WO2020057023A1

WO2020057023A1 - Natural-language semantic parsing method, apparatus, computer device, and storage medium

Info

Publication number: WO2020057023A1
Application number: PCT/CN2019/071251
Authority: WO
Inventors: 江琳; 杨镭
Original assignee: 深圳壹账通智能科技有限公司
Priority date: 2018-09-18
Filing date: 2019-01-11
Publication date: 2020-03-26
Also published as: CN109359295A

Abstract

Provided is a data resource-based method for semantic parsing of natural language, comprising: receiving a request for semantic parsing of natural language information sent by a terminal, and obtaining a semantic context corresponding to the natural language information. In the corresponding semantic context, using a preset semantic parsing method to parse the natural language information to obtain an initial semantic parsing result, and, according to a correlation between a preset keyword and a filter value, obtaining a filter value corresponding to the natural language information. The filter value is used to filter the initial semantic parsing result to obtain a semantic parsing result matching the filter value, and the semantic parsing result is sent to the terminal.

Description

Method, device, computer equipment and storage medium for semantic analysis of natural language

Cross-reference to related applications

This application claims the priority of a Chinese patent application filed on September 18, 2018 with the Chinese Patent Office under the application number 2018110895444. The application name is "Semantic Parsing Method, Apparatus, Computer Equipment, and Storage Medium for Natural Language." Incorporated by reference in this application.

Technical field

The present application relates to a method, an apparatus, a computer device, and a storage medium for semantic analysis of natural language.

Background technique

With the rapid development of computer science and technology, there have been ways to achieve effective communication between people and computers by using natural language processing. To achieve natural language communication between humans and computers, it is necessary for computers to understand the meaning of natural language texts and to express given intentions and ideas in natural language. However, because natural language texts have various ambiguities and ambiguities in different scenarios or different contexts, before storing natural language in a computer system, it needs to be organized and analyzed, according to the corresponding scene And context to disambiguate it and transform it into a format that meets the computer's internal storage requirements.

Traditional natural language semantic analysis, by receiving the natural language information input by the user, and determining the type of the semantic scene according to the received natural language information, under the determined type of the semantic scene, the natural language is performed using a preset semantic analysis method Semantic parsing to obtain parsing results.

However, the inventors realized that due to the ambiguity of natural language itself, when there is a mistake in the determination of the type of the semantic scene, it is easy to get a completely wrong semantic parsing result, reducing the bias of user understanding, requiring repeated operations, resulting in a waste of a large amount of resources .

Summary of the Invention

According to various embodiments disclosed in the present application, a method, an apparatus, a computer device, and a storage medium for a semantic analysis of a natural language are provided.

A natural language semantic analysis method includes:

Receiving a semantic parsing request for natural language information sent by a terminal;

Acquiring a semantic scene corresponding to the natural language information;

Parsing the natural language information by using a preset semantic parsing method under the corresponding semantic scenario to obtain an initial semantic parsing result;

Obtaining a filtering value corresponding to the natural language information according to a preset correspondence between keywords and filtering values; and

The initial semantic parsing result is filtered using the screening value to obtain a semantic parsing result that matches the filtered value, and the semantic parsing result is sent to a terminal.

A semantic analysis device for natural language includes:

Acquiring a semantic scene corresponding to the natural language information;

A computer device includes a memory and one or more processors. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the one or more processors are executed. The following steps:

Acquiring a semantic scene corresponding to the natural language information;

One or more non-volatile computer-readable storage media storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the following steps:

Acquiring a semantic scene corresponding to the natural language information;

Details of one or more embodiments of the present application are set forth in the accompanying drawings and description below. Other features and advantages of the application will become apparent from the description, the drawings, and the claims.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to explain the technical solutions in the embodiments of the present application more clearly, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. Those of ordinary skill in the art can obtain other drawings according to the drawings without paying creative labor.

FIG. 1 is an application scenario diagram of a semantic parsing method of natural language according to one or more embodiments.

FIG. 2 is a schematic flowchart of a semantic parsing method of natural language according to one or more embodiments.

FIG. 3 is a schematic flowchart of a semantic scenario for acquiring natural language information according to one or more embodiments.

FIG. 4 is a block diagram of a semantic parsing apparatus for natural language according to one or more embodiments.

FIG. 5 is a block diagram of a computer device according to one or more embodiments.

detailed description

In order to make the technical solution and advantages of the present application more clear and clear, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application.

The semantic parsing method provided by the present application can be applied to the application environment shown in FIG. 1. The terminal 102 and the server 104 communicate through a network. The server 104 receives a semantic analysis request for natural language information sent by the terminal 102, and acquires a semantic scene corresponding to the natural language information. In the corresponding semantic scenario, the server 104 parses natural language information by using a preset semantic parsing method to obtain an initial semantic parsing result. According to the preset correspondence between keywords and filtered values, obtain filtered values corresponding to natural language information, and use the filtered values to filter the initial semantic parsing results to obtain semantic parsing results that match the filtered values. Send to terminal 102. The terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, and tablet computers. The server 104 may be implemented by an independent server or a server cluster composed of multiple servers.

