Structured data-based question answering method and sports table-based question answering method

Abstract

Provided in the embodiments of the present disclosure are a structured data-based question answering method and a sports table-based question answering method. The structured data-based question answering method comprises: acquiring text information to be processed and raw structured data corresponding to said text information; processing said text information and the raw structured data on the basis of a question-answering model to obtain target structured data; and generating question-answering text on the basis of said text information and the target structured data, and inputting the question-answering text into the question-answering model to obtain target answer information output by the question-answering model, wherein the question-answering text comprises tree-structured logic processing information for said text information, and the question-answering model generates the target answer information on the basis of the tree-structured logic processing information. The present method compresses tables to reduce the data volume of the tables, so as to acquire from a target table the target answer information for said text information, thereby effectively reducing the data processing volume of the question-answering model.

Description

Question answering method based on structured data and question answering method based on sports table TECHNICAL FIELD

[0001] Embodiments of the present disclosure relate to the technical field of computer, and particularly relate to a question answering method based on structured data and a question answering method based on sports table. BACKGROUND

[0002] Tables, as a form of semi-structured data, support various aspects of daily life and professional fields. Tables, including but not limited to databases, web pages, files, etc., are used in key applications such as financial analysis, risk management, health monitoring, and business reporting. The emergence of large language models (LLM) provides a new way of understanding and reasoning about data in tables.

[0003] The application of large language models to table understanding has shown good results on small-scale tables. However, in practical applications, tables are becoming increasingly complex, and the content of data in tables is becoming increasingly rich. Due to the complex interactions between rows and columns of tables, the amount of information that needs to be processed and understood grows exponentially, and this complexity seriously hinders the understanding ability of large language models. In the face of complex problem-solving logic, locating, extracting, and understanding tables become a major challenge for large language models. SUMMARY

[0004] Therefore, the embodiments of the present disclosure provide a question answering method based on structured data. One or more embodiments of the present disclosure also relate to a question answering method based on sports table, a question answering device based on structured data, a computing device, a computer-readable storage medium, and a computer program product to solve the technical defects in the related art.

[0005] According to a first aspect of the embodiments of the present disclosure, a question answering method based on structured data is provided, including: obtaining to-be-processed text information and original structured data corresponding to the to-be-processed text information; processing the to-be-processed text information and the original structured data based on a question answering model to obtain target structured data; generating a question answering text according to the to-be-processed text information and the target structured data, and inputting the question answering text into the question answering model to obtain target answer information output by the question answering model, wherein the question answering text includes tree structure logical processing information for the to-be-processed text information, and the question answering model generates the target answer information based on the tree structure logical processing information.

[0006] According to a second aspect of the embodiments of the present disclosure, a sports table-based question and answer method is provided, including: obtaining to-be-processed sports question information and original sports information table corresponding to the to-be-processed sports question information; processing the to-be-processed sports question information and the original sports information table based on a question and answer model to obtain a target sports information table; generating a question and answer text according to the to-be-processed sports question information and the target sports information table, and inputting the question and answer text into the question and answer model to obtain target answer information output by the question and answer model, wherein the question and answer text includes tree structure logical processing information for the to-be-processed sports question information, and the question and answer model generates the target answer information based on the tree structure logical processing information.

[0007] According to a third aspect of the embodiments of the present disclosure, a sports table-based question and answer method is provided, applied to a cloud side device, including: receiving to-be-processed sports question information sent by an end side device, and obtaining corresponding original sports information table according to the to-be-processed sports question information; processing the to-be-processed sports question information and the original sports information table based on a question and answer model to obtain a target sports information table; generating a question and answer text according to the to-be-processed sports question information and the target sports information table, and inputting the question and answer text into the question and answer model to obtain target answer information output by the question and answer model; and sending the target answer information to the end side device.

[0008] According to a fourth aspect of the embodiments of the present disclosure, an information processing method based on a question and answer model is provided, applied to a task platform, including: receiving a task generation request sent by an end side device, wherein the task generation request includes request information; obtaining a question and answer model based on the request information, wherein the question and answer model determines target structured data according to to-be-processed text information and original structured data, determines target answer information according to a question and answer text generated by the to-be-processed text information and the target structured data, the question and answer text includes tree structure logical processing information for the to-be-processed text information, and the question and answer model generates the target answer information based on the tree structure logical processing information; generating task information based on the question and answer model, wherein the task information is used for the end side device to execute a question and answer task based on structured data.

[0009] According to a fifth aspect of the embodiments of the present disclosure, a task platform is provided, comprising a request interface and a response unit; the request interface is configured to receive a task generation request sent by an end-side device, wherein the task generation request comprises request information; and the response unit is configured to obtain a question and answer model based on the request information, wherein the question and answer model is configured to determine target structured data according to to-be-processed text information and original structured data, and determine target answer information according to a question and answer text generated based on the to-be-processed text information and the target structured data, the question and answer text comprises tree structure logical processing information for the to-be-processed text information, and the question and answer model is configured to generate the target answer information based on the tree structure logical processing information.

[0010] According to a sixth aspect of the embodiments of the present disclosure, a question and answer device based on structured data is provided, comprising: an obtaining module configured to obtain to-be-processed text information and original structured data corresponding to the to-be-processed text information; a screening module configured to obtain target structured data by processing the to-be-processed text information and the original structured data based on a question and answer model; and a question and answer module configured to generate a question and answer text according to the to-be-processed text information and the target structured data, input the question and answer text into the question and answer model, and obtain target answer information output by the question and answer model, wherein the question and answer text comprises tree structure logical processing information for the to-be-processed text information, and the question and answer model is configured to generate the target answer information based on the tree structure logical processing information.

[0011] According to a seventh aspect of the embodiments of the present disclosure, a computing device is provided, comprising a memory and a processor; the memory is configured to store computer programs / instructions, and the processor is configured to execute the computer programs / instructions, so as to implement the steps of the above question and answer method based on structured data, question and answer method based on sports table, and information processing method based on question and answer model.

[0012] According to an eighth aspect of the embodiments of the present disclosure, a computer readable storage medium is provided, which stores computer programs / instructions, and the computer programs / instructions are executed by a processor to implement the steps of the above question and answer method based on structured data, question and answer method based on sports table, and information processing method based on question and answer model.

[0013] According to a ninth aspect of the embodiments of the present disclosure, a computer program product is provided, comprising computer programs / instructions, and the computer programs / instructions are executed by a processor to implement the steps of the above question and answer method based on structured data, question and answer method based on sports table, and information processing method based on question and answer model.

[0014] One embodiment of the present disclosure provides a question and answer method based on structured data, comprising: obtaining to-be-processed text information and original structured data corresponding to the to-be-processed text information; processing the to-be-processed text information and the original structured data based on a question and answer model to obtain target structured data; generating a question and answer text according to the to-be-processed text information and the target structured data, and inputting the question and answer text into the question and answer model to obtain target answer information output by the question and answer model, wherein the question and answer text includes tree structure logical processing information for the to-be-processed text information, and the question and answer model generates the target answer information based on the tree structure logical processing information.

[0015] Through the method provided by the embodiment of the present disclosure, after obtaining the to-be-processed text information and the original structured data, the original structured data is first screened according to the question and answer model, the target structured data related to the to-be-processed text information in the original structured data is determined according to the to-be-processed text information, and the data amount of the target structured data is effectively reduced. Then, the target structured data and the to-be-processed text information are input into the question and answer model for question and answer analysis, so that the question and answer model processes the to-be-processed text information and the target structured data based on the tree structure logical processing information, and obtains the target answer information of the to-be-processed text information from the target structured data, effectively reducing the data processing amount of the question and answer model and improving the data processing efficiency of the question and answer model. BRIEF DESCRIPTION OF DRAWINGS

[0016] FIG. 1 is a processing schematic diagram of a question and answer method based on structured data according to one embodiment of the present disclosure;

[0017] FIG. 2 is a flowchart of a question and answer method based on structured data according to one embodiment of the present disclosure;

[0018] FIG. 3 is a processing process flowchart of a question and answer method based on a sports table according to one embodiment of the present disclosure;

[0019] FIG. 4 is a processing process flowchart of a question and answer method based on a sports table applied to a cloud-side device according to one embodiment of the present disclosure;

[0020] FIG. 5 is a flowchart of an information processing method based on a question and answer model according to one embodiment of the present disclosure;

[0021] FIG. 6 is a schematic diagram of a task platform according to one embodiment of the present disclosure;

[0022] FIG. 7 is a structural schematic diagram of a question and answer device based on structured data according to one embodiment of the present disclosure;

[0023] FIG. 8 is an architecture diagram of a question and answer system based on structured data according to one embodiment of the present disclosure;

[0024] FIG. 9 is a structural block diagram of a computing device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0025] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, the present disclosure can be practiced without the specific details. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to obscure the present disclosure. Some portions of the detailed description are presented in terms of algorithms, procedures, logic blocks, processing and other symbolic representations of operations on data bits that can be stored within a computer memory. These algorithmic descriptions and representations can be the techniques used by those of ordinary skill in the data processing arts to convey the substance of their work to others.

