CN112035506A - Semantic recognition method and equipment - Google Patents
Semantic recognition method and equipment Download PDFInfo
- Publication number
- CN112035506A CN112035506A CN201911028330.0A CN201911028330A CN112035506A CN 112035506 A CN112035506 A CN 112035506A CN 201911028330 A CN201911028330 A CN 201911028330A CN 112035506 A CN112035506 A CN 112035506A
- Authority
- CN
- China
- Prior art keywords
- user
- query
- module
- database
- semantic recognition
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06F16/2433—Query languages
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06F16/245—Query processing
- G06F16/2452—Query translation
- G06F16/24522—Translation of natural language queries to structured queries
Abstract
The invention discloses a semantic recognition method and a device thereof, wherein the semantic recognition method comprises the following steps: receiving user input, wherein the user input is voice input or character input of Chinese natural language; correcting and optimizing user input, and performing syntactic analysis and entity analysis; triggering a Query behavior to determine a retrieval intention and a specific table for retrieving a retrieval subject; aggregating the context information and/or disambiguating fields by context; generating an SQL statement according to the determined Function and Column, and completely collecting necessary elements required by the SQL statement; converting and assembling the SQL sentence execution result into a natural language and/or a chart and outputting the natural language and/or the chart; the invention also discloses semantic recognition equipment, which comprises an input module database module, an automatic voice recognition correction module, an analysis and analysis module, an inquiry behavior module and the like; the semantic recognition method is high in recognition accuracy and speed of natural languages, supports custom expansion and Chinese and English, and can output charts.
Description
Technical Field
The invention relates to the field of artificial intelligent natural language processing, in particular to a semantic recognition method and equipment thereof.
Background
With the development of technology, the human-computer interaction experience has advanced greatly, and from the initial computer language input to the graphic interface, people expect to continuously try new human-computer interaction modes to obtain better interaction experience. The voice direct interaction is an example, and the voice direct interaction is very similar to direct human interaction, is convenient and safe, and can complete computer operation while driving and moving, so the voice direct interaction is popular. Voice interaction and other human-computer interaction can be achieved through interconversion between natural languages and computer languages, i.e., semantic recognition. The existing man-machine interaction technology mostly adopts the traditional natural language query, namely, the characters and the webpage content relevancy are matched and ordered according to the characters input by a user, the characters cannot be understood semantically essentially, and therefore, the fine database query and screening cannot be completed. The traditional database query requires high professional literacy of operators, even programming technicians need a certain period of learning and training to master grammar rules, and meanwhile, the operators need to be familiar with the database structure to know how to organize query logic.
European patent application EP3502928a1, Intelligent Language Query Processor, discloses a system for answering Natural Language Queries (NLQ) to a database system. The system includes a query receiver receiving an NLQ and modeling the NLQ as an entity tuple including at least a subject entity and an intent entity. The system also includes a service discovery component and a query translation unit. The resource calling mode of the invention adopts an ODate API to convert natural language processing into database query. However, because the invention adopts the OData data architecture, some technical limitations exist, such as only adopting a mode of replacing subject or pronoun according to the context analysis and recognizing the conversation intention of the user, and the semantic understanding accuracy of natural language is not high enough; the invention can not realize complex calculation, such as calculation methods of summation, averaging, most value and the like; in addition, the invention can not realize the recognition and query of the natural language of Chinese, and the invention has no function of making the query result into a chart and feeding the chart back to the user.
The invention supports Chinese and English natural language identification, the context analysis of the invention also sets the logics of incremental replacement, Function replacement and timely emptying the above, before carrying out semantic identification on NLQ, a customizable character preprocessing process is configured, and automatic voice identification correction optimization is carried out, thus ensuring the accuracy of the query intention; the invention can also realize various complex calculation methods such as summation, averaging, maximum, minimum, newest, difference, same ratio, ring ratio and the like; the invention can also realize the function of a chart, and the query result is made into the chart and fed back to the user together with the result; further, the present invention is applicable to a relational database, and can learn the possibility of enumerating the value of the type, the granularity of the value of the date type, and the like.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a semantic recognition method and equipment thereof.
The first aspect of the present invention provides a semantic recognition method, including the following steps:
receiving text input by a user;
judging whether to enter a natural language query processing flow;
if the natural language query processing flow is entered, correcting and optimizing the user text by combining the user database;
sentence analysis is carried out on the corrected and optimized text;
triggering inquiry, and acquiring a query subject and a query intention according to user configuration;
determining a table, namely determining a data table which needs to be used by a user query from a database;
if the query topic and the query intention cannot be determined in the query triggering step, judging the query topic and the query intention according to the information input by the user, the sentence analysis information and the disambiguation information;
collecting missing necessary query elements;
creating a structured query language statement;
executing a structured query language statement;
the execution results and/or graphs are returned.
Furthermore, the user input is natural language input, the natural language can be Chinese, and the input mode comprises voice input and character input; the input mode is voice input, correction optimization is automatic voice recognition correction optimization (ASR correction optimization), and automatic voice recognition correction refers to that voice input contents are automatically converted into corresponding texts through a computer technology and semantic correction is carried out on the texts.
In addition, the invention also supports the input of English natural language.
After receiving a text input by a user from the user, judging whether to enter a natural language query processing flow, specifically comprising: judging whether the text input by the user is in multi-round conversation, if so, entering a natural language query process, if not, calling a natural language query interface triggering module to judge whether the text input by the user contains entities related to database query, and if so, entering a natural language query processing process; and if not, invoking a reply in the chat corpus to feed back the user. Further, if the text input by the user is in a multi-turn conversation, judging whether the content input by the user is 'exit', if so, triggering a default reply (backsfill); if not, calling a natural language query interface triggering module to judge whether the text input by the user contains entities related to database query, if so, entering a natural language query flow, and if not, triggering a default reply (backhaul); the default reply comprises a reply feedback user in the chat corpus and quitting multiple rounds of conversations.
Furthermore, the semantic recognition method further includes that a user can select, supplement and correct the retrieved database by himself, the database is composed of a plurality of specific tables, the specific tables are not limited to information sorting displayed in a frame form, such as excel, and the tables are information sets in various forms.
