗黄”修饰“肤质”，因为一次解析文本所属的业务领域是“产品功效”，“暗黄”的权重高于“肤质”，则“暗黄的肤质”消解为“暗黄”，最终分析的句子为“面膜{w＝0.48}可以改善暗黄{w＝0.80}”，之后根据“面膜{w＝0.48}可以改善暗黄{w＝0.80}”，在产品功效领域中查询对应的答案，并返回组装的答案。
本申请的上述实施例提供的方法通过确定一次解析文本所属的业务领域以及在细分的业务领域中解析一次解析文本以得到二次解析文本，提高了解析语义的结果的准确性。
进一步参考图4，作为对上述各图所示方法的实现，本申请提供了一种解析语义的装置的一个实施例，该装置实施例与图2所示的方法实施例相对应，该装置具体可以应用于各种电子设备中。
如图4所示，本实施例所述的解析语义的装置400包括：语义接收单元410、一次解析单元420、相似度计算单元430、语料确定单元440、特征提取单元450、特征获取单元460、权重获取单元470、领域确定单元480和二次解析单元490。
其中，语义接收单元410，配置用于接收用户输入的语义信息； 一次解析单元420，配置用于解析语义信息，得到关联业务的一次文本特征；相似度计算单元430，配置用于计算一次文本特征与语料库中的语料特征的相似度；语料确定单元440，配置用于将与一次文本特征相似度最高的语料特征所属的语料作为一次解析文本；特征提取单元450，配置用于提取一次解析文本的特征，得到二次文本特征；特征获取单元460，配置用于获取业务特征库中与二次文本特征相匹配的业务特征及获取的业务特征所属的业务领域；权重获取单元470，配置用于获取业务特征的权重库中与获取的业务特征相关联的权重值；领域确定单元480，配置用于将权重值最高的业务特征所属的业务领域作为一次解析文本所属的业务领域；二次解析单元490，配置用于在一次解析文本所属的业务领域中解析一次解析文本，得到二次解析文本。
在本实施例的一些可选实现方式中，上述解析语义的装置还包括(图中未示出)：答案查询单元，配置用于在一次解析文本所属的业务领域对应的答案集中查询与二次解析文本对应的答案；答案呈现单元，配置用于呈现与二次解析文本对应的答案。
在本实施例的一些可选实现方式中，一次解析单元进一步配置用于：获取语义信息的文本信息；对文本信息进行清洗和降维；对清洗和降维后的文本信息进行语义和句法分析，得到文本信息特征，提取文本信息特征中关联业务的特征，得到一次文本特征；和/或特征提取单元进一步配置用于：获取一次解析文本；对一次解析文本进行清洗和降维；对清洗和降维后的一次解析文本进行语义和句法分析，得到二次文本特征。
在本实施例的一些可选实现方式中，特征获取单元进一步配置用于：获取业务特征库中与二次文本特征相匹配的业务特征；对与二次文本特征相匹配的业务特征进行指代消解，得到指代消解后的业务特征；将指代消解后的业务特征作为获取的业务特征。
在本实施例的一些可选实现方式中，上述解析语义的装置还包括(图中未示出)：样本获取单元，配置用于获取已标记业务领域的业务信息的样本；词性确定单元，配置用于对样本进行基于统计和机器学 习的分词并确定分词的词性；特征确定单元，配置用于根据分词出现的频率和词性，确定业务特征库中的业务特征及业务特征所属的业务领域。
在本实施例的一些可选实现方式中，上述解析语义的装置还包括(图中未示出)：权重设定单元，用于根据业务特征的粒度和预设的粒度与权重的对应关系，设定业务特征的权重库中业务特征的权重。
本领域技术人员可以理解，上述解析语义的装置400还包括一些其他公知结构，例如处理器、存储器等，为了不必要地模糊本公开的实施例，这些公知的结构在图4中未示出。
应当理解，装置400中记载的诸单元与参考图2描述的方法中的各个步骤相对应。由此，上文针对解析语义的方法描述的操作和特征同样适用于装置400及其中包含的单元，在此不再赘述。装置400中的相应单元可以与终端设备和/或服务器中的单元相互配合以实现本申请实施例的方案。
在本申请的上述实施例中，一次文本特征及二次文本特征仅代表获得文本特征的源数据和得到的文本特征不同，获得文本特征的方法可以相同，也可以不同。本领域技术人员应当理解，其中的一次或二次并不构成对文本特征的特殊限定。
