Table-based Interactive Question Answering Method and System

Abstract

The present application discloses a table-based interactive question and answer method and system. The method includes: receiving an input question in natural language provided by a user; constructing an SQL query statement according to the input question and predefined table metadata; executing the SQL query statement to obtain a query result; determining whether to trigger a rhetorical question mechanism according to the query result, and if so, constructing a rhetorical question according to the query result and providing it to the user, so that the user can make a targeted response, thereby reducing the number of interactions and improving the user experience.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technologies, and in particular, to a table-based interactive question answering method and system. Background Art

[0002] With the development of computer technologies, more and more users obtain the required information through the network, such as online querying information about tickets, catering, etc. In order to save labor costs and improve service efficiency, intelligent question answering systems have started to be used in these scenarios. In the prior art, intelligent question answering systems are classified according to the application fields they involve, and can be divided into restricted-domain question answering systems and open-domain question answering systems. In a restricted-domain question answering system, the system designs tables and table data in advance, and limits the questions input by the user within the range corresponding to the table data. When receiving a question input by the user in natural language, an SQL query statement is constructed based on the table and the table data (this process is called NL2SQL), and an answer is constructed according to the execution result of the SQL query statement. A table is a storage means for structured data, and the structured data is organized and stored in the table according to the table structure, and is called table data.

[0003] However, in the restricted-domain question answering scenario, when the user asks a question, sometimes all requirements can be described clearly at one time, and sometimes multiple interactions are required to describe all requirements clearly. Therefore, in the question answering system, it is necessary to distinguish between scenarios that require interaction and those that do not, and also to reduce the number of interactions as much as possible in scenarios that require interaction to avoid affecting the user experience. Summary of the Invention

[0004] In view of this, the purpose of the present disclosure is to provide a table-based interactive question answering method and system to solve the above technical problems.

[0005] According to the first aspect of the present disclosure, an interactive question answering method is provided, including:

[0006] Receiving an input question in natural language provided by the user;

[0007] Constructing an SQL query statement according to the input question and predefined table metadata;

[0008] Executing the SQL query statement to obtain a query result;

[0009] Determining whether to trigger a rhetorical question mechanism according to the query result. If so, constructing a rhetorical question according to the query result and providing it to the user.

[0010] In some embodiments, the interactive Q&A method further includes: if it is determined according to the query result that the rhetorical question mechanism is not triggered, then determine whether to adopt a fallback strategy according to the query result. If so, provide the default response to the user; if not, organize the query result into a response and provide it to the user.

[0011] In some embodiments, when the query result is multiple records, the rhetorical question mechanism is triggered; when the number of query records is zero, the fallback strategy is adopted; when the query result is a single record, then organize the query result into a response and provide it to the user.

[0012] In some embodiments, when the query result is multiple records, add query columns to the SQL query statement to obtain a new SQL query statement, and organize the column values of the query columns in the query result of the new SQL query statement into multiple options as the rhetorical question.

[0013] In some embodiments, among the multiple records, at least two of the multiple column values of the query column are different.

[0014] In some embodiments, the constructing an SQL query statement according to the input question and predefined table metadata includes:

[0015] Select a first table from multiple candidate tables according to the predefined table metadata;

[0016] Obtain the column information of each column of the first table from the table metadata;

[0017] Input the column information of each column of the first table and the input question into a trained NL2SQL model to obtain the SQL query statement.

[0018] In some embodiments, the selecting a first table from multiple candidate tables according to the predefined table metadata includes:

[0019] Perform word segmentation on the input question to obtain at least one keyword;

[0020] Match the at least one keyword with the topic word of each table in the table metadata, and use the table corresponding to the successfully matched topic word as the first table.

[0021] According to a second aspect of the present disclosure, there is provided an interactive Q&A system, including:

[0022] A natural language understanding module, configured to receive a natural language input question provided by a user, and construct an SQL query statement according to the input question and predefined table metadata;

[0023] A query execution module for executing the SQL query statement to obtain a query result;

[0024] A response generation module for determining whether to trigger a rhetorical question mechanism according to the query result. If so, a rhetorical question is constructed based on the query result and provided to the user.