Furthermore, after the natural language query processing flow is entered, before the modification optimization is performed, a step of performing text preprocessing according to a custom-configured text preprocessing script is further included, and a user can add a processing script for a special text therein.
Further, modifying and optimizing the user input, for example, the user input is a voice input, and further, performing entity parsing and/or grammar parsing on the modified and optimized user input text, where the entity parsing refers to analyzing and identifying entities with specific meanings in the text, and performing entity parsing on query topics according to a set of enumerated values in a database and data Column names of data tables, including names of persons, places, organization names, proper nouns, time, quantity, currency, proportion values and other terms or sentences of data table Column names (Column) classes (host guest classes), expression calculation mode (Function) classes (query intention classes) of the same proportion, ring proportion, maximum, minimum and comparison classes, and logical class terms such as "to", "include", "to", and "to" and "from"; the parsing includes parsing the grammar the text takes, such as positive, negative, turning, emphasizing, question-backing, and the like.
Further, the semantic recognition method further includes that the training database optimized for correction includes a chinese database.
Further, the semantic recognition method further includes that the training database for the grammar analysis and/or the entity analysis includes a chinese database.
Furthermore, the semantic recognition method also comprises the steps that a user can prepare a grammar analysis and/or entity analysis method by himself, and the self-preparation can be realized by adopting a preprocessing script and other modes; the self-configuration syntax analysis method is used for self-defining some special syntaxes, such as omission and inversion; the self-configuration entity analysis method is used for self-defining the contents of entities with specific meanings, such as uncommon place names, or carrying out special treatment on special contents, such as converting festivals into gregories.
Furthermore, the semantic recognition method further includes that a user can perform custom configuration on the Function in advance, where the custom configuration on the Function refers to giving a Function meaning to a certain character or giving a special meaning to a certain Function type character, word or sentence.
Further, before the inquiry is triggered, the method also comprises the step of preprocessing the text according to the query pre-intention preprocessing script configured by the user.
Further, determining a specific retrieved form by triggering a Query (Query) behavior, wherein the Query behavior is to acquire Function and Column existing in a text input by a user, judge whether the specific retrieved form can be determined according to the Function and Column, if so, execute the next step, if not, judge whether the specific retrieved form can be determined through information given by the user, if so, execute the next step, and if not, return to the legal-form-free operation; the no-statutory statement may be to inform the user that the form cannot be confirmed and ask for missing information.
Further, the semantic identification method also comprises the steps of determining whether a user has the access right of the table determining data table after the table is determined, if so, executing the next step, and if not, returning to the no-right operation.
Further, the semantic identification method further comprises the steps of judging whether the user has the access right of the specific data column of the fixed table data table or not if the user has the access right of the fixed table data table, screening out the data column which is not accessed, and executing the next step if the user does not have the access right of the specific data column of the fixed table data table.
Further, the method also comprises the following steps: if the query intention (Function) cannot be determined through the query behavior, the query intention comprises a data query mode or a data calculation mode, and the data query mode or the data calculation mode is determined according to the text input by the user and the analyzed entity.
Further, the method also comprises the following steps: query parameters are collected from the text entered by the user and the context. Generating a Structured Query Language (SQL) statement according to the current Function and the Column, and judging whether an SQL essential element is complete, wherein the SQL essential element is a programming language, the SQL essential element is complete, the SQL essential element means that the current information can form a complete SQL expression, namely the Function exists and Column and Value which are necessary for executing the Function exist, and the Value means the table item content of the data table; if not, returning to the question-pursuing operation; and if so, executing the SQL statement.
Further, the semantic identification method further comprises determining a Function and a Column in the user input text through a context disambiguation field and/or aggregating context information, wherein the context information aggregation is to supplement the user input text with the context of the user input.
Furthermore, the semantic identification method also comprises the step of synthesizing Function and Column information into SQL intermediate representation, wherein the SQL intermediate representation comprises complete SQL logic semantics and can be conveniently expanded into other SQL dialects expressing the same logic semantics, such as Oracle SQL.
And converting and assembling the result of the SQL sentence into a natural language and outputting the natural language, wherein the output can be voice output or non-voice output.
Further, the semantic recognition method further comprises the step of converting the result of executing the SQL statement into a chart and outputting the chart.
Preferably, the step of converting the result of executing the SQL statement into the chart may be performed in a default manner or in a user-defined manner.
The invention also provides a semantic recognition device, comprising: the system comprises an input module, a database module, a correction module, an analysis and analysis module, an inquiry behavior module (Query behavior module), a statement generation module, a judgment module, an execution module and an output module.
The input module is used for receiving user input, wherein the user input can be voice input, and the content of the user input can be natural language.
The database module is used for storing or connecting the retrieved database; the database can be stored locally in the database module or connected through the database module in a cloud or network, the database is composed of a plurality of specific tables, the specific tables are not limited to information sorting displayed in a frame form such as Excel, and the tables are understood to be information sets in various forms.
Further, the database module can select, supplement and modify the retrieved database through user operation.
The correction module is used for correcting and optimizing the user input, and if the user input is voice input, the correction and optimization adopts an automatic voice recognition correction technology; asr (automatic Speech recognition) correction refers to automatically converting Speech input contents into corresponding texts and performing semantic correction on the texts through computer technology.
The analysis and analysis module is used for carrying out syntactic analysis and entity analysis on the corrected user input text, wherein the entity analysis refers to the analysis and identification of entities with specific meanings in the text, including Column (host guest) characters such as name of a person, place name, organization name, proper noun, time, quantity, currency, ratio numerical value and the like; logical words such as "to," "about," "including," "except," and the like; and Function (action) type characters of the same ratio, ring ratio, maximum, minimum and comparison type words; the parsing includes parsing the grammar the text takes, such as positive, negative, turning, emphasizing, question-backing, and the like.
Further, the training database of the ASR modification module includes a chinese database.
Further, the training database of the analysis and analysis module comprises a Chinese database.
Furthermore, in the analysis and analysis module, a user can configure a syntax analysis and/or an entity analysis method by himself, wherein the self-configuration syntax analysis method is to define some special syntaxes by himself, such as omission and inversion; the self-configuration entity analysis method is used for self-defining the contents of entities with specific meanings, such as uncommon place names, or carrying out special treatment on special contents, such as converting festivals into gregories.