在本申请的上述实施例中，一次解析文本及二次解析文本仅代表解析的源数据、方法以及得到的解析文本不同。本领域技术人员应当理解，其中的一次或二次并不构成对解析文本的特殊限定。
下面参考图5，其示出了适于用来实现本申请实施例的终端设备或服务器的计算机系统500的结构示意图。
如图5所示，计算机系统500包括中央处理单元(CPU)501，其可以根据存储在只读存储器(ROM)502中的程序或者从存储部分508加载到随机访问存储器(RAM)503中的程序而执行各种适当的动作和处理。在RAM 503中，还存储有系统500操作所需的各种程序和数据。CPU 501、ROM 502以及RAM 503通过总线504彼此相连。输入/输出(I/O)接口505也连接至总线504。
以下部件连接至I/O接口505：包括键盘、鼠标等的输入部分506； 包括诸如阴极射线管(CRT)、液晶显示器(LCD)等以及扬声器等的输出部分507；包括硬盘等的存储部分508；以及包括诸如LAN卡、调制解调器等的网络接口卡的通信部分509。通信部分509经由诸如因特网的网络执行通信处理。驱动器510也根据需要连接至I/O接口505。可拆卸介质511，诸如磁盘、光盘、磁光盘、半导体存储器等等，根据需要安装在驱动器510上，以便于从其上读出的计算机程序根据需要被安装入存储部分508。
特别地，根据本公开的实施例，上文参考流程图描述的过程可以被实现为计算机软件程序。例如，本公开的实施例包括一种计算机程序产品，其包括有形地包含在机器可读介质上的计算机程序，所述计算机程序包含用于执行流程图所示的方法的程序代码。在这样的实施例中，该计算机程序可以通过通信部分509从网络上被下载和安装，和/或从可拆卸介质511被安装。
附图中的流程图和框图，图示了按照本申请各种实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个单元、程序段、或代码的一部分，所述单元、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或操作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
描述于本申请实施例中所涉及到的单元可以通过软件的方式实现，也可以通过硬件的方式来实现。所描述的单元也可以设置在处理器中，例如，可以描述为：一种处理器包括语义接收单元、一次解析单元、相似度计算单元、语料确定单元、特征提取单元、特征获取单元、权重获取单元、领域确定单元和二次解析单元。其中，这些单元的名称在某种情况下并不构成对该单元本身的限定，例如，语义接收 单元还可以被描述为“接收用户输入的语义信息的单元”。
作为另一方面，本申请还提供了一种非易失性计算机存储介质，该非易失性计算机存储介质可以是上述实施例中所述装置中所包含的非易失性计算机存储介质；也可以是单独存在，未装配入终端中的非易失性计算机存储介质。上述非易失性计算机存储介质存储有一个或者多个程序，当所述一个或者多个程序被一个设备执行时，使得所述设备：接收用户输入的语义信息；解析语义信息，得到关联业务的一次文本特征；计算一次文本特征与语料库中的语料特征的相似度；将与一次文本特征相似度最高的语料特征所属的语料作为一次解析文本；提取一次解析文本的特征，得到二次文本特征；获取业务特征库中与二次文本特征相匹配的业务特征及获取的业务特征所属的业务领域；获取业务特征的权重库中与获取的业务特征相关联的权重值；将权重值最高的业务特征所属的业务领域作为一次解析文本所属的业务领域；在一次解析文本所属的业务领域中解析一次解析文本，得到二次解析文本。
以上描述仅为本申请的较佳实施例以及对所运用技术原理的说明。本领域技术人员应当理解，本申请中所涉及的发明范围，并不限于上述技术特征的特定组合而成的技术方案，同时也应涵盖在不脱离所述发明构思的情况下，由上述技术特征或其等同特征进行任意组合而形成的其它技术方案。例如上述特征与本申请中公开的(但不限于)具有类似功能的技术特征进行互相替换而形成的技术方案。