[0025] In some embodiments, the response generation module further includes: if it is determined not to trigger the rhetorical question mechanism according to the query result, it is determined whether to adopt a fallback strategy according to the query result. If so, a default response is provided to the user; if not, the query result is organized as a response and provided to the user.

[0026] In some embodiments, the response generation module includes:

[0027] When the query result is multiple records, the rhetorical question mechanism is triggered. When the number of query records is zero, the fallback strategy is adopted. When the query result is one record, the query result is organized as a response and provided to the user.

[0028] According to a third aspect of the present disclosure, an electronic device is provided, including a memory and a processor. The memory further stores computer instructions executable by the processor. When the computer instructions are executed, the interactive question-answering method described in any one of the above is implemented.

[0029] According to a fourth aspect of the present disclosure, a computer-readable medium is provided. The computer-readable medium stores computer instructions executable by an electronic device. When the computer instructions are executed, the interactive question-answering method described in any one of the above is implemented.

[0030] The table-based interactive question-answering method and system provided by the embodiments of the present disclosure perform natural language analysis on the input question to obtain an SQL query statement, and determine whether to perform a rhetorical question mechanism according to the query result of executing the SQL query statement. If so, a rhetorical question is constructed based on the query result and provided to the user, so that the user can make a targeted response, thereby reducing the number of interactions and improving the user experience. More specifically, when the query result is multiple records, the rhetorical question mechanism is triggered, and the column values of specific columns are organized into a rhetorical question with multiple options for the user to select from among the multiple options. This not only facilitates user operation but also helps reduce the number of interactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Through the description of the embodiments of the present disclosure with reference to the following drawings, the above and other objects, features, and advantages of the present disclosure will become clearer. In the drawings:

[0032] Figure 1 is an exemplary hardware deployment diagram;

[0033] Figure 2 It is an example diagram of the client interface of an intelligent customer service system;

[0034] Figure 3 It is the structural diagram of the table-based interactive question and answer system provided by the embodiments of the present disclosure;

[0035] Figure 4 It is the flowchart of the table-based interactive question and answer method provided by the embodiments of the present disclosure;

[0036] Figure 5 It is the schematic structural diagram of an exemplary electronic device. Detailed implementation manners

[0037] The following describes the present disclosure based on embodiments, but the present disclosure is not limited to these embodiments. In the following detailed description of the present disclosure, some specific details are described in detail. Those skilled in the art can fully understand the present disclosure without the description of these details. In order to avoid obscuring the essence of the present disclosure, well-known methods, processes, and procedures are not described in detail. Additionally, the drawings are not necessarily drawn to scale.

[0038] Figure 1 It is an exemplary hardware deployment diagram. As Figure 1 shown, various terminals 101 establish communications with a server 103 through a network 102. The terminals 101 are various electronic devices adopting computer systems, such as personal computers, mobile phones, notebooks, tablet computers, and so on. The network 102 can be based on one or a combination of various communication technologies implemented by exchanging signals, including but not limited to wired technologies adopting electrical and / or optical transmission cables, and wireless technologies adopting infrared, radio frequency, and / or other forms. In different application scenarios, the network 102 can be the Internet, a wide area network, or a local area network, such as a company's proprietary network. The network 102 can also be a wired network or a wireless network. The server 103 can be a physical server or a cloud server. A cloud server can integrate the hardware and software resources of a physical server by using virtualization technology, and then provide computing power, storage capacity, and various application services based on the needs of users. The database 104 is used to organize and store data.

[0039] The table-based interactive dialogue system provided by the embodiments of the present disclosure can be deployed on the server 103. Figure 2It is an example diagram of the interface of the client of an interactive response system. The user can open the client interface on the terminal 101 to access the interactive dialogue system on the server 103. As shown in the figure, the selected table Q&A is selected, that is, based on the table data, the intelligent customer service gives corresponding answers to the questions input by the user. Among them, the input questions of the user are displayed on the right side of the interface, and the responses given by the system are displayed on the right side.