In one embodiment, as shown in FIG. 2, a method for semantic parsing of natural language is provided. The method is applied to the server in FIG. 1 as an example, and includes the following steps:

S202. The server receives a semantic parsing request for natural language information sent by the terminal.

Relative to the artificial languages of computer applications, such as programming languages, machine languages, controlled retrieval languages, etc., natural languages represent spoken and written languages that people use every day. The semantic parsing request means a request for semantic parsing of the natural language information input by the user. The server obtains the natural language information input by the user at the terminal, and performs semantic parsing of the corresponding natural language information according to the semantic parsing request sent by the terminal.

Specifically, the server receives the natural language information input by the user at the terminal, and receives a semantic analysis request for the natural language information sent by the terminal, and performs semantic analysis on the received natural language information according to the semantic analysis request. However, natural language information has various ambiguities and ambiguities in different scenarios or different contexts. Before semantic analysis of natural language information, it needs to be sorted and analyzed. According to the corresponding scene and Context eliminates its ambiguity, and transforms it into a format that meets the internal storage requirements of the computer, and performs semantic analysis on the natural language information that meets the preset requirements in the computer.

Further, the natural language information input by the user is "the loan distribution of Shenzhen university students" as an example. The user enters "Shenzhen University Student Loan Distribution" in the terminal, and the terminal sends to the server a semantic analysis request for semantic analysis of the "Shenzhen University Student Loan Distribution". The server receives the semantic analysis request, and according to the semantic analysis request, the "Shenzhen University Student Loan Distribution" "For semantic parsing. According to different scenarios or contexts, the semantics of the natural language information “Shenzhen University Student Loan Distribution” can include: “Shenzhen University Student Loan Education Institution Distribution”, “Shenzhen University Student Loan Situation”, “Shenzhen University Student Loan” Amount distribution "and" Shenzhen university student loan types distribution "and other situations. For a variety of possible situations, the natural language information is analyzed and arranged in a certain context and application scenario, and the semantics of the corresponding natural language information representation in different contexts or application scenarios are obtained.

S204. The server obtains a semantic scene corresponding to the natural language information.

Specifically, the server separately calculates the relevance value between the natural language information and different semantic scenes, sorts the semantic scenes according to the magnitude of the relevance value, and obtains the semantic scene corresponding to the maximum relevance value.

The relevance value is used to judge the relevance between the natural language information input by the user and multiple semantic scenes. The server obtains the degree of association between the natural language information and the semantic scene by calculating the correlation value between the natural language information input by the user and the different semantic scenes. The calculated relevance values are sorted in ascending order, and the semantic scene corresponding to the largest relevance value is obtained as the semantic scene most relevant to the natural language information input by the user, that is, the semantic scene with the highest degree of relevance. .

Further, taking the natural language information entered by the user as "Shenzhen small borrower regional distribution", optional semantic scenarios include: small loans, loans, and large loans, etc., respectively, to calculate "Shenzhen small borrower regional distribution" , The degree of relevance to small loans, loans, and large loans, and the semantic scenes are sorted according to the value of the relevance degree, and the semantic scene corresponding to the obtained maximum relevance value is a small loan.

S206: The server parses the natural language information by using a preset semantic parsing method in a corresponding semantic scene to obtain an initial semantic parsing result.

Different semantic scenarios correspond to different semantic parsing methods. The server can obtain the correspondence between the preset semantic scenes and semantic parsing methods, and obtain according to the obtained correspondence between the preset semantic scenes and semantic parsing methods. The semantic parsing method corresponding to the semantic scene. The server uses the semantic parsing method corresponding to the semantic scene to semantically parse the natural language information input by the user under the corresponding semantic scene.

Specifically, the server obtains the semantic parsing mode corresponding to the semantic scene corresponding to the maximum relevance value according to the correspondence between the semantic scene corresponding to the maximum relevance value and the semantic parsing mode. Parse natural language information according to the semantic parsing method to obtain the original semantic parsing result. Initial inspection is performed on the original semantic parsing result by using a preset inspection rule, and the original semantic parsing result that conforms to the preset inspection rule is obtained, and the initial semantic parsing result is obtained according to the original semantic parsing result that conforms to the preset inspection rule.

The preset inspection rules are used for initial inspection of the original semantic parsing results, including checking the integrity and validity of the original semantic parsing results, and determining whether the original semantic parsing results are complete and valid. If the original semantic parsing result includes the semantic parsing of all keywords of the natural language information input by the user, it means that the original semantic parsing result passes the integrity check performed by the inspection rule. If the original semantic parsing result parses the keywords of natural language information semantically, it can effectively express the original input of natural language information, indicating that the original semantic parsing result has passed the validity check by the inspection rules.

Further, the case where the semantic scene corresponding to the user's natural language information is a borrowing scene is taken as an example. The server obtains the natural language information “Shenzhen university student micro-loan distribution” input by the user, and according to the keyword category to which the keywords in the natural language information entered by the user, obtains the corresponding semantic scene as “micro-loan”, according to the preset The corresponding relationship between the semantic scene and the semantic parsing method, to obtain the semantic parsing method corresponding to the semantic scene as "small loan", and perform a semantic parsing on the natural language information "Shenzhen university student small loan distribution" to obtain the initial semantic analysis The results can include the following situations: "Shenzhen university students' small loan amount distribution", "Shenzhen university students' small loan area distribution" and "Shenzhen university students' small loan".