Furthermore, in the analysis and analysis module, a user may perform custom configuration on the Function in advance, where the custom configuration on the Function refers to giving a Function meaning to a certain character or giving a special meaning to a certain Function type character.
The Query behavior module is used for finishing Query behavior so as to determine a specific retrieved form, wherein the Query behavior is to acquire Function and Column existing in a text input by a user, judge whether the specific retrieved form can be determined according to the Function and Column, if so, execute the next step, if not, judge whether the specific retrieved form can be determined through information given by the user, if so, execute the next step, and if not, return to the legal form-free operation; the no-statutory statement may be to inform the user that the form cannot be confirmed and ask for missing information.
Furthermore, the semantic recognition device further comprises an authority limit module, wherein the authority limit module can judge whether the user has authority to perform a specific behavior, if so, the behavior is continued, and if not, the behavior is blocked; the behavior is prevented by refusing access, returning no-permission speech and the like, so that the user does not execute the behavior and is informed that the behavior is not authorized.
Further, the permission restriction module may determine whether a user has the right to access a particular database and/or the right to access a particular table and/or the right to access a particular column of data.
A statement generating module, configured to generate an SQL statement and/or a natural language, where the SQL (structured query language) is a programming language, and the generating refers to generating a corresponding statement including a corresponding requirement under a certain condition, for example, under a condition that Function, Column, and Value information are partially mastered; or converting the result of executing the SQL statement into a natural language.
The SQL query module is used for generating an SQL statement, and the SQL statement comprises a judging module and a judging module, wherein the judging module is used for judging whether SQL essential elements of the generated SQL statement are complete, and the SQL essential elements refer to that the current information can form a complete SQL expression, namely Function exists and Column and Value which are necessary for executing the Function; if not, returning to the question-pursuing operation; and if so, executing the SQL statement.
And the execution module is used for executing the SQL statement.
Further, the semantic recognition device further comprises a context aggregation module, configured to determine a Function and a Column in the user input text through a context disambiguation field and/or aggregate context information, where the context information aggregation is to supplement the user input text with the context of the user input.
Further, the statement generating module may further generate an SQL intermediate representation, where the generating the SQL intermediate representation refers to a step of synthesizing Function and Column information into the SQL intermediate representation, and the SQL intermediate representation includes complete SQL logical semantics and can be conveniently expanded into other SQL dialects expressing the same logical semantics, such as OracleSQL.
Furthermore, the semantic recognition device further comprises a chart generation module, which is used for converting the result of executing the SQL statement into a chart.
Preferably, the chart generation module can be configured by a user, so that the chart generation mode can be performed according to a default mode or a user-defined mode.
And the output module is used for outputting the natural language or the view.
The invention also provides a user-defined configuration method of the semantic recognition system, which comprises the following steps: the user inputs database link information or a data table document containing data and conforming to a natural language query format; the natural language query system accesses the database; judging whether the user database can be successfully accessed, if so, acquiring all data tables under the database and returning the data tables to the user for selection, and selecting the data tables needing to support NLQ query by the user; if the user database cannot be successfully accessed, returning to the previous step; extracting information such as data samples, types, comments and the like in the data table; and the user self-defines and configures the field attribute of the database or uses default configuration according to the extraction result.
Furthermore, the user-defined configuration method comprises the step of user-defined configuration of the inquiry behaviors by the user, and the user-defined configuration of the corresponding relation between the inquiry problems and the inquiry results.
Further, the user-defined configuration method comprises the step that the user performs user-defined configuration on the synonym thesaurus, and the user expands fields of enumerated values in the data table by adding synonyms.
The technical problems solved and the technical effects achieved by the invention are as follows:
it is an object of the present invention to solve the decoding problem from natural language to database query language.
Specifically, the method can understand the semantics of the natural language to a certain degree, then translate the semantics into a database query language (SQL) for database query, and simultaneously provide a related data statistical chart. The problems that the traditional natural language query cannot semantically understand characters per se and cannot finish fine database query and screening are solved.
The invention can directly organize SQL language and complete database query by understanding natural language, does not need to master data query programming language and depends on technical personnel for querying data programming, and can meet the data query requirement of ordinary people, thereby greatly reducing the use threshold.
The invention is mainly used for Chinese natural language processing, and has an interface for additionally configuring a specific database, thereby being capable of performing additional knowledge supplement on the existing model. The problem of the targeted optimization that the natural language of academic world is converted into the query language of the database and is only limited to the English language and additional specific databases cannot be added is solved.
Thus, the present invention has advantages including: the method has the advantages of high recognition precision and high speed of natural language, low technical threshold of query, support of Chinese natural language query, voice error correction, database configuration and privatization deployment, has the SQL dialect function, can output languages and charts, and can realize data chart reasoning.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate exemplary embodiments of the invention and, together with the description, serve to explain the invention and are not intended to limit the invention in a non-limiting sense. In the drawings:
FIG. 1 is a query flow diagram of a semantic recognition method provided by an embodiment of the present invention;
fig. 2 is a flowchart of a configuration semantic recognition system according to an embodiment of the present invention.
Detailed Description
In order to make the technical problems, technical solutions and technical effects to be solved by the present invention clearer, the technical solutions of the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It is to be understood that the described embodiments are merely exemplary of a portion of the invention, and not all. All other embodiments obtained by a person skilled in the art based on the embodiments of the present invention without any inventive step are also within the scope of the present invention.
The noun defines:
nlq (natural Language query) natural Language query, i.e. the processing flow method of the present invention for identifying, processing, querying and finally feeding back natural Language.
Sql (structured Query language) database Query language, also known as structured Query language, is a specific programming language for managing relational databases.
Asr (automatic Speech recognition), automatic Speech recognition, and computer technology to automatically convert human Speech content into corresponding words.
NER (named Entity recognition), named Entity recognition, and recognition of entities with specific meanings in text, mainly including names of people, names of places, names of institutions, proper nouns, and the like, and characters such as time, quantity, currency, proportional numerical values, and the like.
The Function query intention refers to the content or query mode that the user actually needs to query, or the calculation mode of the data that the user desires to obtain.
The Column names of the data tables in the Column database.