[0040] The intelligent customer service system can be deployed on the server 103. Figure 2 It is an example diagram of the interface of the client of an intelligent customer service system. The user can open the client interface on the terminal 101 to access the intelligent customer service system on the server 103. The intelligent customer service system includes two implementation methods: the response engine (open-domain Q&A) and the table Q&A (restricted-domain Q&A). When the user selects "table Q&A", that is, selects the response provided by the table-based interactive response system provided by the embodiments of the present disclosure. The figure schematically shows an example of the interaction between the table-based user and the system.

[0041] Figure 3 It is the structural diagram of the table-based interactive Q&A system provided by the embodiments of the present disclosure.

[0042] As shown in the figure, the system includes a natural language understanding module 206, a query execution module 203, and a response generation module 205.

[0043] The natural language understanding module 206 is used to construct an SQL query statement according to the input question in natural language and the predefined table metadata after receiving the question input by the user. Here, the table metadata refers to the schema information of the table data stored in the second database 202, and may include columns, column attributes (text type, number type, etc.), and the de-duplicated column values. As shown in the figure, the table metadata is stored in the first database 201.

[0044] Among them, in the step of constructing an SQL query statement according to the input question in natural language and the predefined table metadata, it usually includes three sub-steps: sub-step 1, selecting the table that can meet the input question from all candidate tables based on the predefined table metadata; sub-step 2, obtaining the column information of each column of the selected table; sub-step 3, obtaining the column information of each column of the selected table from the table metadata; inputting the column information of each column of the selected table and the input question into the trained NL2SQL model to obtain the SQL query statement.

[0045] In some embodiments, the table metadata may include subject terms. For example, the table Containers stores relevant information about container transportation, including price, type, etc. In the table metadata, "container transportation" is used as the subject term for the table Containers. That is, the corresponding records included in the table metadata are: Containers (table name), "container transportation" (subject term). Then, based on the predefined table metadata, the tables that can meet the input question are selected from all candidate tables, including: performing word segmentation on the input question and expanding the segmented words to obtain multiple subject retrieval terms, retrieving the table metadata using the subject retrieval terms, obtaining the subject terms that match the subject retrieval terms according to the retrieval results, and using the tables corresponding to the matching subject terms as the selected tables.

[0046] In a further embodiment, after selecting a certain table that can meet the input question from all candidate tables according to the subject term, other tables that have an association relationship with this table are also included. Other tables that have an association relationship with this table, for example, a table that has a primary-foreign key relationship with this table.

[0047] The query execution module 203 uses an SQL query statement to query the table data in the second database 202 to obtain a query result. In some embodiments, the query execution module 203 receives an SQL query statement from the natural language understanding module 206, retrieves the historical context information stored in the third database 204, and updates the SQL query statement according to the historical context information to obtain an updated SQL query statement, and uses the updated SQL query statement to query the table data in the second database 202 to obtain a query result. The historical context information stored in the third database 204 may include the SQL query statements accumulated historically corresponding to each question, whether each question uses a rhetorical mechanism, the execution results of the SQL query statements, and so on.

[0048] The response generation module 205 is used to determine whether to trigger a rhetorical mechanism according to the query result. If so, a rhetorical question is constructed based on the query result and provided to the user so that the user can give a targeted response. If not, it is determined whether to adopt a fallback strategy according to the query result. If so, a default response is provided to the user. If not, the query result is organized as a response and provided to the user.

[0049] In some embodiments, when the query execution module 203 has multiple records in the query result, it triggers the counter-question mechanism. When the number of query records is zero, it adopts the fallback strategy. When the query result has one record, it adopts the response strategy. Correspondingly, when the response generation module 205 receives the counter-question mechanism for the current input question, it constructs a counter-question based on the query result and provides it to the user to supplement information. When it receives the fallback strategy for the current question, it generates a default response and provides it to the user. When it receives the response strategy for the current question, it organizes the query result of one record into a response and provides it to the user. The purpose of the counter-question mechanism is to ask the user questions to supplement information, thereby reducing the number of records in the query result.