[0026] The terminology used in the one or more embodiments of the present disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of the present disclosure. As used in the one or more embodiments of the present disclosure and the accompanying claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in the one or more embodiments of the present disclosure, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0027] It will be understood that, although the terms first, second, etc. can be used herein to describe various information, these terms are not intended to denote a temporal or chronological order. Rather, these terms are used solely to distinguish one from another. For example, without departing from the scope of the one or more embodiments of the present disclosure, first can be termed second, and similarly, second can be termed first. The term "if' as used herein can be interpreted as meaning "when" or "in response to determining" depending on the context.

[0028] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present disclosure are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards in relevant regions, and provide corresponding operation portal for user to choose authorization or refusal.

[0029] In one or more embodiments of the present disclosure, a large model refers to a deep learning model with a large number of model parameters, usually containing hundreds of millions, billions, tens of billions, hundreds of billions, or even tens of billions of model parameters. The large model can also be called a foundation model. Through large-scale unlabeled corpus pre-training, a pre-trained model with hundreds of millions of parameters is output. Such a model can adapt to a wide range of downstream tasks and has good generalization ability. For example, large language models (LLM) and multi-modal pre-training models.

[0030] In practical applications, a large model only needs a small amount of sample data to fine-tune the pre-trained model and can be applied to different tasks. Large models can be widely used in natural language processing (NLP) and computer vision fields. Specifically, they can be applied to computer vision tasks such as visual question answering (VQA), image captioning (IC), and image generation, as well as natural language processing tasks such as text-based sentiment classification, text summarization generation, and machine translation. The main application scenarios of large models include digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc.

[0031] First, the technical terms related to one or more embodiments of the present disclosure are explained.

[0032] Large Language Model (LLM): In the field of machine learning and artificial intelligence, a large language model typically refers to a deep learning model with a large number of parameters and complex structures. These models can learn and store a large amount of knowledge due to their large size, and have good generalization ability and the ability to solve complex problems. Large models can perform tasks in multiple tasks, multiple domains, and multiple modalities, and are widely used in natural language processing, image recognition, speech recognition, and other fields.

[0033] Tabular Understanding: Tabular understanding refers to the process of analyzing, understanding, and extracting structured information from tabular data. In the field of AI, tabular understanding may involve identifying rows, columns, titles, and cell contents in a table, understanding their relationships, and applying these relationships to specific tasks such as data analysis, information retrieval, or table-to-text generation.

[0034] Chain of Thought: Chain of thought is a conceptual model that simulates the process of human logic and reasoning. In AI, chain of thought can be understood as a series of logically linked thought steps that are connected by cause and effect, analogy or other associations, forming a process of solving problems or generating new ideas. By establishing such a chain of thought, AI systems can achieve more advanced reasoning and decision-making functions.

[0035] Tables, as a form of semi-structured data, play an important role in many aspects of daily life and professional fields. Tables include open data repositories, web pages, document tables, etc. They are widely used in financial analysis, risk management, health monitoring, business reporting, and other scenarios. The emergence of large language models has opened up new ways to understand and infer table data.

[0036] Compared with unstructured text, tables provide a dense structured format through the interaction of rows and columns. Rows usually represent the field information included in the table, and columns represent the business data corresponding to each field information. Tables provide a wealth of information sources. However, due to the structural characteristics of rows and columns, large language models face unique challenges because they need to perform high-level reasoning on text and numerical data. Given the increasing reliance on tables for data representation and the complexity of interpretation, research integrating large language models to improve large-scale tables has become an important research direction.

[0037] Currently, large language models often encounter difficulties when applied to more complex and larger tables in real-world scenarios. The limited context capacity of current large language models is a major challenge for table understanding. As the size of the table expands, the amount of information that needs to be processed and understood grows exponentially due to the complex interactions between rows and columns. This complexity makes it difficult for large language models to capture and reason about all the information at once, significantly hindering their understanding capabilities. In the face of complex problem-solving logic spanning long chains, locating, extracting, and understanding key information from tables becomes a significant challenge.

[0038] Currently, there are two methods to solve this problem. The first method is to use the architectural information of the table and use program-assisted methods, such as generating SQL-based answers based on the question. However, this method may result in long and error-prone SQL statements, leading to poor performance. The second method is to divide the table into multiple sub-tables to manage larger tables, but this approach is difficult to apply in practice and lacks practicality.

[0039] Based on this, in the present disclosure, a structured data-based question answering method is provided, and the present disclosure also relates to a sports table-based question answering method, a structured data-based question answering device, a computing device, a computer-readable storage medium, and a computer program product, which are described in detail in the following embodiments.

[0040] Referring to FIG. 1, FIG. 1 shows a processing schematic diagram of a structured data-based question answering method according to an embodiment of the present disclosure.

[0041] As shown in FIG. 1, the text information to be processed and the original structured data are obtained, wherein the text information to be processed can be understood as question information in a question answering scenario, and the original structured data is structured data for answering the question information in the question answering scenario. The structured data contains data for answering the question information, and the structured data can include tables, databases, and other structured or semi-structured data.

[0042] The original structured data in the question answering scenario contains a large amount of data information. Taking a table as an example, it can be one table or multiple tables. In the method provided in the embodiment of the present disclosure, the form of the original structured data is not limited.

[0043] After obtaining the text information to be processed and the original structured data, since the content of the original structured data is relatively large, it contains the content for answering the text information to be processed, but there are also a lot of redundant information, which is not conducive to the processing of all data in the original structured data by the question answering model.

[0044] Based on this, in the method provided in the embodiment of the present disclosure, the original structured data and the text information to be processed are first input into the question answering model for processing, so that the question answering model determines the target structured data related to the text information to be processed. Specifically, in order to improve the processing efficiency of the question answering model, the original structured data identifier in the original structured data and the example structured data, and the text information to be processed are input into the question answering model for processing. The question answering model determines the target structured data identifier related to the text information to be processed from the original structured data identifier. Then, the target structured data identifier and the text information to be processed are filtered in the original structured data to filter out the redundant information in the original structured data, and the target structured data for answering the text information to be processed is generated.

[0045] The target structured data and the to-be-processed text information are input into the question and answer model, so that the question and answer model splits the target structured data into a plurality of target structured sub-data based on given tree structure logic processing information and the to-be-processed text information, and generates reference answer information corresponding to each structured sub-data by analyzing each target structured sub-data. Finally, the final target answer information is generated according to each reference answer information.

[0046] By the method provided in the embodiments of the present disclosure, after obtaining the to-be-processed text information and the original structured data, the original structured data is first screened according to the question and answer model, the target structured data identifier related to the to-be-processed text information is determined in the original structured data according to the to-be-processed text information, and the target structured data is screened out in the original structured data according to the target structured data identifier and the to-be-processed text information, which effectively reduces the data amount of the target structured data. Then, the target structured data and the to-be-processed text information are input into the question and answer model for question and answer analysis, and the target answer information of the to-be-processed text information is obtained from the target structured data, which effectively reduces the data processing amount of the question and answer model and improves the data processing efficiency of the question and answer model.

[0047] Referring to FIG. 2, FIG. 2 shows a flowchart of a question and answer method based on structured data according to an embodiment of the present disclosure, which specifically includes the following steps.

[0048] Step 202: obtaining to-be-processed text information and original structured data corresponding to the to-be-processed text information.

[0049] The method provided in the embodiments of the present disclosure is applied to a terminal, which can be understood as any device with computing capability, such as a personal computer, a smart terminal, a smart wearable device, a server, a cloud server, etc.

[0050] The to-be-processed text information can be understood as a question text asked by a user, and the to-be-processed text information can be one or multiple. The user can input the to-be-processed text information through a user graphical interface of the terminal, and the terminal in the embodiments of the present disclosure obtains the to-be-processed text information.