S208. The server obtains a filtering value corresponding to the natural language information according to a preset correspondence between the keywords and the filtering value.

Specifically, the server obtains a filtering value corresponding to the natural language information by acquiring a filtering mechanism corresponding to the keywords and according to a preset relationship between the preset filtering mechanism and the filtering value.

There is a correspondence between keywords and filtering mechanisms, and different keywords correspond to different filtering mechanisms. There is a preset correspondence between the screening mechanism and the screening value. According to the mapping relationship between the screening mechanism and the screening value, a screening value corresponding to the screening mechanism can be obtained. At the same time, due to the correspondence between different natural language information and keywords, the filtering value corresponding to the keywords can be obtained according to the correspondence between the keywords and the filtering mechanism and the correspondence between the filtering mechanism and the filtering value. Furthermore, the obtained filtering value is a filtering value corresponding to natural language information.

Further, taking the natural language information input by the user obtained by the server as “Shenzhen university student small loan distribution” as an example. The server obtains the natural language information input by the user, and extracts the keyword "small loan" in the natural language information, obtains a filtering mechanism "loan screening" corresponding to the keyword "small loan", and according to the filtering mechanism and the filter value. The preset correspondence relationship between them is obtained by the screening value "amount" corresponding to the screening mechanism "debit screening".

S210: The server uses the filtering value to filter the initial semantic parsing result, obtains a semantic parsing result that matches the filtering value, and sends the semantic parsing result to the terminal.

Specifically, the server obtains the initial parsing data corresponding to the initial semantic parsing result, and uses the filtering value corresponding to the keyword to perform a filtering operation on the initial parsing data to obtain the initial parsing data through the filtering operation. The server obtains the initial semantic parsing result corresponding to the initial parsing data that passed the filtering operation, obtains the semantic parsing result according to the initial semantic parsing result corresponding to the initial parsing data that passes the filtering operation, and sends the semantic parsing result to the terminal.

The filtering value is used to filter the initial parsing data corresponding to the initial semantic parsing result. Different natural language information corresponds to different filtering values. The filtering value can be used to obtain the semantic parsing result consistent with the natural language information. The initial parsing data corresponding to the initial semantic parsing result includes the initial data for parsing natural language information.

Further, taking the natural language information input by the user as the “Shenzhen university student small loan distribution” as an example. The server performs a semantic analysis on the natural language information “Shenzhen University Students’ Small Loan Distribution ”according to the semantic analysis method in the context of small loans. The initial semantic analysis results obtained include the following situations:“ Shenzhen University Students ’Small Loan Amount Distribution "," Regional distribution of Shenzhen university students 'small loans "and" Shenzhen university students' small loans ".

The server uses the obtained filtering value to filter a plurality of possible initial semantic parsing results, and obtains the initial semantic results that match the filtering value. In this embodiment, the available filtering value corresponding to the natural language information is “amount”, that is, for the natural language information “Shenzhen University Student Small Loan Distribution”, using the filtering value “amount” to analyze a plurality of possible initial semantics The results were screened, and the initial semantic parsing result that matched the screened value "amount" was "the distribution of the amount of small loans of Shenzhen university students".

In the above semantic language parsing method, the server receives a semantic parsing request for natural language information sent by the terminal, and takes a semantic scene corresponding to the natural language information. In the corresponding semantic scenario, natural language information is parsed by using a preset semantic parsing method to obtain an initial semantic parsing result. According to the preset correspondence between keywords and filtered values, obtain filtered values corresponding to natural language information, and use the filtered values to filter the initial semantic parsing results to obtain the semantic parsing results that match the filtered values. Send to terminal. Therefore, before the semantic parsing result is sent to the terminal, further screening of the semantic parsing result is realized by using the filtering value, which improves the accuracy of the semantic parsing result, avoids repeated semantic parsing, saves resources, and reduces consumption.

In one embodiment, as shown in FIG. 3, a step for acquiring a semantic scene of natural language information is provided, including:

S302. The server extracts keywords in the natural language information, and obtains keyword attributes corresponding to different keywords.

Specifically, after receiving the semantic parsing request sent by the terminal, the server can obtain the natural language information input by the user on the terminal, and extract keywords in the natural language information. Since keywords in natural language information correspond to different keyword attributes, the server can obtain the correspondence between keywords and keyword attributes, and then obtain the correspondence with keywords according to the correspondence between keywords and keyword attributes. Keyword attributes.

S304. The server obtains a preset keyword category, and obtains a category attribute corresponding to the keyword category.

S306. The server classifies keywords in the natural language information according to the category attribute according to a preset correspondence relationship between the category attribute and the keyword attribute.

Specifically, the server sets a plurality of keyword categories in advance, and respectively sets corresponding category attributes for the plurality of keyword categories. Because there is a correspondence relationship between the keyword attribute and the category attribute corresponding to the keyword category, the keyword category to which each keyword belongs can be obtained according to the preset correspondence relationship between the category attribute and the keyword attribute, and each key Words are classified according to the corresponding category.