[0050] The following is an example based on Figure 2 and Table 1.

[0051] Table 1 Containers

[0052]

[0053]

[0054] For the user's input question "What is the price of the 20QH container trailer in Zhangzhou", the SQL query statement output by the natural language understanding module 206 is: "select 20QH price FROM Containers where city name = Zhangzhou". However, after the query execution module 203 executes this SQL statement and gets multiple records, it can reconstruct a new SQL statement: "SELECT county / district FROM Containers WHERE city name = Zhangzhou". The query execution module 203 executes this SQL statement based on Table 1 above and gets multiple records. Among these multiple records, there are three enumerated values for county / district: Changtai County, Longwen District, and Longhai City. Then these three enumerated values are constructed as Figure 2 The response is displayed on the interface. When the user selects Longhai City, the SQL query statement regenerated by the natural language understanding module 206 is: "select 20QH price FROM Containers where city name = Zhangzhou AND county / district = Longhai City". The query execution module 203 executes this SQL query statement according to Table 1 above and gets multiple records again. Then it reconstructs and executes a new SQL statement: select trailer type FROM Containers where city name = Zhangzhou AND county / district = Longhai City. Among the multiple records obtained, there are three enumerated values: regular trailer, pile-inserted trailer, and towing trailer. Then these three enumerated values are constructed as Figure 2The responses are displayed on the interface for the user to select. When the user further selects a towed trailer, the SQL query statement output by the natural language understanding module 206 is: "select 20QH price where city name = Zhangzhou AND county / district = Longhai City AND trailer type = towed trailer". Then, the query execution module 203 executes this SQL query statement according to Table 1 above and obtains a record, and this record is constructed as Figure 2 The responses are displayed on the interface. Of course, after receiving the SQL statement provided by the natural language understanding module 206, the query execution module 203 can also update it according to the table metadata and then execute the updated SQL statement. For example, update the above SQL statement: select 20QH price FROM Containers where to: SELECT 20QH, county / district FROM Containers WHERE city name = Zhangzhou.

[0055] Figure 4 It is a flowchart of the table-based interactive Q&A method provided by an embodiment of the present disclosure, including the following steps.

[0056] In step S31, receive the user's input question in natural language.

[0057] In step S32, construct an SQL query statement according to the input question in natural language and the predefined table metadata.

[0058] In step S33, execute the SQL query statement and obtain the query result.

[0059] In step S34, judge the query result. If the query result is multiple records, execute step S36; if the query result is one record, execute step S35; if the query result is zero records, execute step S37.

[0060] In step S35, adopt a response strategy to organize the query result into a response and provide it to the user.

[0061] In step S36, trigger a rhetorical mechanism and construct a rhetorical question according to the query result and provide it to the user.

[0062] In step S37, adopt a fallback strategy and provide the default response to the user.

[0063] After steps S36 - S37 are executed, adjust to step S31 to continue receiving the user's input question in natural language.

[0064] This embodiment can be summarized as follows: First, construct an SQL query statement based on the input question and table metadata, then execute the SQL query statement to obtain the query result. When the query result contains multiple records, trigger the rhetorical question mechanism, construct a rhetorical question based on the query result and provide it to the user, so that the user can supplement information to narrow down the number of query results. When the query result has only one record, generate a response based on the numerical values of each column of this record and provide it to the user. When the query result has no records, that is, the input question of the user cannot find the corresponding answer, so provide the default response to the user to let the user re-enter the question.

[0065] In some embodiments, step S32 includes the following sub-steps: Select a table that meets the input question from multiple candidate tables according to the predefined table metadata, then obtain the column information of each column of the table from the table metadata, and then input the column information of each column of the table and the input question into the trained NL2SQL model to obtain an SQL query statement.