[0051] The original structured data corresponding to the to-be-processed text information can be understood as structured data related to the to-be-processed text information, and the original structured data is unprocessed structured data. The original structured data includes a structured data identifier, the structured data identifier includes a plurality of structured data information, and the original structured data further includes data information corresponding to each data field. For example, taking a personnel table as an example, the structured data identifier can include table fields such as "name, department, birth date, age, and gender". Each structured data identifier corresponds to corresponding business data, such as "Zhang San, A department, 19900505, 34, and male".

[0052] In the method provided in the embodiments of the present disclosure, the original structured data can be a table, for example, an Excel table. It can also be multiple tables, and the multiple tables have an association relationship, for example, database tables. In the method provided in the embodiments of the present disclosure, the specific form of the original structured data is not limited. Actual application is used as the criterion.

[0053] In actual application, the original structured data can be sent by a user to a question and answer system in a question and answer scenario together with to-be-processed text information, or can be pre-stored in the question and answer system. In the method provided in the embodiments of the present disclosure, the acquisition manner of the original structured data is not limited.

[0054] In a specific embodiment provided in the present disclosure, taking the case that the original structured data is stored in the question and answer system as an example, the question and answer system in the embodiment can solve multiple types of problems.

[0055] The method provided in the embodiments of the present disclosure includes the following steps.

[0056] In the method provided in the embodiments of the present disclosure, the question and answer system in the terminal receives to-be-processed text information sent by a user. The question and answer system analyzes the to-be-processed text information, and determines text field information corresponding to the to-be-processed text information. The text field information can be understood as a question of what field the to-be-processed text information belongs to, for example, the text field information corresponding to the to-be-processed text information can be an economic field, a technological field, a sports field, an educational field, and the like.

[0057] A large amount of data is stored in the question and answer system, and various data is stored in different structured formats. In order to quickly find the structured data corresponding to the to-be-processed text information from the large amount of data, the various structured data can be pre-divided according to field information, for example, the structured data can be determined to belong to which field according to data information corresponding to each structured data, and each structured data is labeled with field information. When the text field information corresponding to the to-be-processed text information is obtained by analysis, the structured data related to the text field information can be selected as the original structured data by preliminarily screening the structured data in the question and answer system according to the text field information.

[0058] For example, there are 10 tables in the question and answer system, of which 3 tables are related to sports, 3 tables are related to technology, and 4 tables are related to economy. After receiving the text information to be processed, the question and answer system can obtain the text field information corresponding to the text information to be processed by analyzing the text information to be processed, and the text field information is "technology". Then, the 3 tables related to technology can be used as the original table corresponding to the text information to be processed.

[0059] By using the text field information of the text information to be processed to preliminarily screen the structured data in the question and answer system, the structured data related to the text field information is determined as the original structured data, which effectively reduces the data processing amount of the question and answer system in the subsequent processing process. The data processing efficiency of the question and answer system is improved.

[0060] Step 204: processing the text information to be processed and the original structured data based on the question and answer model to obtain target structured data.

[0061] The target structured data can be understood as structured data used to answer the text information to be processed. In the method provided in the present disclosure, after processing the text information to be processed and the original structured data based on the question and answer model, the corresponding target structured data can be obtained. The target structured data used to answer the text information to be processed can be screened from the original structured data, which can effectively reduce the data processing amount and improve the data processing efficiency of the question and answer model.

[0062] Specifically, processing the text information to be processed and the original structured data based on the question and answer model to obtain target structured data includes: generating structured data screening text according to the text information to be processed and the original structured data, and inputting the structured data screening text into the question and answer model to obtain target structured data identification information output by the question and answer model; and generating target structured data according to the target structured data identification information and the original structured data.

[0063] The structured data screening text can be understood as a prompt text used to input the original structured data into the question and answer model for screening. The question and answer model provided in the present disclosure can be understood as a large language model. The purpose of the prompt text is to convey the user's intention to the large language model in an efficient and accurate manner, so that the question and answer model can generate a corresponding response according to the prompt text. Correct and effective prompt text can greatly improve the interaction effect of the large language model. The main functions of the prompt text include clearly expressing the demand, guiding the large language model to think, and providing professional data.

[0064] In an embodiment of the present disclosure, the structured data screening text is provided for the purpose of compressing and screening the original structured data according to the to-be-processed text information. The original structured data is the original structured data form saved in the question and answer system. The amount of business data saved in the original structured data is also relatively large, and the context capacity that the question and answer model can process is limited. If the amount of business data in the original structured data exceeds the processing capacity of the question and answer model, the question and answer model cannot give an effective answer. Based on this, in the method provided in the embodiment of the present disclosure, the corresponding structured data screening text is generated according to the to-be-processed text information and the original structured data, which is used to prompt the question and answer model to compress the original structured data according to the to-be-processed text information. The question and answer model outputs target structured data identification information.

[0065] The target structured data identification information can be understood as structured data identification information related to the to-be-processed text information. For example, there are 100 fields in the table header of the original structured data, and the to-be-processed text information involves 10 fields among them. The structured data screening text generated according to the original structured data and the to-be-processed text information is input into the question and answer model, and the question and answer model outputs the target structured data identification information composed of the 10 fields. The target structured data identification information is a subset of the original structured data identification information of the original structured data.

[0066] In actual application, the size of the target structured data is large, and if all the original structured data is input into the question and answer model, the processing difficulty of the question and answer model will increase. In order to improve the processing efficiency of the question and answer model, in a specific implementation provided in the present disclosure, the structured data screening text is generated according to the to-be-processed text information and the original structured data, comprising: determining a preset data screening prompt text; generating a screening example text according to the original structured data, wherein the screening example text comprises original structured data identification information and example structured data corresponding to the original structured data identification information; and generating a structured data screening text according to the preset data screening prompt text, the screening example text and the to-be-processed text information.

[0067] The preset data screening prompt text can be understood as a default prompt text for structured data screening. For example, the preset data screening prompt text can be "please screen the original structured data identification according to the following screening example text". In actual application, the preset data screening prompt text can be a default text set in advance. When generating the structured data screening text, the preset data screening prompt text can be directly obtained.

[0068] It should be noted that the structured data screening text is text information for input into the data model, which includes preset data screening prompt text, screening example text and to-be-processed text information, and the structured data screening text corresponding to each to-be-processed text information is different. The preset data screening prompt text is a template text for generating structured data screening text, and different to-be-processed text information can correspond to the same preset data screening prompt text.

[0069] The screening example text can be understood as a structured data example text generated based on original structured data, and the screening example text includes original structured data identifiers of the original structured data and example business data corresponding to each original structured data identifier. The example business data can be understood as example data of specific business data in each original structured data identifier. The example business data is a subset of the business data in the original structured data. For example, taking an example of original structured data having 100 business fields and including 100,000 pieces of business data in the original structured data, the screening example text can be 100 business field composed original structured data identifiers and 10 example table data in the 100,000 pieces of business data.

[0070] Specifically, the screening example text is generated according to the original structured data, including: extracting original structured data identifier information in the original structured data; extracting example structured data in the original structured data based on a preset data quantity and the original structured data identifier information; and generating the screening example text according to the original structured data identifier information and the example structured data.

[0071] In actual application, the original structured data is composed of original structured data identifiers and original business data. The original structured data identifier can be directly extracted from the original structured data. In order to facilitate the question and answer model to understand the actual content of each business field in the original structured data identifier in the original structured data, a preset number of example structured data can be extracted from the original business data. Then, the original structured data identifier and the example structured data are spliced to generate the screening example text.

[0072] For example, there are 100 business fields in the original structured data, and there are 100,000 pieces of structured data in the original structured data. The original structured data identifier is extracted from the original structured data. It is determined that the preset data quantity is 5, and 5 pieces of example structured data are selected from the 100,000 pieces of structured data. The way of extracting the example structured data from the original structured data can be random extraction, can be extracting the latest five pieces of structured data, or can be extracting the earliest five pieces of structured data, and the like. In the method provided in the embodiments of the present disclosure, the specific implementation manner of extracting the example structured data in the original structured data based on the preset data quantity and the original structured data identifier is not limited, and is subject to actual application. Finally, the screening example text is spliced according to the original structured data identifier composed of 100 business fields and the 5 pieces of example structured data.

[0073] After obtaining the screening example text, the structured data screening text can be generated according to the preset data screening prompt text, the screening example text and the to-be-processed text information. And the structured data screening text is input into the question and answer model.