S308: The server obtains semantic scenes corresponding to different keyword categories according to a preset correspondence relationship between the keyword categories and semantic scenes.

Specifically, different keyword categories correspond to different semantic scenarios, and the server may obtain a semantic scenario corresponding to a keyword category to which different keywords belong in the natural language information according to a preset correspondence relationship between the keyword categories and the semantic scenarios. .

Further, the natural language information input by the user at the terminal is "how is the education distribution of male borrowers in Shanghai" as an example. The server obtains the natural language information input by the user on the terminal, and receives a semantic parsing request for the natural language information sent by the terminal. According to the natural language information "how is the distribution of the education of male borrowers in Shanghai", the keywords "Shanghai", "male" "Borrowers" and "educational distribution" to obtain the keyword category to which each keyword belongs. "Shanghai" represents the geographical category and is used to restrict the geographical scope of the executing subject. "Male borrower" belongs to the category of behavior subject and is used to limit execution The subject of operations in natural language information, "educational distribution" represents a further definition of the question to be retrieved. After the server classifies each keyword in the natural language information according to the keyword category, it obtains a semantic scene corresponding to the keyword category. In this embodiment, for the natural language information “how is the education of male borrowers in Shanghai distributed” input by the user, the semantic scene corresponding to the keyword category is the loan scene.

In the above step of acquiring a semantic scene of natural language information, the server extracts keywords in the natural language information and obtains keyword attributes corresponding to different keywords. Obtain a preset keyword category, and obtain category attributes corresponding to the keyword category. Furthermore, according to the preset correspondence between the category attribute and the keyword attribute, the keywords in the natural language information are classified according to the category attribute, and according to the preset correspondence between the keyword category and the semantic scene, the difference is obtained. Semantic scenes corresponding to keyword categories. By obtaining the semantic scene corresponding to the keywords in the natural language information, the targeted semantic parsing in the corresponding semantic scene can be achieved, which further improves the accuracy of the semantic parsing.

In one embodiment, a step of obtaining a semantic scene corresponding to natural language information is provided, including:

The server calculates the relevance value between the natural language information and different semantic scenes respectively; sorts the semantic scenes according to the magnitude of the relevance value, and obtains the semantic scene corresponding to the maximum relevance value.

In the above step of obtaining the semantic scene of natural language information, the server calculates the correlation degree values between the natural language information and different semantic scenes separately, and sorts the semantic scenes according to the magnitude of the correlation degree value, so as to obtain the maximum correlation value corresponding Semantic scene. Since the semantic scene with the highest degree of relevance to natural language information is obtained, it is possible to determine whether the semantic scene meets the requirements of the corresponding natural language information before performing semantic parsing, and further improve the accuracy of semantic parsing.

In one embodiment, a step of parsing natural language information to obtain an initial semantic parsing result by using a preset semantic parsing method in a corresponding semantic scene is provided, including:

The server obtains the correspondence between the semantic scene corresponding to the maximum relevance value and the semantic parsing mode, and obtains the semantic parsing mode corresponding to the semantic scene corresponding to the maximum relevance value; parses the natural language information according to the semantic parsing mode to obtain the original Semantic analysis results; use the preset inspection rules to perform initial inspection on the original semantic analysis results to obtain the original semantic analysis results that meet the preset inspection rules; and obtain the initial semantic analysis results based on the original semantic analysis results that meet the preset inspection rules .

Specifically, the preset inspection rules are used for initial inspection of the original semantic parsing results, including checking the integrity and validity of the original semantic parsing results, and determining whether the original semantic parsing results are complete and valid. If the original semantic parsing result includes the semantic parsing of all keywords of the natural language information input by the user, it means that the original semantic parsing result passes the integrity check performed by the inspection rule. If the original semantic parsing result parses the keywords of natural language information semantically, it can effectively express the original input of natural language information, indicating that the original semantic parsing result has passed the validity check by the inspection rules.

Further, the case where the semantic scene corresponding to the user's natural language information is a borrowing scene is taken as an example. The server obtains the natural language information “Shenzhen university student micro-loan distribution” input by the user, and according to the keyword category to which the keywords in the natural language information entered by the user, obtains the corresponding semantic scene as “micro-loan”, according to the preset The corresponding relationship between the semantic scenes and the semantic parsing methods is used to obtain the semantic parsing method corresponding to the semantic scene as "small loans", and the natural language information "Shenzhen university students' small loan distribution" is semantically analyzed.

In the above step of obtaining the initial semantic parsing result, the server obtains the original semantic parsing result by obtaining the semantic parsing method corresponding to the semantic scene corresponding to the maximum relevance value, and analyzing the natural language information according to the semantic parsing method. Furthermore, a preset inspection rule is used to perform initial inspection on the original semantic parsing result to obtain an initial semantic parsing result that conforms to the preset inspection rule. It can test the completeness and validity of the original semantic parsing result, and obtain the initial semantic parsing result that meets the preset requirements. It further realizes the multi-faceted detection of the initial semantic parsing result, and improves the accuracy of the initial semantic parsing result.