[0066] In some embodiments, the table metadata may include subject words. Then, selecting a table that can meet the input question from all candidate tables based on the predefined table metadata includes: performing word segmentation on the input question and expanding the word segmentation to obtain multiple subject retrieval words, retrieving the table metadata using the subject retrieval words, obtaining the subject words that match the subject retrieval words according to the retrieval result, and using the table corresponding to the matching subject words as the selected table.

[0067] In some embodiments, when the query result contains multiple records, for example, in the above example, using the SQL query statement: SELECT 20QH, County / District FROM Containers WHERE City Name = Zhangzhou, querying Table 1 to obtain multiple records, then construct a rhetorical question based on the enumeration values of County / District in the multiple records and provide it to the user.

[0068] In summary, the table-based interactive question-and-answer method and system provided by the embodiments of the present disclosure perform natural language analysis on the input question to obtain an SQL query statement, and determine whether to perform the rhetorical question mechanism according to the query result of executing the SQL query statement. If so, construct a rhetorical question based on the query result and provide it to the user, so that the user can make a targeted response, thereby reducing the number of interactions and improving the user experience. More specifically, when the query result contains multiple records, trigger the rhetorical question mechanism, and organize the column numerical values of specific columns into a rhetorical question with multiple options for the user to select from among the multiple options. This not only facilitates user operations but also helps reduce the number of interactions.

[0069] Figure 5 It is a schematic structural diagram of an exemplary electronic device. The electronic device can be Figure 1The server 103 therein is used to run the form response system of the embodiments of the present disclosure. As Figure 5 shown, at the hardware level, it includes a memory 502 and a processor 501. In addition, in some cases, it also includes an input / output device 503 and other hardware 504. The memory 502 is, for example, a high-speed random access memory (RAM), and may also be a non-volatile memory, such as at least one disk memory, etc. The input / output device 503 is, for example, devices such as a display, a keyboard, a mouse, a network controller, etc. The processor 501 can be constructed based on various models of processors currently on the market. The processor 501, the memory 502, the input / output device 503, and other hardware 504 are interconnected by a bus, and the bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, only one line is shown in the figure, but it does not mean that there is only one bus or one type of bus.

[0070] The memory 502 is used to store programs. Specifically, the program may include program code, and the program code includes computer instructions. The memory may include a memory and a non-volatile memory, and provide computer instructions and data to the processor 501. The processor 501 reads the corresponding computer program from the memory 502 into the memory and then runs it, and implements the dialogue method provided in the above embodiments at the logical level.

[0071] Those skilled in the art can understand that the present disclosure can be implemented as a system, a method, and a computer program product. Therefore, the present disclosure can be specifically implemented in the following forms, namely, completely hardware, completely software (including firmware, resident software, microcode), and can also be implemented in the form of a combination of software and hardware.

[0072] In addition, in some embodiments, the present disclosure can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable media contains computer-readable program code.

[0073] A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium, for example but not limited to, is a system, apparatus, or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples of the computer-readable storage medium include: an electrical connection of one or more specific wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical memory, a magnetic memory, or any suitable combination of the above. In this article, a computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by, or in combination with, a processing unit, apparatus, or device.

[0074] A computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any other suitable combination. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, and this computer-readable medium can send, propagate, or transmit a program for use by, or in combination with, an instruction system, apparatus, or device.

[0075] The program code contained on a computer-readable medium can be transmitted by any appropriate medium, including but not limited to wireless, wire, optical cable, RF, etc., and any suitable combination of the above.

[0076] The above are only the preferred embodiments of the present disclosure and are not used to limit the present disclosure. For those skilled in the art, various modifications and changes can be made to the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present disclosure shall be included within the protection scope of the present disclosure.