[0074] For example, the structured data screening text is: “Please screen the original structured data identifier according to the following screening example text; screening example text 1; to-be-processed text information 1”; wherein “Please screen the original structured data identifier according to the following screening example text” is the preset data screening prompt text, screening example text 1 is the screening example text, and to-be-processed text information 1 is the to-be-processed text information.

[0075] In the foregoing embodiments, the target structured data can be generated according to the target structured data identifier information and the original structured data. If the quantity of business data in the structured data is large, although the original structured data can be screened through the target structured data identifier information, the large business data can still affect the processing of the model.

[0076] Therefore, in a specific embodiment provided in the present disclosure, the target structured data is generated according to the target structured data identifier information and the original structured data, and the method comprises: generating the target structured data according to the target structured data identifier information, the to-be-processed text information and the original structured data.

[0077] In actual application, in addition to using the target structured data identifier information to compress the original structured data, the original structured data can be further compressed according to the content of the to-be-processed text information, so as to obtain the target structured data.

[0078] In practical applications, after obtaining the target structured data identification information, the original structured data can be filtered according to the target structured data identification information, but the filtered structured data obtained at this time may still have the problem that the business data volume is too large to exceed the processing capacity of the question and answer model. At this time, the filtered structured data can be further filtered according to the to-be-processed text information, so as to obtain the target structured data.

[0079] For example, still taking the original structured data with 100 business fields and 100,000 pieces of structured data as an example for explanation and description. After filtering the original structured data according to the target structured data identification information, the 100 business fields are compressed to 10 business fields. However, the 10 business fields still correspond to 100,000 pieces of business data. At this time, further filtering can be performed according to the to-be-processed text information, for example, there are certain business data filtering conditions in the to-be-processed text information, and by filtering the business data, the number of business data can be further reduced. Thus, the target structured data is obtained.

[0080] In a specific embodiment provided in the present disclosure, generating target structured data according to the target structured data identification information, the to-be-processed text information and the original structured data includes: extracting a data filtering parameter in the to-be-processed text information; generating reference structured data according to the target structured data identification information and the original structured data; filtering the reference structured data according to the data filtering parameter to generate target structured data corresponding to the reference structured data.

[0081] The data filtering parameter is a limitation condition parameter in the to-be-processed text information, for example, the to-be-processed text information is “which is the largest and smallest of the sales growth percentage of A enterprise, B enterprise and C enterprise in a certain region during the period from 2020 to 2022”. Among them, the query time is limited to 2020-2022, and the query range is limited to A enterprise, B enterprise and C enterprise and a certain region. 2020-2022, A enterprise, B enterprise, C enterprise and a certain region can be understood as data filtering parameters.

[0082] According to the target structured data identification information and the original structured data, preliminary filtering is performed, and only the business data related to the target structured data identification information in the original structured data is retained to generate reference structured data. The reference structured data is generated by filtering and compressing only according to the target structured data identification information.

[0083] According to the data filtering parameter, reference structured data is filtered, and business data meeting the data filtering parameter is retained. Target structured data corresponding to the reference structured data is generated. The business data in the target structured data at this time is business data related to the to-be-processed text information. Through the compression filtering of the structured data identification information and the data filtering parameter, the volume of the target structured data obtained is much smaller than the volume of the original structured data. This facilitates subsequent understanding of the target structured data, and improves the processing efficiency of the question and answer model.

[0084] Through the method provided in the embodiments of the present disclosure, the structured data filtering text is generated, and the structured data filtering text is input into the question and answer model for processing. The understanding ability of the question and answer model is used to determine the target structured data identification information related to the to-be-processed text information, which reduces the processing workload of the subsequent question and answer model in understanding the structured data, and facilitates the improvement of the subsequent processing efficiency of the question and answer model.

[0085] Step 206: generating a question and answer text according to the to-be-processed text information and the target structured data, and inputting the question and answer text into the question and answer model to obtain target answer information output by the question and answer model, wherein the question and answer text includes tree structure logical processing information for the to-be-processed text information, and the question and answer model generates the target answer information based on the tree structure logical processing information.

[0086] The question and answer text can be understood as a prompt text input into the question and answer model to assist the question and answer model in question and answer. According to the content in the question and answer text, the question and answer model can process the to-be-processed text information and the target structured data, and obtain the target answer information corresponding to the to-be-processed text information from the target structured data. The target answer information can be understood as an answer generated based on the original structured data to the to-be-processed text information.

[0087] In the application process of the large language model, the prompt text plays a key role. In the question and answer text provided in the embodiments of the present disclosure, the prompt information includes the tree structure logical processing information. The tree structure logical processing information facilitates the large language model to split the target structured data into smaller and more easily understood structured sub-data, so as to realize the analysis of each structured sub-data, and obtain the final target answer information according to the processing result of each structured sub-data.

[0088] Tree-structured logic processing organizes structured data into a tree structure. This structure mimics the human cognitive process of step-by-step reasoning and problem-solving, providing a clear navigation path for large language models when processing complex relationships and large amounts of structured data. Each node in tree-structured logic processing represents a logical step in information data processing. By using tree-structured logic processing, the parsing and application of complex structured data can be made more efficient and systematic.

[0089] In one specific embodiment provided in this disclosure, generating question-and-answer text based on the text information to be processed and the target structured data includes: determining a parsing logic prompt text based on the text information to be processed; and concatenating the parsing logic prompt text, the text information to be processed, and the target structured data to generate question-and-answer text.

[0090] The question-and-answer text in this embodiment includes the aforementioned tree-structured logical processing information, that is, determining the parsing logic prompt text for the text to be processed based on the text information to be processed. The parsing logic prompt text can be understood as the tree-structured logical processing information in the process of answering the text information to be processed. It defines the order of processing logic in the process of answering the text information to be processed.

[0091] Tree-structured logic processing reorganizes information extracted from structured data into a hierarchical tree structure. Nodes in this structure represent logical information blocks or steps in a reasoning chain. It breaks down complex problems into smaller, more manageable subproblems, with the arrangement of nodes mimicking the thought process used to answer questions about textual information. By constructing a structured data tree, it effectively performs step-by-step reasoning, simulating the human process of reasoning through complex problems. The real-time generation of tree structures within a large language model ensures that each traversal allows for reasonable step-by-step reasoning based on the current node's information, demonstrating the characteristics of iterative reasoning and ultimately providing the target answer.

[0092] Specifically, determining the parsing logic prompt text based on the text information to be processed includes: parsing the text information to be processed to obtain the tree structure logic processing information corresponding to the text information to be processed; and generating the parsing logic prompt text corresponding to the text information to be processed based on the logic processing information.

[0093] In practical applications, the text information to be processed varies greatly. It's impossible to set a fixed template for the parsing logic prompt text for ease of use. Therefore, the text information to be processed can be parsed to obtain its tree-structured logical processing information. Based on this tree-structured logical processing information, the corresponding parsing logic prompt text for the text to be processed can be generated.

[0094] Furthermore, the text information to be processed can be input into a pre-trained information understanding model for processing. The information understanding model can generate corresponding tree-structured logical processing information based on the text information to be processed.

[0095] For example, let's take the text message to be processed as "During the period from 2020 to 2022, which of Company A, Company B, and Company C had the largest and smallest percentage sales growth in a certain region?" as an example for explanation. By parsing this text message, we can determine its corresponding tree structure logical processing information as follows.

[0096] 1. Calculate the maximum and minimum percentage sales growth of Company A in a certain region between 2020 and 2022.

[0097] 2. Calculate the maximum and minimum percentage sales growth of Company B in a certain region between 2020 and 2022.

[0098] 3. Calculate the maximum and minimum percentage sales growth of Company C in a certain region between 2020 and 2022.

[0099] 4. Determine the maximum and minimum percentage of sales growth among the maximum and minimum percentages of sales growth for Company A, Company B, and Company C.

[0100] Based on the tree structure logic processing information, corresponding parsing logic prompt text is generated. It should be noted that the parsing logic prompt text may also include methods for calculating the maximum and minimum percentage of sales growth.

[0101] In another specific embodiment provided in this disclosure, the question-and-answer text includes tree structure logic processing information corresponding to the text information to be processed; inputting the question-and-answer text into the question-and-answer model to obtain the target answer information output by the question-and-answer model includes: inputting the question-and-answer text into the question-and-answer model to obtain the target answer information output by the question-and-answer model, wherein the question-and-answer model splits the target structured data into at least one target structured sub-data based on the tree structure logic processing information and the text information to be processed, parses each target structured sub-data to generate reference answer information corresponding to each target structured sub-data, and generates target answer information based on the tree structure logic processing information and each reference answer information.