Claims (14)

  1. 一种解析语义的方法，其特征在于，所述方法包括：
    接收用户输入的语义信息；
    解析所述语义信息，得到关联业务的一次文本特征；
    计算所述一次文本特征与语料库中的语料特征的相似度；
    将与所述一次文本特征相似度最高的语料特征所属的语料作为一次解析文本；
    提取所述一次解析文本的特征，得到二次文本特征；
    获取业务特征库中与所述二次文本特征相匹配的业务特征及获取的业务特征所属的业务领域；
    获取业务特征的权重库中与获取的业务特征相关联的权重值；
    将权重值最高的业务特征所属的业务领域作为所述一次解析文本所属的业务领域；
    在所述一次解析文本所属的业务领域中解析所述一次解析文本，得到二次解析文本。
  2. 根据权利要求1所述的解析语义的方法，其特征在于，所述方法还包括：
    在所述一次解析文本所属的业务领域对应的答案集中查询与所述二次解析文本对应的答案；
    呈现所述与所述二次解析文本对应的答案。
  3. 根据权利要求1所述的解析语义的方法，其特征在于，所述解析所述语义信息，得到关联业务的一次文本特征包括：获取所述语义信息的文本信息；对所述文本信息进行清洗和降维；对清洗和降维后的文本信息进行语义和句法分析，得到文本信息特征，提取所述文本信息特征中关联业务的特征，得到所述一次文本特征；和/或
    所述提取所述一次解析文本的特征，得到二次文本特征包括：获取所述一次解析文本；对所述一次解析文本进行清洗和降维；对清洗 和降维后的一次解析文本进行语义和句法分析，得到所述二次文本特征。
  4. 根据权利要求1或3所述的解析语义的方法，其特征在于，所述获取业务特征库中与所述二次文本特征相匹配的业务特征包括：
    获取业务特征库中与所述二次文本特征相匹配的业务特征；
    对所述与所述二次文本特征相匹配的业务特征进行指代消解，得到指代消解后的业务特征；
    将指代消解后的业务特征作为获取的业务特征。
  5. 根据权利要求4所述的解析语义的方法，其特征在于，所述方法还包括：
    获取已标记业务领域的业务信息的样本；
    对所述样本进行基于统计和机器学习的分词并确定所述分词的词性；
    根据所述分词出现的频率和所述词性，确定所述业务特征库中的业务特征及业务特征所属的业务领域。
  6. 根据权利要求5所述的解析语义的方法，其特征在于，所述方法还包括：
    根据业务特征的粒度和预设的粒度与权重的对应关系，设定所述业务特征的权重库中业务特征的权重。
  7. 一种解析语义的装置，其特征在于，所述装置包括：
    语义接收单元，用于接收用户输入的语义信息；
    一次解析单元，用于解析所述语义信息，得到关联业务的一次文本特征；
    相似度计算单元，用于计算所述一次文本特征与语料库中的语料特征的相似度；
    语料确定单元，用于将与所述一次文本特征相似度最高的语料特 征所属的语料作为一次解析文本；
    特征提取单元，用于提取所述一次解析文本的特征，得到二次文本特征；
    特征获取单元，用于获取业务特征库中与所述二次文本特征相匹配的业务特征及获取的业务特征所属的业务领域；
    权重获取单元，用于获取业务特征的权重库中与获取的业务特征相关联的权重值；
    领域确定单元，用于将权重值最高的业务特征所属的业务领域作为所述一次解析文本所属的业务领域；
    二次解析单元，用于在所述一次解析文本所属的业务领域中解析所述一次解析文本，得到二次解析文本。
  8. 根据权利要求7所述的解析语义的装置，其特征在于，所述装置还包括：
    答案查询单元，用于在所述一次解析文本所属的业务领域对应的答案集中查询与所述二次解析文本对应的答案；
    答案呈现单元，用于呈现所述与所述二次解析文本对应的答案。
  9. 根据权利要求7所述的解析语义的装置，其特征在于，所述一次解析单元进一步用于：获取所述语义信息的文本信息；对所述文本信息进行清洗和降维；对清洗和降维后的文本信息进行语义和句法分析，得到文本信息特征，提取所述文本信息特征中关联业务的特征，得到所述一次文本特征；和/或
    所述特征提取单元进一步用于：获取所述一次解析文本；对所述一次解析文本进行清洗和降维；对清洗和降维后的一次解析文本进行语义和句法分析，得到所述二次文本特征。
  10. 根据权利要求7或9所述的解析语义的装置，其特征在于，所述特征获取单元进一步用于：获取业务特征库中与所述二次文本特征相匹配的业务特征；对所述与所述二次文本特征相匹配的业务特征 进行指代消解，得到指代消解后的业务特征；将指代消解后的业务特征作为获取的业务特征。
  11. 根据权利要求10所述的解析语义的装置，其特征在于，所述装置还包括：
    样本获取单元，用于获取已标记业务领域的业务信息的样本；
    词性确定单元，用于对所述样本进行基于统计和机器学习的分词并确定所述分词的词性；
    特征确定单元，用于根据所述分词出现的频率和所述词性，确定所述业务特征库中的业务特征及业务特征所属的业务领域。
  12. 根据权利要求11所述的解析语义的装置，其特征在于，所述装置还包括：
    权重设定单元，用于根据业务特征的粒度和预设的粒度与权重的对应关系，设定所述业务特征的权重库中业务特征的权重。
  13. 一种设备，包括：
    处理器；和
    存储器，
    所述存储器中存储有能够被所述处理器执行的计算机可读指令，在所述计算机可读指令被执行时，所述处理器执行如权利要求1-6中任一项所述的方法。
  14. 一种非易失性计算机存储介质，所述计算机存储介质存储有能够被处理器执行的计算机可读指令，当所述计算机可读指令被处理器执行时，所述处理器执行如权利要求1-6中任一项所述的方法。
PCT/CN2017/081197 2016-05-19 2017-04-20 解析语义的方法和装置 Ceased WO2017198031A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
RU2018142588A RU2704531C1 (ru) 2016-05-19 2017-04-20 Способ и устройство для анализа семантической информации
US16/303,000 US10824816B2 (en) 2016-05-19 2017-04-20 Semantic parsing method and apparatus