In one embodiment, a step of obtaining a filtering value corresponding to natural language information according to a preset correspondence between a keyword and a filtering value is provided, including:

The server obtains a filtering mechanism corresponding to the keywords; and obtains a filtering value corresponding to the natural language information according to the preset relationship between the filtering mechanism and the filtering value.

In the above step of obtaining the filtering value corresponding to the natural language information, the server obtains the filtering value corresponding to the natural language information by acquiring the filtering mechanism corresponding to the keywords and according to the preset relationship between the filtering mechanism and the filtering value. It is possible to further determine the association relationship between keywords and filter values, ensure that the obtained filter values correspond to natural language information input by the user, and improve work efficiency.

In one embodiment, a step of filtering an initial semantic parsing result by using a filtering value to obtain a semantic parsing result that meets the filtering value, and sending the semantic parsing result to a terminal includes:

The server obtains the initial parsing data corresponding to the initial semantic parsing result; uses the filtering value corresponding to the keyword to perform the filtering operation on the initial parsing data; obtains the initial parsing data that passes the filtering operation, and obtains the corresponding parsing data that passes the filtering operation The initial semantic parsing result of the; based on the initial semantic parsing result corresponding to the initial parsing data through the filtering operation to obtain the semantic parsing result and send the semantic parsing result to the terminal.

In the above step of obtaining a semantic parsing result that matches the filtering value, the server obtains the initial parsing data corresponding to the initial semantic parsing result and uses the filtering value corresponding to the keyword to perform a filtering operation on the initial parsing data. Therefore, by implementing the screening of the initial semantic parsing results, the initial semantic parsing results matching the filtered values, and generating the semantic parsing results according to the initial semantic parsing results matching the filtered values, further ensuring the accuracy of the semantic parsing results.

It should be understood that although the steps in the flowchart of FIG. 2-3 are sequentially displayed in accordance with the directions of the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless explicitly stated in this document, the execution of these steps is not strictly limited, and these steps can be performed in other orders. Moreover, at least a part of the steps in Figure 2-3 may include multiple sub-steps or stages. These sub-steps or stages are not necessarily performed at the same time, but may be performed at different times. These sub-steps or stages The execution order of is not necessarily performed sequentially, but may be performed in turn or alternately with at least a part of another step or a sub-step or stage of another step.

In one embodiment, as shown in FIG. 4, a natural language semantic parsing device is provided, including: a receiving module 402, a semantic scene acquisition module 404, an initial semantic parsing result acquisition module 406, a filtering value acquisition module 408, and The semantic parsing result obtaining module 410, wherein:

The receiving module 402 is configured to receive a semantic parsing request for natural language information sent by a terminal.

The semantic scene acquisition module 404 is configured to acquire a semantic scene corresponding to natural language information.

The initial semantic parsing result acquisition module 406 is configured to parse natural language information by using a preset semantic parsing method in a corresponding semantic scene to obtain an initial semantic parsing result.

The filtering value obtaining module 408 is configured to obtain a filtering value corresponding to natural language information according to a preset correspondence between a keyword and a filtering value.

The semantic parsing result acquisition module 410 is configured to filter the initial semantic parsing result by using the screening value to obtain a semantic parsing result that matches the screening value, and send the semantic parsing result to the terminal.

In the above-mentioned natural language semantic analysis device, the server receives a semantic analysis request for natural language information sent by the terminal, and takes a semantic scene of the natural language information. In the corresponding semantic scenario, natural language information is parsed by using a preset semantic parsing method to obtain an initial semantic parsing result. According to the preset correspondence between keywords and filtered values, obtain filtered values corresponding to natural language information, and use the filtered values to filter the initial semantic parsing results to obtain the semantic parsing results that match the filtered values. Send to terminal. Therefore, before the semantic parsing result is sent to the terminal, further screening of the semantic parsing result is realized by using the filtering value, which improves the accuracy of the semantic parsing result, avoids repeated semantic parsing, saves resources, and reduces consumption.

In one embodiment, a semantic scene acquisition module is provided for:

Extract keywords in natural language information, and obtain keyword attributes corresponding to different keywords; obtain preset keyword categories, and obtain category attributes corresponding to keyword categories; according to the preset category attributes and keyword attributes The corresponding relationship is used to classify keywords in natural language information according to category attributes; according to a preset correspondence relationship between keyword categories and semantic scenes, semantic scenes corresponding to different keyword categories are obtained.

In the above semantic scene acquisition module, the server extracts keywords in natural language information and obtains keyword attributes corresponding to different keywords. Obtain a preset keyword category, and obtain category attributes corresponding to the keyword category. Furthermore, according to the preset correspondence between the category attribute and the keyword attribute, the keywords in the natural language information are classified according to the category attribute, and according to the preset correspondence between the keyword category and the semantic scene, the difference is obtained. Semantic scenes corresponding to keyword categories. By obtaining the semantic scene corresponding to the keywords in the natural language information, the targeted semantic parsing in the corresponding semantic scene can be achieved, which further improves the accuracy of the semantic parsing.