[0102] In practical applications, the question-and-answer text includes tree-structured logical processing information corresponding to the text information to be processed. When the question-and-answer text is input into the question-and-answer model for processing, the model, based on the tree-structured logical processing information, splits the target structured data into at least one target structured sub-data. By parsing each target structured sub-data, the model obtains the reference answer information corresponding to each target structured sub-data. Finally, the final target answer information is generated based on the reference answer information.

[0103] Let's continue using the example of the text to be processed: "During the period from 2020 to 2022, which companies, A, B, and C, had the highest and lowest percentage sales growth in a certain region?" Based on the content of nodes 1-3 in the tree structure logic, the target structured data is split into three target structured sub-data, corresponding to the sales information of company A, company B, and company C in a certain region from 2020 to 2022. After parsing and analyzing each of the three target structured sub-data, we obtain the maximum percentage sales growth (Amax) and maximum percentage sales growth (Amin) for company A in a certain region during the period from 2020 to 2022; the maximum percentage sales growth (Bmax) and maximum percentage sales growth (Bmin) for company B in a certain region during the same period; and the maximum percentage sales growth (Cmax) and maximum percentage sales growth (Cmin) for company C in a certain region during the same period. Then, based on the content of the fourth node in the tree structure logical processing information, the maximum value Cmax is selected from Amax, Bmax, and Cmax; the minimum value Amin is selected from Amin, Bmin, and Cmin, thereby determining the target answer information: "During the period from 2020 to 2022, the maximum percentage increase in sales of Company A, Company B, and Company C in a certain region is Cmax, and the minimum value is Amin."

[0104] The method provided in this disclosure generates target structured data based on the text information to be processed and the original structured data, effectively reducing the volume of the original structured data. This allows the large language model to process less data in subsequent processing of the structured data and better grasp the key information in the target structured data. During the processing of question-and-answer text, the question-and-answer model can enhance its reasoning ability for target structured data by processing information based on the tree structure logic within the text. This involves breaking down complex structured data into multiple structured sub-data and determining the final target answer information based on the reference answer information of each structured sub-data.

[0105] Secondly, in the process of generating target structured data, the target structured data identification information can be determined by the text information to be processed and the original structured data. The original structured data is then filtered using the target structured data identification information. The filtered structured data is then filtered again based on the text information to be processed to obtain the target structured data. Through these two filtering processes, the original structured data can be compressed more effectively, reducing the amount of data to be processed by the large language model.

[0106] Furthermore, the question-and-answer text generated from the text information to be processed and the target structured data includes tree-structured logical processing information. Constructing the answer logic into a tree-like thought chain helps the large language model gradually understand complex structured data content. The tree-like thought chain simulates the step-by-step reasoning process of problem-solving, providing a logical and orderly operational method. This enables the large language model to query information within complex structured data, simulating the user's segmented understanding of information and providing answers. Each node in the tree-like thought chain has a specific purpose, simplifying the interaction between the large language model and complex structured data, allowing the large language model to more effectively process and understand business data within structured data.

[0107] The following description, in conjunction with Figure 3, uses the application of the structured data-based question-answering method provided in this disclosure in a sports scenario as an example to further illustrate the structured data-based question-answering method. Figure 3 shows a flowchart of the processing procedure of a sports table-based question-answering method according to an embodiment of this disclosure, specifically including the following steps.

[0108] Step 302: Obtain the sports problem information to be processed and the original sports information table corresponding to the sports problem information to be processed.

[0109] Step 304: Process the sports problem information to be processed and the original sports information table based on the question-answering model to obtain the target sports information table.

[0110] Step 306: Generate question-and-answer text based on the sports problem information to be processed and the target sports information table, and input the question-and-answer text into the question-and-answer model to obtain the target answer information output by the question-and-answer model. The question-and-answer text includes tree structure logic processing information for the sports problem information to be processed, and the question-and-answer model generates the target answer information based on the tree structure logic processing information.

[0111] The question-answering method based on structured data provided in this implementation is applied to query scenarios involving sports tabular data. For example, a specific sports event, such as a football match, a basketball match, or a multi-sport event.

[0112] The sports-related problem information to be processed is problem information specific to sports scenarios. For example, the sports-related problem information to be processed could be "Who are the oldest and youngest players in all the finals of event A in competition 1?"

[0113] Based on the sports problem information to be processed, the original sports information table corresponding to the sports problem information is obtained. The original sports information table includes various sports-related information, such as event 1, event 2, event 3, etc. Event 1 can be a comprehensive event, event 2 can be a football-related event, event 3 can be a gymnastics-related event, and so on.

[0114] If the original sports information table needs to be filtered, then the corresponding table filtering text is generated based on the sports question information to be processed and the original sports information table. The table filtering text is used to prompt the question answering model to filter based on the original sports information table and determine the target header information related to the sports question information to be processed.

[0115] Based on the target header information, the sports problem information to be processed, and the original sports information table, further compression and filtering are performed to obtain the target sports information table. The target sports information table can be understood as a sports information table related to the sports problem information to be processed.

[0116] Based on the information of the sports problem to be processed, the parsing logic prompt text is determined. The parsing logic prompt text, the text information to be processed, and the target sports information table are concatenated to generate the question and answer text, which is then input into the question and answer model for processing.

[0117] The question-answering model splits the target sports information table into multiple sub-tables based on the parsing logic prompts and the sports question information to be processed. Each sub-table is then processed to obtain corresponding reference answer information, and the final target answer information is generated based on these reference answers.

[0118] For example, let's take the sports question "Who are the oldest and youngest athletes in all the finals of Event A in Competition 1?" as an example for explanation. Based on the parsing logic prompts and the sports question information, the target sports information table can be split according to the finals of Event A in Competition 1, resulting in sub-tables for each final. In each sub-table, the age of each athlete is calculated based on their birth date and the date of the final, determining the oldest and youngest athletes in each final. Comparing the oldest and youngest athletes in each final further determines the oldest and youngest athletes in all the finals of Event A in Competition 1. Therefore, the target answer is "The oldest athlete is Zhang San, and the youngest athlete is Li Si." Furthermore, the target answer information can be further used to determine which final the oldest athlete, Zhang San, participated in, and related information about Li Si's age (such as date of birth, date of the final, etc.); and which final the youngest athlete, Li Si, participated in, and related information about Li Si's age (such as date of birth, date of the final, etc.).

[0119] The method provided in this disclosure generates a table filtering text based on the sports question information to be processed and the original sports information table. The target header information generated from the table filtering text is then used to filter the original sports information table, generating a target sports information table. This effectively reduces the size of the original sports information table, allowing the large language model to process less data in subsequent table processing and better grasp the key information in the target sports information table. During the question-and-answer model's processing of the question-and-answer text, processing information based on the tree structure logic within the text enhances the model's reasoning ability for target structured data. Complex structured data is broken down into multiple structured sub-data, and the final target answer information is determined based on the reference answer information of each structured sub-data.

[0120] In the process of generating the target sports information table, the target header information can be determined by the sports problem information to be processed and the original sports information table. The original sports information table is then filtered using the target header information. The filtered sports information table is then filtered a second time based on the sports problem information to be processed to obtain the target sports information table. Through these two filtering processes, the original sports information table can be compressed more effectively, reducing the amount of data processed by the large language model.

[0121] Furthermore, the question-and-answer text generated from the sports problem information and the target sports information table includes parsing logic hints, constructing the answer logic as a tree-like thought chain. This helps the large language model gradually understand the complex table content. The tree-like thought chain simulates the step-by-step reasoning process of problem-solving, providing a logical and orderly operational method. This allows the large language model to query information within complex tables, simulating the user's segmented understanding of information and providing solutions. Each node in the tree-like thought chain has a specific purpose, simplifying the interaction between the large language model and complex tables, enabling the large language model to more effectively process and understand the business data within the tables.

[0122] Referring to Figure 4, Figure 4 shows a flowchart of a sports table-based question-and-answer method for cloud-side devices provided in an embodiment of this disclosure, which specifically includes the following steps.

[0123] Step 402: Receive the sports problem information to be processed sent by the receiving end device, and obtain the corresponding original sports information table based on the sports problem information to be processed.