The name or content of the data table entry in the Value database.
Example 1:
as shown in fig. 1, the present embodiment provides a semantic identification method, which includes the following steps:
(1) flow judgment
The method comprises the steps that user input can be natural language input and comprises text input and voice input, the user input supports Chinese input and English input, the content of the user input is called user text in the invention, whether the user input is in multiple rounds of conversation is judged by combining context, if the user input is in the multiple rounds of conversation, an NLQ processing flow is entered, if the user input is not in the multiple rounds of conversation, an NLQ triggering module is called, whether the user input contains entities related to database query is judged, and if the user input contains the entities related to the database query, the NLQ processing flow is entered; if the user input does not contain the database query related entity, a default reply (backfill) is triggered, and the consequences of the default reply comprise that the user feeds back a reply of a sentence in the corpus, and the user exits from multiple rounds of conversations.
(2) Text pre-processing script
After entering NLQ, judging whether a user configures a text preprocessing script, if the user configures the text preprocessing script, executing the text preprocessing script, wherein the user can set the processing script of the text in a self-defined way, for example, the user can set that some special festival names are converted into calendar dates, and the like, and after the text preprocessing is executed, the correction optimization is carried out, if the input is voice input, the correction optimization uses ASR correction optimization; and if the user does not configure the text preprocessing script, directly performing ASR correction optimization.
(3) ASR correction optimization
After ASR correction optimization is carried out, the ASR identifies input of a user, the identified text is corrected through information of enumerated field values in a user database, for example, a place name named 'high key' exists in the user database, general ASR can be recognized as more common 'plaster', the word can be restored to 'high key' through ASR correction optimization processing, and the word is closer to the idea of searching the database by the user.
(4) Sentence parsing
Performing sentence analysis, wherein the sentence analysis comprises entity analysis and grammar analysis, and the entity analysis performs noun entity analysis according to a set of enumerated values in a database of a user and a Column name of a data table, and comprises date type, numerical type, Function type entities and the like; the syntax analysis includes reference analysis of Column and specific values and analysis of negative word classes, numerical range classes, and the like.
(5) Pre-Function custom scripting
Entering a Function workflow, judging whether a user configures a Function preprocessing script, if so, executing the Function preprocessing script, wherein the script processing target is an entity subjected to sentence analysis, and processing the analyzed entity according to the Function preprocessing script configured by the user. For example, some words are specially defined in some fields, for example, in a database in the weather field, which defines the rainfall in days, such as "2 month and 5 days rainfall" means the rainfall in the period of "2 month and 3 days 08:00:00-2 month and 4 days 08:00: 00", so that the user can specially process the analyzed time information in the system, and the range of the time information and the statistical data can be customized; after the Function preprocessing script is executed, entering triggering Query;
and if the user does not configure the Function preprocessing script, directly entering to trigger Query.
(6) Triggering Query
Entering triggering Query, which is to trigger a text after the subsequent processing according to a combination of a Function and a Column configured by a user, for example, the user configures a Query of "how the june cola sells" as a trigger "Query sales", and when the user asks for a similar sentence, for example, "how the june cola sells" in the last year, the Query such as Function ═ Query and Column ═ sales "is directly triggered, and meanwhile, if the user has a plurality of data tables, the Query further determines which table to use for querying.
(7) Fixed watch
Judging whether the Function and the Column can be obtained through the triggering operation matching of the Query, if so, obtaining the Function and the Column, and determining a used data table (fixed table) for next permission check; if the Function and the Column cannot be acquired through Query triggering, judging whether to determine which data table can be used according to information analyzed by user input, and if so, performing next permission check; if the data tables used cannot be determined, return no legal tables.
The data table refers to a table which is already manufactured and can be used for inquiring related data, the table is stored in a database, the patent only aims at a relational database, namely the form of resources can be described by a regular data table, and therefore information such as all possible fields of enumerated type values, granularity of date type values and the like can be learned. The user may additionally configure some data table related information, such as whether a field of a value type is of a unit price type or a rate type. Such as whether one enumerated value is a sub-level of another enumerated value. These can optimize the process of natural language translation to SQL.
(8) Authority checking
Because of the sensitivity and the confidentiality of the data, the invention is provided with the authority check of the user data table and the data column, after the table is determined, whether the user has the authority of the data table is checked, and if the user has the authority of the data table, whether the user has the authority of the data column is further checked; if no data table authority exists, returning to the no-authority word operation;
when checking whether the user has the data column authority, if so, entering a combined context disambiguation field; if the data column has no authority, the resolving entity of the data column without the authority is screened out, and then the disambiguating field combining the context is entered.
(9) Disambiguating fields in conjunction with context
For the NLQ system, the user's input is not only the text currently input by the user, but also includes the user's previous round of inquiry and answer, and in this step, the information of the user's inquiry can be supplemented and disambiguated in combination with the context, for example, "jilin" can refer to both "jilin city" and "jilin province", when the user asks "how to say for the jilin" after asking "sales performance of each province", the user is referred to "jilin province" rather than "jilin city" in a large probability. There are also incremental replacements, such as when the user gets an answer to "Shanghai sales Performance" above, then the user asks "Biguang? "automatically merge the context information, and query the result of" compare sales performance of Shanghai and Guangzhou ". As a Function replacement, for example, the user asks "20 stores in the upper sea before sales. "follow the question after the last," forward in Guangzhou? "at this point, the store was queried, in conjunction with the contextual information, 20 Guangzhou. Furthermore, when the next question is a relatively complete sentence, and there is no need to inherit the entity information from the previous question, the above information can be emptied, for example, is the first question user ask for the sales of iPad in 3 months of shanghai? The second question asked iPad in several colors? At this time, the second question does not need to inherit the entities such as Shanghai, March, sales volume and the like from the first question, and the above can be cleared. When no context exists, the disambiguation can be carried out by adopting a trial query mode, if a certain entity can represent the condition A and the condition B, and if the condition A is met and the condition B is certainly met, the condition B is adopted, namely the condition with looser limitation is adopted. For example, the user inputs "sales of Jilin for the last year", and if there is no context, "Jilin" may represent both "Jilin province" and "Jilin City", and "province" is "Jilin" which is necessarily satisfied when "City" is "Jilin", and thus Jilin province is adopted as a condition.