Claims

1. A table - based interactive question - answering method, comprising: Receiving an input question in natural language provided by a user; Constructing an SQL query statement according to the input question and predefined table metadata, wherein the SQL query statement is determined by the input question and column information, and the column information is obtained by matching the predefined table metadata with the input question from multiple candidate tables; Executing the SQL query statement to obtain a query result; Determining whether to trigger a rhetorical question mechanism according to the number of records in the query result. If so, constructing a rhetorical question based on the query result and providing it to the user, wherein the rhetorical question mechanism is used to construct a rhetorical question with multiple options based on the column values of the query columns in the query result and provide it to the user; Wherein, the method further comprises: If it is determined not to trigger the rhetorical question mechanism, determining whether to adopt a fallback strategy according to the number of records in the query result; if the fallback strategy is adopted, providing a default response to the user, wherein the default response is used to prompt the user to re - enter the question.

2. The interactive question and answer method according to claim 1 further includes: If it is determined not to trigger the rhetorical question mechanism according to the query result, determining whether to adopt a fallback strategy according to the query result. If not, organizing the query result into a response and providing it to the user.

3. The interactive question and answer method according to claim 2, wherein When the query result is multiple records, triggering the rhetorical question mechanism; when the query result is zero records, adopting the fallback strategy; when the query result is one record, organizing the query result into a response and providing it to the user.

4. The interactive question-and-answer method according to claim 3, wherein, When the query result is multiple records, adding query columns to the SQL query statement to obtain a new SQL query statement, and organizing the column values of the query columns in the query result of the new SQL query statement into multiple options as the rhetorical question.

5. The interactive Q&A method according to claim 4, wherein Among the multiple records, at least two of the multiple column values of the query column are different.

6. The interactive question and answer method according to claim 1, wherein, The constructing of the SQL query statement according to the input question and predefined table metadata includes: Selecting a first table from multiple candidate tables according to the predefined table metadata; Obtaining the column information of each column of the first table from the table metadata; Inputting the column information of each column of the first table and the input question into a trained NL2SQL model to obtain the SQL query statement.

7. The interactive question-answering method according to claim 1, wherein, The selecting of the first table from multiple candidate tables according to the predefined table metadata includes: Performing word - segmentation processing on the input question to obtain at least one keyword; Matching the at least one keyword with the topic words of each table in the table metadata, and taking the table corresponding to the successfully - matched topic word as the first table.

8. A table - based interactive question - answering system, comprising: A natural language understanding module, configured to receive an input question in natural language provided by a user, and construct an SQL query statement according to the input question and predefined table metadata, wherein the SQL query statement is determined by the input question and column information, and the column information is obtained by matching the predefined table metadata with the input question from multiple candidate tables; A query execution module, configured to execute the SQL query statement to obtain a query result; A response generation module, configured to determine whether to trigger a rhetorical question mechanism according to the number of records in the query result. If so, construct a rhetorical question based on the query result and provide it to the user, where the rhetorical question mechanism is used to construct a rhetorical question with multiple options based on the column values of the query columns in the query result and provide it to the user; The response generation module is further configured to, if it is determined not to trigger the rhetorical question mechanism, determine whether to adopt a fallback strategy according to the number of records in the query result; if the fallback strategy is adopted, provide a default response to the user, where the default response is used to prompt the user to re-enter the question.

9. The interactive question and answer system according to claim 8, wherein the response generation module further comprises: If it is determined not to trigger the rhetorical question mechanism according to the query result, determine whether to adopt a fallback strategy according to the query result. If not, organize the query result into a response and provide it to the user.

10. The interactive question and answer system according to claim 9, wherein, The response generation module includes: When the query result is multiple records, trigger the rhetorical question mechanism. When the query result is zero records, adopt the fallback strategy. When the query result is one record, organize the query result into a response and provide it to the user.

11. An electronic device, including a memory and a processor, where the memory further stores computer instructions executable by the processor. When the computer instructions are executed, the interactive question-answering method according to any one of claims 1 to 7 is implemented.

12. A computer-readable medium, where the computer-readable medium stores computer instructions executable by an electronic device. When the computer instructions are executed, the interactive question-answering method according to any one of claims 1 to 7 is implemented.

Images