[0124] Step 404: Process the sports problem information to be processed and the original sports information table based on the question-answering model to obtain the target sports information table.

[0125] Step 406: Generate question-and-answer text based on the sports problem information to be processed and the target sports information table, and input the question-and-answer text into the question-and-answer model to obtain the target answer information output by the question-and-answer model. The question-and-answer text includes tree structure logic processing information for the sports problem information to be processed, and the question-and-answer model generates the target answer information based on the tree structure logic processing information.

[0126] Step 408: Send the target answer information to the end-side device.

[0127] In practical applications, question-answering models are large language models, and their deployment requires substantial computing resources. Edge devices may lack the necessary processing capabilities. Therefore, question-answering methods based on sports tables can be implemented on cloud devices. The cloud device receives the sports questions to be processed from the edge device and retrieves the corresponding raw sports table information based on the questions. It then further obtains the target answer information corresponding to the sports question information and finally returns the target answer information to the edge device.

[0128] It should be noted that the implementation methods of steps 402 to 406 are the same as those of steps 202 to 206 described above, and will not be repeated in this embodiment.

[0129] The method provided in this embodiment further includes: receiving adjustment information sent by the end-side device for the target answer information; inputting the adjustment information and the target answer information into the question-answering model to obtain the adjusted answer information output by the question-answering model; and sending the adjusted answer information to the end-side device.

[0130] After receiving the target answer information, the client can further send adjustment information to modify the content of the target answer information. For example, if the target answer information includes entity 1 and entity 2, the adjustment information could be "Please provide relevant information about entity 1".

[0131] The cloud-side device inputs the adjustment information and target answer information into the question-answering model for further processing, obtains the adjustment answer information output by the question-answering model, and returns the adjustment answer information to the edge device.

[0132] The method provided in this disclosure generates a table filtering text based on the sports problem information to be processed and the original sports information table. The target header information generated by the table filtering text is used to filter the original sports information table to generate a target sports information table. This effectively reduces the size of the original sports information table, so that the large language model can better grasp the key information in the target sports information table during the subsequent processing of the table, which reduces the amount of data processed by the large language model.

[0133] Furthermore, the question-and-answer text generated from the sports problem information and the target sports information table includes parsing logic hints, constructing the answer logic as a tree-like thought chain. This helps the large language model gradually understand the complex table content. The tree-like thought chain simulates the step-by-step reasoning process of problem-solving, providing a logical and orderly operational method. This allows the large language model to query information within complex tables, simulating the user's segmented understanding of information and providing solutions. Each node in the tree-like thought chain has a specific purpose, simplifying the interaction between the large language model and complex tables, enabling the large language model to more effectively process and understand the business data within the tables.

[0134] Figure 5 shows a flowchart of an information processing method based on a question-answering model provided in an embodiment of the present disclosure. The method is applied to a task platform and includes the following steps.

[0135] Step 502: Receive a task generation request sent by the receiving end device, wherein the task generation request includes request information.

[0136] Step 504: Based on the request information, obtain a question-answering model, wherein the question-answering model determines target structured data based on the text information to be processed and the original structured data, and determines target answer information based on the question-answering text generated based on the text information to be processed and the target structured data, wherein the question-answering text includes tree structure logic processing information for the text information to be processed, and the question-answering model generates the target answer information based on the tree structure logic processing information.

[0137] Step 506: Generate task information based on the question-answering model, wherein the task information is used by the end device to perform a question-answering task based on structured data.

[0138] A task platform is an online platform that provides services or functions, allowing developers, enterprises, or individual users to submit tasks, obtain resources, or implement specific functions through API calls, web interfaces, or other interactive methods. In this embodiment, the task platform is a cloud service platform capable of providing question-and-answer services based on structured data. It allows external applications (such as content recommendation, social media, digital twin systems, etc.) to access and use its question-and-answer processing capabilities based on structured data to generate target answer information. The task platform ensures efficient and accurate responses to requests from edge devices by managing model training, updates, optimization, and load balancing.

[0139] A task generation request is a specific operation request initiated by an edge device to the task platform, designed to trigger the platform to execute a question-and-answer task based on structured data. Task generation requests typically contain necessary parameters and request information, such as the text to be processed.

[0140] The task platform can obtain the corresponding question-answering model based on the request information. For relevant information about the question-answering model, please refer to the relevant description in the above embodiments, which will not be repeated here.

[0141] After the task platform processes the task generation request, it generates task information based on the output of the scene localization model, serving as a response to the edge device. This task information guides the edge device to operate the question-answering model to execute corresponding question-answering tasks based on structured data.

[0142] In another specific embodiment provided in this disclosure, the request information includes a task scenario identifier for a question-and-answer task based on structured data, or a task model identifier; obtaining a question-and-answer model based on the request information includes: determining a target scenario template from multiple preset scenario templates based on the task scenario identifier, and searching for a question-and-answer model from a model library based on the target scenario template, wherein the model library stores multiple question-and-answer models; or, searching for a question-and-answer model from the model library based on the task model identifier.

[0143] The task scenario identifier for question-answering tasks based on structured data is a unique label or code that identifies or classifies a question-answering task based on structured data. It helps the task platform understand which specific task category a question-answering task based on structured data belongs to.

[0144] The task scenario identifier for question answering tasks based on structured data is a label used to distinguish and identify different question answering task models based on structured data. Each model may be designed for different data features, processing logic, or optimization goals.

[0145] In practical applications, the target scenario template corresponding to the task scenario identifier can be found from multiple preset scenario templates based on the task scenario identifier, and the corresponding question-answering model can be found from the model library based on the target scenario template. Alternatively, the corresponding question-answering model can be found directly from the model library based on the task model identifier.

[0146] Figure 6 shows a schematic diagram of a task platform provided in an embodiment of this disclosure. The task platform includes a request interface 602 and a response unit 604, wherein: the request interface 602 is used to receive a task generation request sent by an end-side device, wherein the task generation request includes request information. The response unit 604 is used to obtain a question-and-answer model based on the request information, wherein the question-and-answer model determines target structured data based on the text information to be processed and the original structured data, and determines target answer information based on question-and-answer text generated based on the text information to be processed and the target structured data, wherein the question-and-answer text includes tree structure logic processing information for the text information to be processed, and the question-and-answer model generates the target answer information based on the tree structure logic processing information.

[0147] Corresponding to the above method embodiments, this disclosure also provides an embodiment of a question-answering device based on structured data. Figure 7 shows a schematic diagram of the structure of a question-answering device based on structured data provided in one embodiment of this disclosure. As shown in Figure 7, the device includes: an acquisition module 702, configured to acquire text information to be processed and the original structured data corresponding to the text information to be processed; a filtering module 704, configured to process the text information to be processed and the original structured data based on a question-answering model to obtain target structured data; and a question-answering module 706, configured to generate question-answer text based on the text information to be processed and the target structured data, and input the question-answer text into the question-answering model to obtain target answer information output by the question-answering model. The question-answer text includes tree structure logic processing information for the text information to be processed, and the question-answering model generates the target answer information based on the tree structure logic processing information.

[0148] Optionally, the acquisition module 702 is further configured to: receive text information to be processed; parse the text information to be processed to obtain text domain information corresponding to the text information to be processed; and determine the original structured data corresponding to the text information to be processed based on the text domain information.

[0149] Optionally, the filtering module 704 is further configured to: generate structured data filtering text based on the text information to be processed and the original structured data, and input the structured data filtering text into the question-answering model to obtain the target structured data identification information output by the question-answering model; and generate target structured data based on the target structured data identification information and the original structured data.

[0150] Optionally, the filtering module 704 is further configured to: determine a preset data filtering prompt text; generate a filtering example text based on the original structured data, wherein the filtering example text includes original structured data identification information and example structured data corresponding to the original structured data identification information; and generate structured data filtering text based on the preset data filtering prompt text, the filtering example text, and the text information to be processed.

[0151] Optionally, the filtering module 704 is further configured to: extract original structured data identification information from the original structured data; extract example structured data from the original structured data based on a preset data quantity and the original structured data identification information; and generate filtering example text based on the original structured data identification information and the example structured data.

[0152] Optionally, the filtering module 704 is further configured to generate target structured data based on the target structured data identifier information, the text information to be processed, and the original structured data.