(10) Function determination and parameter collection
Judging whether the Function is determined through Query, if so, collecting Function parameters from the user text and the context; if the Function cannot be determined through Query, determining the Function according to the user text and the analyzed entity, and then collecting Function parameters from the user text and the context;
for example: the basic functions include: query one or more subjects, sum, maximum, difference, TopN (refer to the top N names, i.e., return the top N names after sorting the query results), parity ring ratio, average, etc. The Function parameters include: such as N in TopN, etc.
(11) Function deletion element Collection
Judging whether necessary Column and Value required by executing the Function are complete, if so, synthesizing the obtained information into SQL intermediate representation, if not, judging whether the necessary Column and Value can be inherited from the above, and if so, synthesizing the obtained information into SQL intermediate representation after inheriting the necessary Column and Value from the above; if the user can not inherit from the above, the user returns to the question-following operation to request the user to supplement necessary information.
(12) SQL intermediate representation
And synthesizing the obtained information into an SQL intermediate representation.
And a data structure expressed in the middle of SQL is arranged from the Function plus the entity information to the last SQL character string, and the data structure contains complete SQL logic semantics and is conveniently expanded into other SQL dialects expressing the same logic semantics, such as OracleSQL. The SQL intermediate representation may support complex SQL sentence assembly, such as "who are the first three actors in the movies of all ginger directors in the last decade? "such a complicated sentence can be queried only by analyzing the syntactic structure of the sentence itself. Meanwhile, the SQL intermediate representation can be packaged into high-level components such as computing "newest", "timeframe", "same-proportion", and the like. The components can be reused in various relational databases, and expression capacity and development efficiency are considered.
(13) Executing SQL to obtain results
And after the SQL character string is established, the NLQ system transmits the SQL character string to the database system to obtain a database operation result.
(14) Chart generation
The chart is a chart which is made according to the content of the user query according to the result finally obtained by the client query or calculation, for example, when the user only asks for 'sales in Shanghai district for march', the answer obtained by the user is an accurate answer that 'sales in Shanghai is xx'. Meanwhile, the chart system automatically expands the Shanghai region, so that the sales volume of all stores in Mount Shanghai can be obtained, for example, a user can add a parent node of a city which is 'store' into the data table configuration, the displayed table is expanded according to the stores (the stores are used as abscissa), the sales volume chart of each store in Shanghai in Mount Shanghai is displayed, if the data table knowledge that the 'stores' are children nodes of the 'Shanghai' is not available, the chart is automatically expanded in time (the time is used as abscissa), and a sales volume curve of each day in Shanghai in Mount Shanghai is inquired and displayed.
(15) Returning results
And finally merging the answers to the questions in the input of the user and the relevant charts together and returning the merged answers to the user.
Example 2: the invention also provides a semantic recognition device, comprising: the system comprises an input module, a database module, a correction module, an analysis and analysis module, an inquiry behavior module (Query behavior module), a statement generation module, a judgment module, an execution module and an output module.
The input module is used for receiving user input, wherein the user input can be voice input, and the content of the user input can be natural language.
The database module is used for storing or connecting the retrieved database; the database can be stored locally in the database module or connected through the database module in a cloud or network, the database is composed of a plurality of specific tables, the specific tables are not limited to information sorting displayed in a frame form such as Excel, and the tables are understood to be information sets in various forms.
Further, the database module can select, supplement and modify the retrieved database through user operation.
The correction module is used for correcting and optimizing the user input, and if the user input is voice input, the correction and optimization adopts an automatic voice recognition correction technology; asr (automatic Speech recognition) correction refers to automatically converting Speech input contents into corresponding texts and performing semantic correction on the texts through computer technology.
The analysis and analysis module is used for carrying out syntactic analysis and entity analysis on the corrected user input text, wherein the entity analysis refers to the analysis and identification of entities with specific meanings in the text, including Column (host guest) characters such as name of a person, place name, organization name, proper noun, time, quantity, currency, ratio numerical value and the like; logical words such as "to," "about," "including," "except," and the like; and Function (action) type characters of the same ratio, ring ratio, maximum, minimum and comparison type words; the parsing includes parsing the grammar the text takes, such as positive, negative, turning, emphasizing, question-backing, and the like.
Further, the training database of the ASR modification module includes a chinese database.
Further, the training database of the analysis and analysis module comprises a Chinese database.
Furthermore, in the analysis and analysis module, a user can configure a syntax analysis and/or an entity analysis method by himself, wherein the self-configuration syntax analysis method is to define some special syntaxes by himself, such as omission and inversion; the self-configuration entity analysis method is used for self-defining the contents of entities with specific meanings, such as uncommon place names, or carrying out special treatment on special contents, such as converting festivals into gregories.
Furthermore, in the analysis and analysis module, a user may perform custom configuration on the Function in advance, where the custom configuration on the Function refers to giving a Function meaning to a certain character or giving a special meaning to a certain Function type character.
The Query behavior module is used for finishing Query behavior so as to determine a specific retrieved form, wherein the Query behavior is to acquire Function and Column existing in a text input by a user, judge whether the specific retrieved form can be determined according to the Function and Column, if so, execute the next step, if not, judge whether the specific retrieved form can be determined through information given by the user, if so, execute the next step, and if not, return to the legal form-free operation; the no-statutory statement may be to inform the user that the form cannot be confirmed and ask for missing information.
Furthermore, the semantic recognition device further comprises an authority limit module, wherein the authority limit module can judge whether the user has authority to perform a specific behavior, if so, the behavior is continued, and if not, the behavior is blocked; the behavior is prevented by refusing access, returning no-permission speech and the like, so that the user does not execute the behavior and is informed that the behavior is not authorized.
Further, the permission restriction module may determine whether a user has the right to access a particular database and/or the right to access a particular table and/or the right to access a particular column of data.
A statement generating module, configured to generate an SQL statement and/or a natural language, where the SQL (structured query language) is a programming language, and the generating refers to generating a corresponding statement including a corresponding requirement under a certain condition, for example, under a condition that Function, Column, and Value information are partially mastered; or converting the result of executing the SQL statement into a natural language.