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610338834.2 2016-05-19
CN201610338834.2A CN107402912B (zh) 2016-05-19 2016-05-19 解析语义的方法和装置

Publications (1)

Publication Number Publication Date
WO2017198031A1 true WO2017198031A1 (zh) 2017-11-23

Family

ID=60324770

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/081197 Ceased WO2017198031A1 (zh) 2016-05-19 2017-04-20 解析语义的方法和装置

Country Status (4)

Country Link
US (1) US10824816B2 (zh)
CN (1) CN107402912B (zh)
RU (1) RU2704531C1 (zh)
WO (1) WO2017198031A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109829045A (zh) * 2018-12-30 2019-05-31 贝壳技术有限公司 一种问答方法和装置
CN120354147A (zh) * 2025-04-08 2025-07-22 山东联数信息科技有限公司 一种基于多类型数据库文件的多维数据匹配训练处理方法

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10529323B2 (en) * 2017-05-19 2020-01-07 UBTECH Robotics Corp. Semantic processing method of robot and semantic processing device
CN109035001A (zh) * 2018-06-28 2018-12-18 招联消费金融有限公司 智能语音应答处理方法及系统
CN110837356B (zh) * 2018-08-15 2024-08-20 北京京东尚科信息技术有限公司 一种数据处理方法和装置
CN109033457A (zh) * 2018-08-29 2018-12-18 广州中赢财富信息科技有限公司 多源数据库关联的审计方法及系统
CN109830231B (zh) * 2018-12-17 2021-03-16 北京百度网讯科技有限公司 会话处理方法、装置和存储介质
CN111984766B (zh) * 2019-05-21 2023-02-24 华为技术有限公司 缺失语义补全方法及装置
CN110674363B (zh) * 2019-08-30 2022-04-22 中国人民财产保险股份有限公司 接口服务之间的相似匹配方法、装置及电子设备
CN111309903B (zh) * 2020-01-20 2023-06-16 北京大米未来科技有限公司 一种数据处理方法、装置、存储介质和电子设备
CN114139537B (zh) * 2021-11-30 2025-05-09 杭州数梦工场科技有限公司 词向量的生成方法及装置
CN114118062B (zh) * 2021-12-15 2025-04-04 广州小鹏汽车科技有限公司 客户特征提取方法、装置、电子设备及存储介质
CN115981554A (zh) * 2022-12-20 2023-04-18 平安科技（深圳）有限公司 存储内存降低方法、装置、电子设备及介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120078062A1 (en) * 2010-09-24 2012-03-29 International Business Machines Corporation Decision-support application and system for medical differential-diagnosis and treatment using a question-answering system
CN104615755A (zh) * 2015-02-12 2015-05-13 北京航空航天大学 一种基于众包的新型问答系统
CN104657463A (zh) * 2015-02-10 2015-05-27 乐娟 应用于自动问答系统的问句分类方法及装置
CN104699708A (zh) * 2013-12-09 2015-06-10 中国移动通信集团北京有限公司 一种客服机器人的自学习方法及装置
CN105354300A (zh) * 2015-11-05 2016-02-24 上海智臻智能网络科技股份有限公司 一种信息推荐方法及装置