In one embodiment, a semantic scene acquisition module is provided, which is further used for:

Calculate the relevance value between natural language information and different semantic scenes; sort the semantic scenes according to the magnitude of the relevance value, and obtain the semantic scene corresponding to the maximum relevance value.

In the above semantic scene acquisition module, the server calculates the correlation degree values between the natural language information and different semantic scenes separately, and sorts the semantic scenes according to the magnitude of the correlation degree values, so as to obtain the semantic scene corresponding to the maximum correlation degree value. Since the semantic scene with the highest degree of relevance to natural language information is obtained, it is possible to determine whether the semantic scene meets the requirements of the corresponding natural language information before performing semantic parsing, and further improve the accuracy of semantic parsing.

In one of the embodiments, an initial semantic parsing result acquisition module is provided for:

Obtain the correspondence between the semantic scene corresponding to the maximum relevance value and the semantic parsing mode, and obtain the semantic parsing mode corresponding to the semantic scene corresponding to the maximum relevance value; parse the natural language information according to the semantic parsing mode to obtain the original semantics Analyze the results; use the preset inspection rules to perform initial inspection on the original semantic parsing results to obtain the original semantic parsing results that meet the preset inspection rules; and obtain the initial semantic parsing results according to the original semantic parsing results that conform to the preset inspection rules.

In the above initial semantic parsing result acquisition module, the server obtains the original semantic parsing result by acquiring the semantic parsing method corresponding to the semantic scene corresponding to the maximum relevance value, and analyzing the natural language information according to the semantic parsing method. Furthermore, a preset inspection rule is used to perform initial inspection on the original semantic parsing result to obtain an initial semantic parsing result that conforms to the preset inspection rule. It can test the completeness and validity of the original semantic parsing result, and obtain the initial semantic parsing result that meets the preset requirements. It further realizes the multi-faceted detection of the initial semantic parsing result, and improves the accuracy of the initial semantic parsing result.

In one embodiment, a filtering value obtaining module is provided, which is used for:

Obtain the filtering mechanism corresponding to the keywords; according to the corresponding relationship between the preset filtering mechanism and the filtering value, obtain the filtering value corresponding to the natural language information.

In the above-mentioned filtering value acquisition module, the server obtains a filtering value corresponding to natural language information by acquiring a filtering mechanism corresponding to a keyword, and according to a correspondence relationship between a preset filtering mechanism and the filtering value. It is possible to further determine the relationship between keywords and filtered values, ensure that the obtained filtered values correspond to natural language information input by the user, and improve work efficiency.

In one of the embodiments, a semantic parsing result acquisition module is provided for:

Obtain initial parsing data corresponding to the initial semantic parsing result; use the filtering value corresponding to the keyword to perform the filtering operation on the initial parsing data; obtain the initial parsing data that passes the filtering operation, and obtain the parsing data that corresponds to the initial parsing data that passes the filtering operation Initial semantic parsing result; obtain the semantic parsing result according to the initial semantic parsing result corresponding to the initial parsing data through the filtering operation, and send the semantic parsing result to the terminal.

In the above semantic parsing result acquisition module, the server obtains the initial parsing data corresponding to the initial semantic parsing result, and uses the filtering value corresponding to the keyword to perform a filtering operation on the initial parsing data. Therefore, by implementing the screening of the initial semantic parsing results, the initial semantic parsing results matching the filtered values, and generating the semantic parsing results according to the initial semantic parsing results matching the filtered values, further ensuring the accuracy of the semantic parsing results.

For the specific limitation of the semantic parsing device of natural language, please refer to the limitation on the semantic parsing method of natural language mentioned above, which will not be repeated here. Each module in the above-mentioned natural language semantic parsing device may be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules may be embedded in the hardware form or independent of the processor in the computer device, or may be stored in the memory of the computer device in the form of software, so that the processor calls and performs the operations corresponding to the above modules.

In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. 5. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile computer-readable storage medium and an internal memory. The non-volatile computer-readable storage medium stores an operating system, computer-readable instructions, and a database. The internal memory provides an environment for operating systems and computer-readable instructions in a non-volatile computer-readable storage medium. The database of the computer equipment is used to store semantic parsing data of natural language. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer-readable instructions are executed by a processor to implement a natural language semantic parsing method.

Those skilled in the art can understand that the structure shown in FIG. 5 is only a block diagram of a part of the structure related to the solution of the application, and does not constitute a limitation on the computer equipment to which the solution of the application is applied. The specific computer equipment may be Include more or fewer parts than shown in the figure, or combine certain parts, or have a different arrangement of parts.

A computer device includes a memory and one or more processors. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the method for preprocessing the imbalanced sample data provided in any one of the embodiments of the present application is implemented. A step of.

One or more non-volatile computer-readable storage media storing computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors implement one of the embodiments of the present application Provides steps for pre-processing methods for unbalanced sample data.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by computer-readable instructions to instruct related hardware. The computer-readable instructions can be stored in a non-volatile computer. In the readable storage medium, the computer-readable instructions, when executed, may include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Synchlink DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above embodiments can be arbitrarily combined. In order to make the description concise, all possible combinations of the technical features in the above embodiments have not been described. However, as long as there is no contradiction in the combination of these technical features, they should be It is considered to be the range described in this specification.