[0153] Optionally, the filtering module 704 is further configured to: extract data filtering parameters from the text information to be processed; generate reference structured data based on the target structured data identifier information and the original structured data; filter the reference structured data based on the data filtering parameters to generate target structured data corresponding to the reference structured data.

[0154] Optionally, the question-and-answer module 706 is further configured to: determine the parsing logic prompt text based on the text information to be processed; and concatenate the parsing logic prompt text, the text information to be processed, and the target structured data to generate question-and-answer text.

[0155] Optionally, the question-answering module 706 is further configured to: parse the text information to be processed to obtain tree structure logic processing information corresponding to the text information to be processed; and generate parsing logic prompt text corresponding to the text information to be processed based on the logic processing information.

[0156] Optionally, the question-answering module 706 is further configured to: parse the text information to be processed, obtain at least one logical processing step corresponding to the text information to be processed and the execution order between the logical processing steps; and determine tree structure logical processing information based on the information of each logical processing step and the execution order.

[0157] Optionally, the question-and-answer text includes tree structure logic processing information corresponding to the text information to be processed; the question-and-answer module 706 is further configured to: input the question-and-answer text into the question-and-answer model to obtain the target answer information output by the question-and-answer model, wherein the question-and-answer model splits the target structured data into at least one target structured sub-data based on the tree structure logic processing information and the text information to be processed, parses each target structured sub-data to generate reference answer information corresponding to each target structured sub-data, and generates target answer information based on the tree structure logic processing information and each reference answer information.

[0158] The apparatus provided in this disclosure generates target structured data based on the text information to be processed and the original structured data, effectively reducing the volume of the original structured data. This allows the large language model to process less data in subsequent processing of the structured data and better grasp the key information in the target structured data. During the processing of question-and-answer text, the question-and-answer model enhances its reasoning ability for target structured data by processing information based on the tree structure logic within the text. It breaks down complex structured data into multiple structured sub-data, and determines the final target answer information based on the reference answer information of each structured sub-data.

[0159] Secondly, in the process of generating target structured data, the target structured data identification information can be determined by the text information to be processed and the original structured data. The original structured data is then filtered using the target structured data identification information. The filtered structured data is then filtered again based on the text information to be processed to obtain the target structured data. Through these two filtering processes, the original structured data can be compressed more effectively, reducing the amount of data to be processed by the large language model.

[0160] Furthermore, the question-and-answer text generated from the text information to be processed and the target structured data includes tree-structured logical processing information. Constructing the answer logic into a tree-like thought chain helps the large language model gradually understand complex structured data content. The tree-like thought chain simulates the step-by-step reasoning process of problem-solving, providing a logical and orderly operational method. This enables the large language model to query information within complex structured data, simulating the user's segmented understanding of information and providing answers. Each node in the tree-like thought chain has a specific purpose, simplifying the interaction between the large language model and complex structured data, allowing the large language model to more effectively process and understand business data within structured data.

[0161] The above is an illustrative scheme of a question-and-answer device based on structured data according to this embodiment. It should be noted that the technical solution of this question-and-answer device based on structured data and the technical solution of the question-and-answer method based on structured data described above belong to the same concept. For details not described in detail in the technical solution of the question-and-answer device based on structured data, please refer to the description of the technical solution of the question-and-answer method based on structured data described above.

[0162] Referring to Figure 8, which illustrates an architecture diagram of a question-answering system based on structured data according to an embodiment of this disclosure, the question-answering system based on structured data may include a client 100 and a server 200; the client 100 is used to send text information to be processed to the server 200; the server 200 is used to obtain the original structured data corresponding to the text information to be processed; process the text information to be processed and the original structured data based on a question-answering model to obtain target structured data; generate question-answer text based on the text information to be processed and the target structured data; input the question-answer text into the question-answering model to obtain target answer information output by the question-answering model, wherein the question-answer text includes tree structure logic processing information for the text information to be processed, and the question-answering model generates the target answer information based on the tree structure logic processing information; send the target answer information to the client 100; the client 100 is also used to receive the target answer information sent by the server 200.

[0163] A question-answering system based on structured data may include multiple clients 100 and a server 200. Clients 100 can be referred to as edge devices, and server 200 can be referred to as cloud devices. Multiple clients 100 can establish communication connections through server 200. In a question-answering scenario based on structured data, server 200 is used to provide structured data-based question-answering services between multiple clients 100. Each client 100 can act as either a sender or a receiver, communicating through server 200.

[0164] Users can interact with server 200 through client 100 to receive data sent by other clients 100, or send data to other clients 100, etc. In a question-and-answer scenario based on structured data, users can publish data streams to server 200 through client 100, server 200 can generate target answer information based on the data stream, and push the target answer information to other clients that have established communication.

[0165] In this system, client 100 and server 200 establish a connection via a network. The network provides the medium for communication between client 100 and server 200. The network can include various connection types, such as wired or wireless communication links or fiber optic cables. Data transmitted by client 100 may need to undergo encoding, transcoding, compression, or other processing before being published to server 200.

[0166] Client 100 can be a browser, an app (application), a web application such as an H5 (HyperText Markup Language 5) application, a lightweight application (also known as a mini-program), or a cloud application. Client 100 can be developed based on the software development kit (SDK) of the corresponding service provided by server 200, such as a real-time communication (RTC) SDK. Client 100 can be deployed on electronic devices and depends on the device or certain apps on the device to run. Electronic devices may have displays and support information browsing, such as personal mobile terminals like mobile phones, tablets, and personal computers. Various other types of applications can also be configured on electronic devices, such as human-computer interaction applications, model training applications, text processing applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, and social media platform software.

[0167] Server 200 may include servers providing various services, such as servers providing communication services to multiple clients, servers supporting backend training of models used on clients, and servers processing data sent by clients. It should be noted that server 200 can be implemented as a distributed server cluster composed of multiple servers, or as a single server. The server can also be a server in a distributed system, or a server integrated with blockchain. The server can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology.

[0168] It is worth noting that the question-answering method based on structured data provided in this disclosure is generally executed by the server. However, in other embodiments of this disclosure, the client may also have similar functionality to the server, thereby executing the question-answering method based on structured data provided in this disclosure. In other embodiments, the question-answering method based on structured data provided in this disclosure may also be executed jointly by the client and the server.

[0169] Figure 9 shows a structural block diagram of a computing device 900 according to an embodiment of the present disclosure. The components of the computing device 900 include, but are not limited to, a memory 910 and a processor 920. The processor 920 is connected to the memory 910 via a bus 930, and a database 950 is used to store data.

[0170] The computing device 900 also includes an access device 940, which enables the computing device 900 to communicate via one or more networks 960. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 940 may include one or more of any type of wired or wireless network interface (e.g., a network interface controller (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0171] In one embodiment of this disclosure, the aforementioned components of the computing device 900, as well as other components not shown in FIG. 9, may also be connected to each other, for example, via a bus. It should be understood that the computing device structural block diagram shown in FIG. 9 is merely for illustrative purposes and is not intended to limit the scope of this disclosure. Those skilled in the art can add or replace other components as needed.

[0172] The computing device 900 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 900 can also be a mobile or stationary server.

[0173] The processor 920 is used to execute the following computer program / instructions, which, when executed by the processor, implement the steps of the above-mentioned question-and-answer method based on structured data, question-and-answer method based on sports tables, and information processing method based on question-and-answer models.

[0174] The various embodiments in this disclosure are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the computing device embodiments are basically similar to the question-answering methods based on structured data, question-answering methods based on sports tables, and information processing methods based on question-answering models, so the description is relatively simple. Relevant parts can be referred to the descriptions of the question-answering methods based on structured data, question-answering methods based on sports tables, and information processing methods based on question-answering models.

[0175] An embodiment of this disclosure also provides a computer-readable storage medium storing a computer program / instructions that, when executed by a processor, implement the steps of the above-described question-and-answer method based on structured data, the question-and-answer method based on sports tables, and the information processing method based on question-and-answer models.

[0176] The various embodiments in this disclosure are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the computer-readable storage medium embodiments are relatively simple in description because they are fundamentally similar to the question-answering methods based on structured data, the question-answering methods based on sports tables, and the information processing methods based on question-answering models. Relevant details can be found in the descriptions of the question-answering methods based on structured data, the question-answering methods based on sports tables, and the information processing methods based on question-answering models.

[0177] An embodiment of this disclosure also provides a computer program product, including a computer program / instruction that, when executed by a processor, implements the steps of the above-described question-answering method based on structured data, question-answering method based on sports tables, and information processing method based on question-answering models.