Family Cites Families (56)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5819258A (en) * 1997-03-07 1998-10-06 Digital Equipment Corporation Method and apparatus for automatically generating hierarchical categories from large document collections
WO2000046701A1 (en) * 1999-02-08 2000-08-10 Huntsman Ici Chemicals Llc Method for retrieving semantically distant analogies
US6766316B2 (en) * 2001-01-18 2004-07-20 Science Applications International Corporation Method and system of ranking and clustering for document indexing and retrieval
US7024400B2 (en) * 2001-05-08 2006-04-04 Sunflare Co., Ltd. Differential LSI space-based probabilistic document classifier
US20030066025A1 (en) * 2001-07-13 2003-04-03 Garner Harold R. Method and system for information retrieval
US8612208B2 (en) * 2004-04-07 2013-12-17 Oracle Otc Subsidiary Llc Ontology for use with a system, method, and computer readable medium for retrieving information and response to a query
CN101431573B (zh) * 2007-11-08 2013-02-20 上海撼世网络科技有限公司 通过人机交互技术实现自动客户服务的方法和设备
US9646078B2 (en) * 2008-05-12 2017-05-09 Groupon, Inc. Sentiment extraction from consumer reviews for providing product recommendations
US9519636B2 (en) * 2008-10-15 2016-12-13 Business Objects S.A. Deduction of analytic context based on text and semantic layer
RU2399959C2 (ru) * 2008-10-29 2010-09-20 Закрытое акционерное общество "Авикомп Сервисез" Способ автоматизированной обработки текста на естественном языке путем его семантической индексации, способ автоматизированной обработки коллекции текстов на естественном языке путем их семантической индексации и машиночитаемые носители
US8335754B2 (en) * 2009-03-06 2012-12-18 Tagged, Inc. Representing a document using a semantic structure
US9183288B2 (en) * 2010-01-27 2015-11-10 Kinetx, Inc. System and method of structuring data for search using latent semantic analysis techniques
US9298818B1 (en) * 2010-05-28 2016-03-29 Sri International Method and apparatus for performing semantic-based data analysis
US20170262529A1 (en) * 2010-10-01 2017-09-14 Google Inc. Sponsor answers and user-approved, system-suggested links in a social search engine
WO2012088591A1 (en) * 2010-12-30 2012-07-05 Primal Fusion Inc. System and method for performing a semantic operation on a digital social network
US9400778B2 (en) * 2011-02-01 2016-07-26 Accenture Global Services Limited System for identifying textual relationships
US9201868B1 (en) * 2011-12-09 2015-12-01 Guangsheng Zhang System, methods and user interface for identifying and presenting sentiment information
JP5825676B2 (ja) * 2012-02-23 2015-12-02 国立研究開発法人情報通信研究機構 ノン・ファクトイド型質問応答システム及びコンピュータプログラム
US9037452B2 (en) * 2012-03-16 2015-05-19 Afrl/Rij Relation topic construction and its application in semantic relation extraction
US9575963B2 (en) * 2012-04-20 2017-02-21 Maluuba Inc. Conversational agent
CN103425640A (zh) * 2012-05-14 2013-12-04 华为技术有限公司 一种多媒体问答系统及方法
US20140006012A1 (en) * 2012-07-02 2014-01-02 Microsoft Corporation Learning-Based Processing of Natural Language Questions
US8577671B1 (en) * 2012-07-20 2013-11-05 Veveo, Inc. Method of and system for using conversation state information in a conversational interaction system
US10169456B2 (en) * 2012-08-14 2019-01-01 International Business Machines Corporation Automatic determination of question in text and determination of candidate responses using data mining
US10031968B2 (en) * 2012-10-11 2018-07-24 Veveo, Inc. Method for adaptive conversation state management with filtering operators applied dynamically as part of a conversational interface
US10055418B2 (en) * 2014-03-14 2018-08-21 Highspot, Inc. Narrowing information search results for presentation to a user
US20140344261A1 (en) * 2013-05-20 2014-11-20 Chacha Search, Inc Method and system for analyzing a request