The above-mentioned embodiments only express several implementation manners of the present application, and their descriptions are more specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be noted that, for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the protection scope of this application patent shall be subject to the appended claims.

Claims

A natural language semantic analysis method, including:

Receiving a semantic parsing request for natural language information sent by a terminal;

Acquiring a semantic scene corresponding to the natural language information;

Parsing the natural language information by using a preset semantic parsing method under the corresponding semantic scenario to obtain an initial semantic parsing result;

Obtaining a filtering value corresponding to the natural language information according to a preset correspondence between keywords and filtering values; and

The initial semantic parsing result is filtered using the screening value to obtain a semantic parsing result that matches the filtered value, and the semantic parsing result is sent to a terminal.
The method according to claim 1, wherein the acquiring a semantic scene corresponding to the natural language information comprises:

Calculate the degree of relevance between natural language information and different semantic scenes; and

Sort the semantic scenes according to the magnitude of the relevance degree value, and obtain the semantic scene corresponding to the largest relevance degree value.
The method according to claim 1, wherein the parsing the natural language information by using a preset semantic parsing method under the corresponding semantic scene to obtain an initial semantic parsing result comprises:

Acquiring the correspondence between the semantic scene corresponding to the maximum relevance value and the semantic parsing mode, and acquiring the semantic parsing mode corresponding to the semantic scene corresponding to the maximum relevance value;

Parse the natural language information according to the semantic parsing method to obtain the original semantic parsing result;

Perform an initial inspection on the original semantic analysis result using a preset inspection rule to obtain an original semantic analysis result that conforms to the preset inspection rule; and

An initial semantic parsing result is obtained according to the original semantic parsing result that conforms to a preset inspection rule.
The method according to any one of claims 1 to 2, wherein the acquiring a filtering value corresponding to the natural language information according to a preset correspondence between a keyword and a filtering value comprises:

Obtaining a filtering mechanism corresponding to the keywords; and

A filtering value corresponding to the natural language information is obtained according to a preset relationship between a preset filtering mechanism and a filtering value.
The method according to any one of claims 1 to 2, wherein the initial semantic parsing result is filtered by using the filtering value to obtain a semantic parsing result that matches the filtering value, and the The semantic parsing results are sent to the terminal, including:

Acquiring initial parsing data corresponding to the initial semantic parsing result;

Performing a filtering operation on the initial parsed data by using a filtering value corresponding to the keyword;

Acquiring initial parsing data that passed the filtering operation, and acquiring initial semantic parsing results corresponding to the initial parsing data that passed the filtering operation; and

Obtaining a semantic parsing result according to an initial semantic parsing result corresponding to the initial parsing data through the filtering operation, and sending the semantic parsing result to a terminal.
The method according to claim 1, wherein the acquiring a semantic scene of the natural language information comprises:

Extract keywords in the natural language information, and obtain keyword attributes corresponding to different keywords;

Acquiring a preset keyword category, and acquiring category attributes corresponding to the keyword category;

Classify keywords in the natural language information according to a category attribute according to a preset correspondence relationship between category attributes and keyword attributes; and obtain according to a preset correspondence relationship between keyword categories and semantic scenes Semantic scenes corresponding to different said keyword categories.
A semantic parsing device for natural language, including:

A receiving module, configured to receive a semantic parsing request for natural language information sent by a terminal;

A semantic scene acquisition module, configured to acquire a semantic scene of the natural language information;

An initial semantic parsing result acquisition module, configured to parse the natural language information by using a preset semantic parsing method under the corresponding semantic scenario to obtain an initial semantic parsing result;

A filtering value obtaining module, configured to obtain a filtering value corresponding to the natural language information according to a preset correspondence between a keyword and the filtering value; and

The semantic parsing result acquisition module is configured to filter the initial semantic parsing result by using the screening value to obtain a semantic parsing result that matches the filtered value, and send the semantic parsing result to a terminal.
The apparatus according to claim 7, wherein the semantic scene acquisition module is further configured to:

Extract keywords in the natural language information, and obtain keyword attributes corresponding to different keywords;

Acquiring a preset keyword category, and acquiring category attributes corresponding to the keyword category;

Classifying keywords in the natural language information according to a category attribute according to a preset correspondence relationship between category attributes and keyword attributes; and

According to a preset correspondence relationship between a keyword category and a semantic scene, a semantic scene corresponding to a different keyword category is obtained.
The apparatus according to claim 7, wherein the initial semantic parsing result obtaining module is further configured to:

Acquiring the correspondence between the semantic scene corresponding to the maximum relevance value and the semantic parsing mode, and acquiring the semantic parsing mode corresponding to the semantic scene corresponding to the maximum relevance value;

Parse the natural language information according to the semantic parsing method to obtain the original semantic parsing result;

Perform an initial inspection on the original semantic analysis result using a preset inspection rule to obtain an original semantic analysis result that conforms to the preset inspection rule; and

An initial semantic parsing result is obtained according to the original semantic parsing result that conforms to a preset inspection rule.
A computer device includes a memory and one or more processors. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the one or more processors, the one or more processors are Each processor performs the following steps:

Receiving a semantic parsing request for natural language information sent by a terminal;

Acquiring a semantic scene corresponding to the natural language information;

Parsing the natural language information by using a preset semantic parsing method under the corresponding semantic scenario to obtain an initial semantic parsing result;

Obtaining a filtering value corresponding to the natural language information according to a preset correspondence between keywords and filtering values; and

The initial semantic parsing result is filtered using the screening value to obtain a semantic parsing result that matches the filtered value, and the semantic parsing result is sent to a terminal.
The computer device according to claim 10, wherein the processor further executes the following steps when executing the computer-readable instructions:

Calculate the degree of relevance between natural language information and different semantic scenes; and

Sort the semantic scenes according to the magnitude of the relevance degree value, and obtain the semantic scene corresponding to the largest relevance degree value.
The computer device according to claim 10, wherein the processor further executes the following steps when executing the computer-readable instructions:

Acquiring the correspondence between the semantic scene corresponding to the maximum relevance value and the semantic parsing mode, and acquiring the semantic parsing mode corresponding to the semantic scene corresponding to the maximum relevance value;

Parse the natural language information according to the semantic parsing method to obtain the original semantic parsing result;

Perform an initial inspection on the original semantic analysis result using a preset inspection rule to obtain an original semantic analysis result that conforms to the preset inspection rule; and

An initial semantic parsing result is obtained according to the original semantic parsing result that conforms to a preset inspection rule.
The computer device according to any one of claims 10 to 11, wherein the processor further executes the following steps when executing the computer-readable instructions:

Obtaining a filtering mechanism corresponding to the keywords; and

A filtering value corresponding to the natural language information is obtained according to a preset relationship between a preset filtering mechanism and a filtering value.
The computer device according to any one of claims 10 to 11, wherein the processor further executes the following steps when executing the computer-readable instructions:

Acquiring initial parsing data corresponding to the initial semantic parsing result;

Performing a filtering operation on the initial parsed data by using a filtering value corresponding to the keyword;

Acquiring initial parsing data that passed the filtering operation, and acquiring initial semantic parsing results corresponding to the initial parsing data that passed the filtering operation; and

Obtaining a semantic parsing result according to an initial semantic parsing result corresponding to the initial parsing data through the filtering operation, and sending the semantic parsing result to a terminal.
The computer device according to claim 10, wherein the processor further executes the following steps when executing the computer-readable instructions:

Extract keywords in the natural language information, and obtain keyword attributes corresponding to different keywords;

Acquiring a preset keyword category, and acquiring category attributes corresponding to the keyword category;

Classify keywords in the natural language information according to a category attribute according to a preset correspondence relationship between category attributes and keyword attributes; and obtain according to a preset correspondence relationship between keyword categories and semantic scenes Semantic scenes corresponding to different said keyword categories.
One or more non-transitory computer-readable storage media storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the following steps:

Receiving a semantic parsing request for natural language information sent by a terminal;

Acquiring a semantic scene corresponding to the natural language information;

Parsing the natural language information by using a preset semantic parsing method under the corresponding semantic scenario to obtain an initial semantic parsing result;

Obtaining a filtering value corresponding to the natural language information according to a preset correspondence between keywords and filtering values; and

The initial semantic parsing result is filtered using the screening value to obtain a semantic parsing result that matches the filtered value, and the semantic parsing result is sent to a terminal.
The storage medium according to claim 16, wherein when the computer-readable instructions are executed by the processor, the following steps are further performed:

Calculate the degree of relevance between natural language information and different semantic scenes; and

Sort the semantic scenes according to the magnitude of the relevance degree value, and obtain the semantic scene corresponding to the largest relevance degree value.
The storage medium according to claim 16, wherein when the computer-readable instructions are executed by the processor, the following steps are further performed:

Acquiring the correspondence between the semantic scene corresponding to the maximum relevance value and the semantic parsing mode, and acquiring the semantic parsing mode corresponding to the semantic scene corresponding to the maximum relevance value;

Parse the natural language information according to the semantic parsing method to obtain the original semantic parsing result;

Perform an initial inspection on the original semantic analysis result using a preset inspection rule to obtain an original semantic analysis result that conforms to the preset inspection rule; and

An initial semantic parsing result is obtained according to the original semantic parsing result that conforms to a preset inspection rule.
The storage medium according to any one of claims 16 to 17, wherein when the computer-readable instructions are executed by the processor, the following steps are further performed:

Obtaining a filtering mechanism corresponding to the keywords; and

A filtering value corresponding to the natural language information is obtained according to a preset relationship between a preset filtering mechanism and a filtering value.
The storage medium according to any one of claims 16 to 17, wherein when the computer-readable instructions are executed by the processor, the following steps are further performed:

Extract keywords in the natural language information, and obtain keyword attributes corresponding to different keywords;

Acquiring a preset keyword category, and acquiring category attributes corresponding to the keyword category;

Classify keywords in the natural language information according to a category attribute according to a preset correspondence relationship between category attributes and keyword attributes; and obtain according to a preset correspondence relationship between keyword categories and semantic scenes Semantic scenes corresponding to different said keyword categories.