[0178] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product belongs to the same concept as the above-mentioned question-answering method based on structured data, question-answering method based on sports tables, and information processing method based on question-answering models. For details not described in detail in the technical solution of the computer program product, please refer to the descriptions of the above-mentioned question-answering method based on structured data, question-answering method based on sports tables, and information processing method based on question-answering models.

[0179] The foregoing has described specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0180] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0181] It should be noted that the above description describes specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments of this disclosure.

[0182] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0183] The preferred embodiments disclosed above are merely illustrative of this disclosure. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments of this disclosure. These embodiments are selected and specifically described in this disclosure to better explain the principles and practical applications of the embodiments of this disclosure, thereby enabling those skilled in the art to better understand and utilize this disclosure. This disclosure is limited only by the claims and their full scope and equivalents.

Claims

1. A method for question answering based on structured data, comprising: obtaining text information to be processed and original structured data corresponding to the text information to be processed; processing the text information to be processed and the original structured data based on a question answering model to obtain target structured data; generating question answering text according to the text information to be processed and the target structured data, and inputting the question answering text into the question answering model to obtain target answer information output by the question answering model, wherein the question answering text includes tree structure logical processing information for the text information to be processed, and the question answering model generates the target answer information based on the tree structure logical processing information.

2. The method of claim 1, wherein obtaining text information to be processed and original structured data corresponding to the text information to be processed comprises: receiving text information to be processed; parsing the text information to be processed to obtain text field information corresponding to the text information to be processed; determining original structured data corresponding to the text information to be processed according to the text field information.

3. The method of claim 1, wherein processing the text information to be processed and the original structured data based on a question answering model to obtain target structured data comprises: generating structured data screening text according to the text information to be processed and the original structured data, and inputting the structured data screening text into a question answering model to obtain target structured data identification information output by the question answering model; generating target structured data according to the target structured data identification information and the original structured data.

4. The method of claim 3, wherein generating structured data screening text according to the text information to be processed and the original structured data comprises: determining a preset data screening prompt text; generating a screening example text according to the original structured data, wherein the screening example text includes original structured data identification information and example structured data corresponding to the original structured data identification information; generating structured data screening text according to the preset data screening prompt text, the screening example text, and the text information to be processed.

5. The method of claim 4, wherein generating a screening example text according to the original structured data comprises: extracting original structured data identification information in the original structured data; extracting example structured data in the original structured data based on a preset data quantity and the original structured data identification information; generating a screening example text according to the original structured data identification information and the example structured data.

6. The method of claim 3, wherein generating target structured data according to the target structured data identification information and the original structured data comprises: generating target structured data according to the target structured data identification information, the text information to be processed, and the original structured data.

7. The method of claim 6, wherein generating target structured data according to the target structured data identification information, the text information to be processed, and the original structured data comprises: extracting a data screening parameter in the to-be-processed text information; generating reference structured data according to the target structured data identification information and the original structured data; screening the reference structured data according to the data screening parameter to generate target structured data corresponding to the reference structured data.

8. The method of claim 1, wherein generating a question and answer text according to the to-be-processed text information and the target structured data comprises: determining a parsing logic prompt text according to the to-be-processed text information; splicing the parsing logic prompt text, the to-be-processed text information and the target structured data to generate a question and answer text.

9. The method of claim 8, wherein determining a parsing logic prompt text according to the to-be-processed text information comprises: parsing the to-be-processed text information to obtain tree structure logic processing information corresponding to the to-be-processed text information; generating a parsing logic prompt text corresponding to the to-be-processed text information according to the logic processing information.

10. The method of claim 9, wherein parsing the to-be-processed text information to obtain tree structure logic processing information corresponding to the to-be-processed text information comprises: parsing the to-be-processed text information to obtain at least one logic processing step and an execution order between the logic processing steps corresponding to the to-be-processed text information; determining tree structure logic processing information according to the logic processing step information and the execution order.

11. The method of claim 1, wherein the question and answer text comprises tree structure logic processing information corresponding to the to-be-processed text information; inputting the question and answer text into the question and answer model to obtain target answer information output by the question and answer model comprises: inputting the question and answer text into the question and answer model to obtain target answer information output by the question and answer model, wherein the question and answer model splits the target structured data into at least one target structured sub-data based on the tree structure logic processing information and the to-be-processed text information, parses each target structured sub-data to generate reference answer information corresponding to each target structured sub-data, and generates target answer information according to the tree structure logic processing information and each reference answer information.

12. A question and answer method based on sports tables, comprising: obtaining to-be-processed sports question information and original sports information tables corresponding to the to-be-processed sports question information; processing the to-be-processed sports question information and the original sports information tables based on a question and answer model to obtain target sports information tables; generating a question and answer text according to the to-be-processed sports question information and the target sports information tables, and inputting the question and answer text into the question and answer model to obtain target answer information output by the question and answer model, wherein the question and answer text comprises tree structure logic processing information for the to-be-processed sports question information, and the question and answer model generates target answer information based on the tree structure logic processing information.

13. A question and answer method based on sports tables, applied to a cloud side device, comprising: receiving to-be-processed sports question information sent by an end side device, and obtaining corresponding original sports information tables according to the to-be-processed sports question information; obtaining a target sports information table based on the to-be-processed sports question information and the original sports information table; generating a question and answer text according to the to-be-processed sports question information and the target sports information table, inputting the question and answer text into the question and answer model, and obtaining target answer information output by the question and answer model, wherein the question and answer text includes tree structure logical processing information for the to-be-processed sports question information, and the question and answer model generates target answer information based on the tree structure logical processing information; sending the target answer information to the terminal device.

14. The method of claim 13, further comprising: receiving adjustment information sent by the terminal device for the target answer information; inputting the adjustment information and the target answer information into the question and answer model to obtain adjustment answer information output by the question and answer model; sending the adjustment answer information to the terminal device.

15. An information processing method based on a question and answer model, applied to a task platform, comprising: receiving a task generation request sent by a terminal device, wherein the task generation request includes request information; obtaining a question and answer model based on the request information, wherein the question and answer model determines target structured data according to to-be-processed text information and original structured data, and determines target answer information according to a question and answer text generated based on the to-be-processed text information and the target structured data, the question and answer text includes tree structure logical processing information for the to-be-processed text information, and the question and answer model generates the target answer information based on the tree structure logical processing information; generating task information based on the question and answer model, wherein the task information is used for the terminal device to perform a question and answer task based on structured data.

16. The method of claim 15, wherein the request information includes a task scene identifier of the question and answer task based on structured data, or a task model identifier. Obtaining a question and answer model based on the request information includes: determining a target scene template from a plurality of preset scene templates based on the task scene identifier, and searching for a question and answer model from a model library based on the target scene template, wherein the model library stores a plurality of question and answer models; or, searching for a question and answer model from the model library based on the task model identifier.

17. A task platform, comprising a request interface and a response unit; the request interface is configured to receive a task generation request sent by a terminal device, wherein the task generation request includes request information; the response unit is configured to obtain a question and answer model based on the request information, wherein the question and answer model determines target structured data according to to-be-processed text information and original structured data, and determines target answer information according to a question and answer text generated based on the to-be-processed text information and the target structured data, the question and answer text includes tree structure logical processing information for the to-be-processed text information, and the question and answer model generates the target answer information based on the tree structure logical processing information.

18. A question and answer device based on structured data, comprising: An acquisition module configured to acquire text information to be processed and original structured data corresponding to the text information to be processed; A screening module configured to obtain target structured data by processing the text information to be processed and the original structured data based on a question and answer model; A question and answer module configured to generate question and answer text according to the text information to be processed and the target structured data, input the question and answer text into the question and answer model, and obtain target answer information output by the question and answer model, wherein the question and answer text includes tree structure logical processing information for the text information to be processed, and the question and answer model generates the target answer information based on the tree structure logical processing information.

19. A computing device comprising: a memory and a processor; the memory is configured to store computer programs / instructions, and the processor is configured to execute the computer programs / instructions, and the computer programs / instructions, when executed by the processor, implement the steps of the method of any one of claims 1 to 16.

20. A computer readable storage medium storing computer programs / instructions, and the computer programs / instructions, when executed by a processor, implement the steps of the method of any one of claims 1 to 16.

21. A computer program product comprising computer programs / instructions, and the computer programs / instructions, when executed by a processor, implement the steps of the method of any one of claims 1 to 16.

Images