US10346621B2 (en) * 2013-05-23 2019-07-09 yTrre, Inc. End-to-end situation aware operations solution for customer experience centric businesses
US9996588B2 (en) * 2013-12-09 2018-06-12 International Business Machines Corporation Managing a search
US11921715B2 (en) * 2014-01-27 2024-03-05 Microstrategy Incorporated Search integration
US9286910B1 (en) * 2014-03-13 2016-03-15 Amazon Technologies, Inc. System for resolving ambiguous queries based on user context
CN103970898A (zh) * 2014-05-27 2014-08-06 重庆大学 一种基于多级规则库的信息提取方法及装置
US9858261B2 (en) * 2014-06-23 2018-01-02 International Business Machines Corporation Relation extraction using manifold models
US9740771B2 (en) * 2014-09-26 2017-08-22 International Business Machines Corporation Information handling system and computer program product for deducing entity relationships across corpora using cluster based dictionary vocabulary lexicon
CN104410662A (zh) * 2014-10-23 2015-03-11 山东大学 物联网并行海量数据传输中间件及其工作方法
US10509814B2 (en) * 2014-12-19 2019-12-17 Universidad Nacional De Educacion A Distancia (Uned) System and method for the indexing and retrieval of semantically annotated data using an ontology-based information retrieval model
CN104598532A (zh) * 2014-12-29 2015-05-06 中国联合网络通信有限公司广东省分公司 一种信息处理方法及装置
US20160196490A1 (en) * 2015-01-02 2016-07-07 International Business Machines Corporation Method for Recommending Content to Ingest as Corpora Based on Interaction History in Natural Language Question and Answering Systems
US10460720B2 (en) * 2015-01-03 2019-10-29 Microsoft Technology Licensing, Llc. Generation of language understanding systems and methods
US9588959B2 (en) * 2015-01-09 2017-03-07 International Business Machines Corporation Extraction of lexical kernel units from a domain-specific lexicon
US9984310B2 (en) * 2015-01-23 2018-05-29 Highspot, Inc. Systems and methods for identifying semantically and visually related content
US9514124B2 (en) * 2015-02-05 2016-12-06 International Business Machines Corporation Extracting and recommending business processes from evidence in natural language systems
US10033714B2 (en) * 2015-06-16 2018-07-24 Business Objects Software, Ltd Contextual navigation facets panel
US10783179B2 (en) * 2015-08-03 2020-09-22 International Business Machines Corporation Automated article summarization, visualization and analysis using cognitive services
CN105205104A (zh) * 2015-08-26 2015-12-30 成都布林特信息技术有限公司 一种云平台数据获取方法
US11663254B2 (en) * 2016-01-29 2023-05-30 Thomson Reuters Enterprise Centre Gmbh System and engine for seeded clustering of news events
US10565309B2 (en) * 2016-02-17 2020-02-18 International Business Machines Corporation Interpreting the meaning of clinical values in electronic medical records
CN105653738B (zh) * 2016-03-01 2020-05-22 北京百度网讯科技有限公司 基于人工智能的搜索结果播报方法和装置
US10489393B1 (en) * 2016-03-30 2019-11-26 Amazon Technologies, Inc. Quasi-semantic question answering
US10332508B1 (en) * 2016-03-31 2019-06-25 Amazon Technologies, Inc. Confidence checking for speech processing and query answering
US10388274B1 (en) * 2016-03-31 2019-08-20 Amazon Technologies, Inc. Confidence checking for speech processing and query answering
US10585898B2 (en) * 2016-05-12 2020-03-10 International Business Machines Corporation Identifying nonsense passages in a question answering system based on domain specific policy
US10210201B2 (en) * 2016-05-13 2019-02-19 TCL Research America Inc. Method and system for App page recommendation via inference of implicit intent in a user query
US10503767B2 (en) * 2016-09-13 2019-12-10 Microsoft Technology Licensing, Llc Computerized natural language query intent dispatching
US10540513B2 (en) * 2016-09-13 2020-01-21 Microsoft Technology Licensing, Llc Natural language processor extension transmission data protection
US20180089681A1 (en) * 2016-09-27 2018-03-29 Marie Fenimore System and method for compliance screening