The SQL query module is used for generating an SQL statement, and the SQL statement comprises a judging module and a judging module, wherein the judging module is used for judging whether SQL essential elements of the generated SQL statement are complete, and the SQL essential elements refer to that the current information can form a complete SQL expression, namely Function exists and Column and Value which are necessary for executing the Function; if not, returning to the question-pursuing operation; and if so, executing the SQL statement.
And the execution module is used for executing the SQL statement.
Further, the semantic recognition device further comprises a context aggregation module, configured to determine a Function and a Column in the user input text through a context disambiguation field and/or aggregate context information, where the context information aggregation is to supplement the user input text with the context of the user input.
Further, the statement generating module may further generate an SQL intermediate representation, where the generating the SQL intermediate representation refers to a step of synthesizing Function and Column information into the SQL intermediate representation, and the SQL intermediate representation includes complete SQL logical semantics and can be conveniently expanded into other SQL dialects expressing the same logical semantics, such as OracleSQL.
Furthermore, the semantic recognition device further comprises a chart generation module, which is used for converting the result of executing the SQL statement into a chart.
Preferably, the chart generation module can be configured by a user, so that the chart generation mode can be performed according to a default mode or a user-defined mode.
And the output module is used for outputting the natural language or the view, outputting the query result and outputting the user query result in a chart mode according to a default or user-defined mode.
Example 3:
as shown in fig. 2, this embodiment discloses a user-defined configuration method of a semantic recognition system of the present invention, which includes:
(1) database connection
The user inputs database link information including a database access URI, a user name password and the like, then the NLQ system accesses the database and judges whether the database of the user can be successfully accessed, if so, all data tables under the database are acquired and returned to the user for selection, and the user only needs to select the data table which the NLQ wants to support query; and if the user database cannot be successfully accessed, returning to the previous step. Or the user directly uploads the excel file containing the data which conforms to the NLQ system format, and after the user successfully accesses the database, the user selects a specific table which needs to support NLQ query.
(2) Multi-table configuration
Extracting information such as data samples, types, comments and the like in a specific form of a user;
for each query form, the system extracts the data type, comments and other information of each field in the data table and returns the information to the user for further configuration, wherein the user can directly use the default configuration without any modification, and the user can modify or increase the configuration at any time later.
(3) Data table configuration
The user determines whether to configure field attributes or to use default configuration according to the extraction result, for example, personalized configuration items such as parent and child nodes, default values, units, numerical formats, etc. can be set in some fields, allowing the user to freely configure, specifically, if "city" is set as a child node of "province", then is the highest GDP in query "where a province a is? And when the node is in the 'time', the child nodes belonging to the city of province A are automatically expanded to be inquired.
(4) Query configuration
The user chooses to use default or custom configuration for Query behavior, such as by asking directly "how did product a sell last month? The effect of ' achieving the Query ' sales volume ' can be achieved by adding the sentence under the Query of ' sales volume ' to increase the availability of the NLQ, and the NLQ can also automatically learn a similar expression mode; in addition, the user may set a special condition for a special Query, such as "how is product a sold in the last month? "this Query must define a" city "condition, and when the Query lacks that condition, the user will be asked back to determine that condition.
(5) Synonym configuration
The user configures the synonym thesaurus, expands the field of enumerated values in the data table, and if the sales volume and the sales volume are configured as synonyms, the same result as the query sales volume is generated when the user inputs the query sales volume.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the examples, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and variations can be made by persons skilled in the art without departing from the principles of the invention and should be considered as within the scope of the invention.
Claims (31)
1. A semantic recognition method is characterized by comprising the following steps:
receiving text input by a user;
judging whether to enter a natural language query processing flow;
if the natural language query processing flow is entered, correcting and optimizing the text input by the user by combining the user database;
sentence analysis is carried out on the corrected and optimized text;
triggering inquiry, and acquiring a query subject and a query intention according to user configuration;
determining a table, namely determining a data table which needs to be used by a user query from a database;
if the query topic and the query intention cannot be determined in the query triggering step, judging the query topic and the query intention according to the text input by the user, the information of sentence analysis and the information of disambiguation;
collecting missing necessary query elements;
creating a structured query language statement;
executing a structured query language statement;
the execution results and/or graphs are returned.
2. The semantic recognition method according to claim 1, wherein the user input is natural language input, the natural language is chinese or english, and the input means includes voice input and text input; and when the input mode is voice input, the correction optimization adopts voice recognition correction optimization.
3. The semantic recognition method according to claim 1, wherein determining whether to enter a natural language query processing flow specifically comprises: judging whether the text input by the user is in multi-round conversation, if so, entering a natural language query process, if not, calling a natural language query interface triggering module to judge whether the text input by the user contains entities related to database query, and if so, entering a natural language query processing process; and if not, invoking a reply in the chat corpus to feed back the user.
4. The semantic recognition method according to claim 3, wherein if the text input by the user is in a plurality of rounds of dialog, whether the content input by the user is 'exit' is judged, and if yes, a default reply is triggered; if not, calling a natural language query interface triggering module to judge whether the text input by the user contains entities related to database query, if so, entering a natural language query flow, and if not, triggering a default reply; the default reply comprises a reply feedback user in the chat corpus and quitting multiple rounds of conversations.
5. The semantic recognition method of claim 1, further comprising the steps of:
the user can select, supplement and correct the used database.
6. The semantic identification method according to claim 1, further comprising the step of performing text preprocessing according to a custom-configured text preprocessing script after entering the natural language query processing flow and before performing modification optimization.
7. The semantic recognition method according to claim 1, wherein the sentence parsing comprises entity parsing and/or grammar parsing, wherein the entity parsing comprises noun entity parsing, and the specific steps comprise entity parsing of the query topic according to the set of enumerated values in the database and the data column names of the data table, and wherein the entity parsing further comprises query intent parsing.
8. The semantic recognition method of claim 7, further comprising the step of self-configuring entity parsing and/or parsing methods.
9. The semantic recognition method according to claim 1, further comprising the step of text preprocessing according to a user-configured pre-query-intent preprocessing script prior to triggering the query.
10. The semantic identification method according to claim 1, wherein after the query is triggered, if there are a plurality of query data tables, the user is asked to make a table, or the table is made according to the parsed information of the sentence.
11. The semantic recognition method of claim 1, further comprising, after the tabulating, the steps of: and judging whether the user has the access authority of the fixed table data table, if so, executing the next step, and if not, returning to the no-authority operation.
12. The semantic identification method according to claim 11, further comprising the steps of, if the user has access rights to the spreadsheet: and judging whether the user has the access authority of the data column of the fixed table data table, if so, executing the next step, and if not, screening out the data column without the authority.
13. The semantic recognition method of claim 1, further comprising the steps of: if the data query mode or the data calculation mode cannot be determined through the query behavior, determining query intentions according to the text input by the user and the analyzed entity, wherein the query intentions comprise the data query mode and/or the data calculation mode.
14. The semantic recognition method of claim 1, further comprising the steps of: query parameters are collected from the text entered by the user and the context.
15. The semantic recognition method of claim 1, further comprising the steps of: information aggregation and/or disambiguation fields are performed in conjunction with the context of the user input.
16. The semantic recognition method according to claim 1, wherein the chart is generated according to the execution result, and the step of converting the result of executing the structured query language sentence into the chart can be performed according to a default mode or a user-defined mode.
17. A semantic recognition device comprising:
an input module for receiving user input;
the database module is used for storing or connecting the retrieved database;
the correction module is used for correcting and optimizing the user input;
the analysis and analysis module is used for carrying out syntactic analysis and entity analysis on the corrected user input text;
the inquiry behavior module is used for completing inquiry behaviors;
the sentence generation module is used for generating a structured query language sentence and/or a natural language;
the judging module is used for judging whether the structural query language requirements of the generated structural query language sentences are complete or not;
an execution module for executing the structured query language statement;
and the output module is used for outputting the natural language or the chart.
18. Semantic recognition device according to claim 17, characterized in that the database module can select, supplement, modify the retrieved database by means of user actions.
19. The semantic recognition device of claim 17, wherein the training database of the rework module comprises a chinese database.
20. The semantic recognition apparatus of claim 17, wherein the training database of the parsing module comprises a chinese database.
21. The semantic recognition device according to claim 17, wherein in the parsing module, a user can configure a parsing method and/or an entity parsing method.
22. The semantic recognition device of claim 17, wherein the analysis and parsing module allows a user to customize the query intent in advance.
23. The semantic recognition device of claim 17, further comprising a permission restriction module.
24. The semantic recognition device of claim 23, wherein the permission restriction module can determine whether a user has permission to access a particular database and/or permission to access a particular table and/or permission to access a particular column of data.
25. The semantic recognition device of claim 17, further comprising a context aggregation module to aggregate context information and/or by context disambiguating fields.
26. The semantic recognition device of claim 17, wherein the statement generation module is further operable to generate a structured query language intermediate representation.
27. The semantic recognition device of claim 17, further comprising a graph generation module to convert results of executing the structured query language statement into a graph.
28. The semantic recognition device of claim 27, wherein the chart generation module is self-configurable by a user.
29. A user-defined configuration method of a semantic recognition system is characterized by comprising the following steps: the user inputs database link information or a data table document containing data and conforming to a natural language query format; the natural language query system accesses the database; judging whether the user database can be successfully accessed, if so, acquiring all data tables under the database and returning the data tables to the user for selection, and selecting the data tables needing to support natural language query by the user; if the user database cannot be successfully accessed, returning to the previous step; extracting information such as data samples, types, comments and the like in the data table; and the user self-defines and configures the field attribute of the database or uses default configuration according to the extraction result.
30. The user-defined configuration method of claim 29, comprising a step of user-defined configuration of query behavior, wherein the user-defined configuration of the corresponding relationship between query question and query intention.
31. The user-defined configuration method of claim 29, comprising the step of user-defined configuration of the thesaurus by the user, wherein the user expands the field of the enumerated value in the data table by adding the synonym.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911028330.0A CN112035506A (en) | 2019-10-28 | 2019-10-28 | Semantic recognition method and equipment |
| PCT/CN2020/083555 WO2021082353A1 (en) | 2019-10-28 | 2020-04-07 | Semantic recognition method and device therefor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911028330.0A CN112035506A (en) | 2019-10-28 | 2019-10-28 | Semantic recognition method and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112035506A true CN112035506A (en) | 2020-12-04 |
Family
ID=73576261
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911028330.0A Pending CN112035506A (en) | 2019-10-28 | 2019-10-28 | Semantic recognition method and equipment |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN112035506A (en) |
| WO (1) | WO2021082353A1 (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112632110A (en) * | 2021-01-06 | 2021-04-09 | 上海合合信息科技股份有限公司 | Data extraction method and device for automatically generating query statement |
| CN112989011A (en) * | 2021-04-26 | 2021-06-18 | 广州思迈特软件有限公司 | Data query method, data query device and electronic equipment |
| CN113032418A (en) * | 2021-02-08 | 2021-06-25 | 浙江大学 | Method for converting complex natural language query into SQL (structured query language) based on tree model |
| CN113609864A (en) * | 2021-08-06 | 2021-11-05 | 珠海市鸿瑞信息技术股份有限公司 | Text semantic recognition processing system and method based on industrial control system |