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120078062A1 (en) * 2010-09-24 2012-03-29 International Business Machines Corporation Decision-support application and system for medical differential-diagnosis and treatment using a question-answering system
CN104699708A (zh) * 2013-12-09 2015-06-10 中国移动通信集团北京有限公司 一种客服机器人的自学习方法及装置
CN104657463A (zh) * 2015-02-10 2015-05-27 乐娟 应用于自动问答系统的问句分类方法及装置
CN104615755A (zh) * 2015-02-12 2015-05-13 北京航空航天大学 一种基于众包的新型问答系统
CN105354300A (zh) * 2015-11-05 2016-02-24 上海智臻智能网络科技股份有限公司 一种信息推荐方法及装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109829045A (zh) * 2018-12-30 2019-05-31 贝壳技术有限公司 一种问答方法和装置
CN120354147A (zh) * 2025-04-08 2025-07-22 山东联数信息科技有限公司 一种基于多类型数据库文件的多维数据匹配训练处理方法

Also Published As

Publication number Publication date
CN107402912A (zh) 2017-11-28
US20190171713A1 (en) 2019-06-06
RU2704531C1 (ru) 2019-10-29
CN107402912B (zh) 2019-12-31
US10824816B2 (en) 2020-11-03

Similar Documents

Publication Publication Date Title
WO2017198031A1 (zh) 解析语义的方法和装置
US20220318275A1 (en) Search method, electronic device and storage medium
CN102866989B (zh) 基于词语依存关系的观点抽取方法
US9183274B1 (en) System, methods, and data structure for representing object and properties associations
CN112417846B (zh) 文本自动化生成方法、装置、电子设备及存储介质
CN111143576A (zh) 一种面向事件的动态知识图谱构建方法和装置
US20160299955A1 (en) Text mining system and tool
CN113076735B (zh) 目标信息的获取方法、装置和服务器
Mills et al. Graph-based methods for natural language processing and understanding—a survey and analysis
WO2019080863A1 (zh) 文本情感分类方法、存储介质及计算机
CN114547303B (zh) 基于Bert-LSTM的文本多特征分类方法及装置
CN110889292B (zh) 一种基于句义结构模型的文本数据生成观点摘要的方法及系统
CN101634983A (zh) 一种文本分类方法和装置
CN109472022B (zh) 基于机器学习的新词识别方法及终端设备
US20220365956A1 (en) Method and apparatus for generating patent summary information, and electronic device and medium
CN106294505B (zh) 一种反馈答案的方法和装置
CN103514213A (zh) 词语提取方法及装置
Fan et al. Prismatic: Inducing knowledge from a large scale lexicalized relation resource
Petrov et al. Sparse multi-scale grammars for discriminative latent variable parsing
CN112949293A (zh) 一种相似文本生成方法、相似文本生成装置及智能设备
Honkela et al. WordICA—emergence of linguistic representations for words by independent component analysis
CN112668286A (zh) 一种动态标签生成方法、装置、电子设备及存储介质
CN114020865A (zh) 一种搜索语句的处理方法、装置、设备及介质
Akhtar et al. Unsupervised morphological expansion of small datasets for improving word embeddings
Tianwen et al. Evaluate the chinese version of machine translation based on perplexity analysis

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17798595

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 25.02.2019)

122 Ep: pct application non-entry in european phase

Ref document number: 17798595

Country of ref document: EP

Kind code of ref document: A1