| CN113918701A (en) * | 2021-10-20 | 2022-01-11 | 北京亿信华辰软件有限责任公司 | Billboard display method and device |
| CN115346690A (en) * | 2022-07-08 | 2022-11-15 | 中国疾病预防控制中心慢性非传染性疾病预防控制中心 | System for guiding operator to ask help seeker |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140236572A1 (en) * | 2013-02-20 | 2014-08-21 | Jinni Media Ltd. | System Apparatus Circuit Method and Associated Computer Executable Code for Natural Language Understanding and Semantic Content Discovery |
| CN108829858A (en) * | 2018-06-22 | 2018-11-16 | 北京京东金融科技控股有限公司 | Data query method, apparatus and computer readable storage medium |
| CN109063166A (en) * | 2018-08-15 | 2018-12-21 | 中国联合网络通信集团有限公司 | A kind of voice inquiry method and device, user equipment |
| CN109408526A (en) * | 2018-10-12 | 2019-03-01 | 平安科技（深圳）有限公司 | SQL statement generation method, device, computer equipment and storage medium |
| CN109697201A (en) * | 2018-12-27 | 2019-04-30 | 清华大学 | A kind of method of query processing, system, equipment and computer readable storage medium |
| CN109947794A (en) * | 2019-02-21 | 2019-06-28 | 东华大学 | A kind of interactive natural language inquiry conversion method |
| CN110188163A (en) * | 2019-04-13 | 2019-08-30 | 上海策友信息科技有限公司 | Data intelligence processing system based on natural language |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107016012A (en) * | 2015-09-11 | 2017-08-04 | 谷歌公司 | Handle the failure in processing natural language querying |
| US9959311B2 (en) * | 2015-09-18 | 2018-05-01 | International Business Machines Corporation | Natural language interface to databases |
| CN105868249A (en) * | 2015-12-15 | 2016-08-17 | 乐视网信息技术（北京）股份有限公司 | Data query control method and device |
| US10430447B2 (en) * | 2018-01-31 | 2019-10-01 | International Business Machines Corporation | Predicting intent of a user from anomalous profile data |
-
2019
- 2019-10-28 CN CN201911028330.0A patent/CN112035506A/en active Pending
-
2020
- 2020-04-07 WO PCT/CN2020/083555 patent/WO2021082353A1/en active Application Filing
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140236572A1 (en) * | 2013-02-20 | 2014-08-21 | Jinni Media Ltd. | System Apparatus Circuit Method and Associated Computer Executable Code for Natural Language Understanding and Semantic Content Discovery |
| CN108829858A (en) * | 2018-06-22 | 2018-11-16 | 北京京东金融科技控股有限公司 | Data query method, apparatus and computer readable storage medium |
| CN109063166A (en) * | 2018-08-15 | 2018-12-21 | 中国联合网络通信集团有限公司 | A kind of voice inquiry method and device, user equipment |
| CN109408526A (en) * | 2018-10-12 | 2019-03-01 | 平安科技（深圳）有限公司 | SQL statement generation method, device, computer equipment and storage medium |
| CN109697201A (en) * | 2018-12-27 | 2019-04-30 | 清华大学 | A kind of method of query processing, system, equipment and computer readable storage medium |
| CN109947794A (en) * | 2019-02-21 | 2019-06-28 | 东华大学 | A kind of interactive natural language inquiry conversion method |
| CN110188163A (en) * | 2019-04-13 | 2019-08-30 | 上海策友信息科技有限公司 | Data intelligence processing system based on natural language |
Non-Patent Citations (1)
| Title |
|---|
| 北京东方人华科技有限公司: "《普通高等院校应用型人才培养"十三五"规划教材 Python数据分析》", 中国铁道出版社, pages: 175 - 177 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112632110A (en) * | 2021-01-06 | 2021-04-09 | 上海合合信息科技股份有限公司 | Data extraction method and device for automatically generating query statement |
| CN113032418A (en) * | 2021-02-08 | 2021-06-25 | 浙江大学 | Method for converting complex natural language query into SQL (structured query language) based on tree model |
| CN112989011A (en) * | 2021-04-26 | 2021-06-18 | 广州思迈特软件有限公司 | Data query method, data query device and electronic equipment |
| CN113609864A (en) * | 2021-08-06 | 2021-11-05 | 珠海市鸿瑞信息技术股份有限公司 | Text semantic recognition processing system and method based on industrial control system |
| CN113918701A (en) * | 2021-10-20 | 2022-01-11 | 北京亿信华辰软件有限责任公司 | Billboard display method and device |
| CN113918701B (en) * | 2021-10-20 | 2022-04-15 | 北京亿信华辰软件有限责任公司 | Billboard display method and device |
| CN115346690A (en) * | 2022-07-08 | 2022-11-15 | 中国疾病预防控制中心慢性非传染性疾病预防控制中心 | System for guiding operator to ask help seeker |
| CN115346690B (en) * | 2022-07-08 | 2023-12-01 | 中国疾病预防控制中心慢性非传染性疾病预防控制中心 | System for guiding operator to ask help seeker |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021082353A1 (en) | 2021-05-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10579657B2 (en) | Answering questions via a persona-based natural language processing (NLP) system | |
| CN112035506A (en) | Semantic recognition method and equipment | |
| US10262062B2 (en) | Natural language system question classifier, semantic representations, and logical form templates | |
| RU2637992C1 (en) | Method of extracting facts from texts on natural language | |
| US9448995B2 (en) | Method and device for performing natural language searches | |
| US11080295B2 (en) | Collecting, organizing, and searching knowledge about a dataset | |
| RU2564629C1 (en) | Method of clustering of search results depending on semantics | |
| KR101732342B1 (en) | Trusted query system and method | |
| RU2509350C2 (en) | Method for semantic processing of natural language using graphic intermediary language | |
| US10467122B1 (en) | Methods, systems, and computer program product for capturing and classification of real-time data and performing post-classification tasks | |
| RU2626555C2 (en) | Extraction of entities from texts in natural language | |
| US20150324422A1 (en) | Natural Language Query | |
| EP0387226A1 (en) | Natural language analyzing apparatus and method | |
| EP1814047A1 (en) | Linguistic user interface | |
| CN114547329A (en) | Method for establishing pre-training language model, semantic analysis method and device | |
| JP2003228564A (en) | Context-sensitive flag setting for information in natural language text, and service for centralized control of meta-data related to the information via computer network | |
| RU2697647C1 (en) | System and method for automatic creation of templates | |
| US9075864B2 (en) | Method and system for semantic searching using syntactic and semantic analysis | |
| CN110175585A (en) | It is a kind of letter answer correct system and method automatically | |
| TWI735380B (en) | Natural language processing method and computing apparatus thereof | |
| RU61442U1 (en) | SYSTEM OF AUTOMATED ORDERING OF UNSTRUCTURED INFORMATION FLOW OF INPUT DATA | |
| CN112507089A (en) | Intelligent question-answering engine based on knowledge graph and implementation method thereof | |
| CN112579733A (en) | Rule matching method, rule matching device, storage medium and electronic equipment | |
| Bais et al. | A model of a generic natural language interface for querying database | |
| JPWO2019167281A1 (en) | Response processing program, response processing method, response processing device and response processing system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |