Question and answer knowledge base, establishment method thereof, intelligent question and answer method and system

Abstract

The invention relates to a question and answer knowledge base and an establishment method thereof, and an intelligent question and answer method and system, and the system comprises a source data collection unit which is used for collecting data in a computer system and marking the data as source data; and the feature extraction and classification unit is used for extracting feature data such as a text semantic structure, a keyword type and an information format of the source data to classify the source data, obtaining and identifying distinguished data QB, and analyzing associated data in the distinguished data QB to generate preprocessed data Yc. In particular to the technical field of question and answer knowledge bases, through a source data processing module, feature data such as text semantic structures, keyword types and information formats of source data can be accurately extracted, so that effective classification of the source data is realized, and distinguishing data QB is obtained and identified. Therefore, data related to questions and answers can be quickly positioned and screened, the efficiency and accuracy of data processing are improved, and a solid foundation is laid for subsequent question and answer knowledge construction.

Description

Technical Field

[0001] The present invention belongs to the technical field of question-answering knowledge bases, and in particular, relates to a question-answering knowledge base and a method for establishing the same, and an intelligent question-answering method and system. Background Art

[0002] With the rapid development of information technology, a huge amount of data resources have been accumulated in computer systems. In various application scenarios, such as customer service, intelligent assistants, knowledge retrieval, etc., how to efficiently use this data to accurately answer users' questions has become a key challenge. Traditional question-answering systems often face problems such as low data processing efficiency, low accuracy of question-answering, and lack of effective data organization and storage mechanisms. For example, a simple question-answering system based on keyword matching has difficulty understanding the semantic connotation of user questions, and is prone to misjudgment or failure to provide accurate answers. Moreover, in the face of large-scale data, without a reasonable classification, integration and storage strategy, data query and call delays will occur, affecting the overall performance of the system. In addition, the lack of deep mining and utilization of data associations makes it difficult for the question-answering system to provide comprehensive and in-depth answers and cannot meet the increasingly complex needs of users. Therefore, it is particularly necessary to develop an efficient, intelligent question-answering knowledge base with a complete data processing process. Summary of the invention

[0003] The purpose of the present invention is to provide a question-answering knowledge base and a method for establishing the same, an intelligent question-answering method and system, which solve the problems existing in the prior art.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] A question-answering knowledge base, comprising:

[0006] A source data processing module is configured to collect data in a computer system and mark it as source data, extract feature data such as text semantic structure, keyword type and information format of the source data to classify the source data, obtain distinguishing data QB and mark it, and analyze the associated data in the distinguishing data QB to generate preprocessing data Yc, wherein the associated data is text data with the same semantic theme or related keywords;

[0007] The data organization module integrates the pre-processed data Yc obtained from different classifications into a pre-processed data bundle containing multiple sets of question-answering knowledge-related databases, wherein the database is composed of data chains of various lengths covering question texts, answer texts, and knowledge-related labels. The module also performs group processing according to the type of data chain, and transmits the same type of data chains to the storage unit with the help of the data identification channel. The storage unit is pre-set with question classification templates and answer matching templates, etc., which are used to match, sort, re-match and store the data chains for the second time, and calculate the matched data chains in sequence, starting from the data chain with the smallest capacity, and storing the results in the result template. If the subsequent calculation encounters the same type of question-answering knowledge data group, the calculation is skipped and the result template data is read. At the same time, the matching template is generated and stored in the memory storage unit for calling;

[0008] The data storage and display module includes a storage module and a data display terminal. The storage module is used to store and record analysis and calculation result template data and match and verify it with the result template. The data display terminal is used to display the question and answer data results processed by the storage module and the invalid question and answer data transmitted by the invalid data module. The invalid data module is used to process the input invalid question and answer data.

[0009] The present invention also discloses a method for establishing a question-answer knowledge base, which specifically comprises the following steps:

[0010] Collect the data information obtained by the computer and mark it as source data, extract the feature data of the text semantic structure, keyword type and information format of the source data, classify the source data according to the feature data to obtain the distinguishing data QB and mark it;

[0011] Analyze and process the associated data in the distinguishing data QB, i.e., the text data of the same semantic topic or related keywords, to generate preprocessed data Yc;

[0012] The preprocessed data Yc obtained by different classifications are bundled to form a preprocessed data bundle package containing multiple sets of question-answer knowledge related databases. The database consists of data links of various lengths. The data link types include question text, answer text and knowledge related tags. The data links are grouped according to the data link type and the data links of the same type are transmitted to the storage unit with the help of a data identification channel.

[0013] The storage unit matches the data chain using a pre-set template, sorts and stores the data chain from large to small according to factors such as the number of question-answer knowledge or text length contained in the data chain, and then performs more accurate matching on the sorted data chain based on keyword range, semantic depth, etc. and feeds it back to the storage unit for secondary storage;

[0014] The matched data chains are calculated sequentially starting from the one with the smallest capacity. The calculation method includes the integration of question and answer knowledge and association analysis. The calculation results are stored in the result template. If the subsequent calculation encounters the same type of question and answer knowledge data group, the calculation is skipped and the result template data is read. At the same time, a matching template is generated and stored in the memory storage unit.

[0015] The present invention also discloses an intelligent question-answering method for a question-answering knowledge base, which specifically comprises the following steps:

[0016] Data acquisition operation: the computer acquires data information related to the question and answer;

[0017] Data processing operation: data is transmitted to the data processing center. The data grouping unit of the data processing center groups the data according to the data link type. The data sorting unit sorts the grouped data links according to the capacity. The data analysis and calculation unit calculates the sorted and matched data links starting from the smallest capacity. The calculation method includes the integration and association analysis of question and answer knowledge. If the same type of question and answer knowledge data group is encountered, the calculation is skipped and the result template data is read. The memory storage unit stores the matching template for easy calling.

[0018] Result display operation: The data results are presented to the operator through the data display terminal. The storage module stores and records the data results and matches and verifies them with the result template. The invalid data module processes invalid question and answer data and transmits it to the data display terminal for the operator to observe.

[0019] The present invention also discloses a system for establishing a question-answer knowledge base, comprising:

[0020] A source data acquisition unit, used for acquiring data in a computer system and marking it as source data;

[0021] The feature extraction and classification unit extracts the feature data such as the text semantic structure, keyword type and information format of the source data to classify the source data, obtain the distinguishing data QB and mark it, and can analyze the associated data in the distinguishing data QB to generate pre-processed data Yc;

[0022] A data integration and transmission unit integrates the pre-processed data Yc obtained from different classifications into a pre-processed data bundle containing multiple sets of question-answer knowledge related databases;

[0023] The data chain processing unit and the storage unit are pre-set with question classification templates and answer matching templates. The unit is used to match, sort, re-match and store the data chains, and calculate the matched data chains in sequence, starting from the data chain with the smallest capacity, and storing the results in the result template. If the subsequent calculation encounters the same type of question and answer knowledge data group, the calculation is skipped and the result template data is read. At the same time, a matching template is generated and stored in the memory storage unit;

[0024] The storage and display unit includes a storage module and a data display terminal. The storage module is used to store and record analysis and calculation result template data and match and verify it with the result template. The data display terminal is used to display the question and answer data results processed by the storage module and the invalid question and answer data transmitted by the invalid data module. The invalid data module is used to process the input invalid question and answer data.

[0025] The beneficial effects of the present invention are as follows: through the source data processing module, it is possible to accurately extract the text semantic structure, keyword type, information format and other characteristic data of the source data, thereby realizing effective classification of the source data, obtaining the distinguishing data QB and marking it. This helps to quickly locate and screen the data related to the question and answer, improve the efficiency and accuracy of data processing, and lay a solid foundation for the subsequent question and answer knowledge construction; in-depth analysis of the associated data in the distinguishing data QB to generate preprocessed data Yc can dig out the intrinsic connection between text data with the same semantic theme or related keywords. This kind of association analysis enables the question and answer knowledge base to provide more comprehensive and logical answers, avoids simple fragmented information replies, and improves the quality and depth of users' knowledge acquisition.

[0026] The data organization module integrates the preprocessed data Yc obtained from different classifications into a preprocessed data bundle containing multiple sets of question-answer knowledge-related databases, and performs group processing and intelligent transmission according to the data link type. With the help of pre-set question classification templates and answer matching templates, efficient matching, sorting, re-matching and secondary storage of data links are achieved. At the same time, through a unique calculation and result storage mechanism, such as starting calculation from the data link with the smallest capacity and using result templates and memory storage units, repeated calculations are greatly reduced, storage and retrieval efficiency is improved, and system resource consumption is reduced.

[0027] In the intelligent question-and-answer method, the various units of the data processing center work together to carefully group, sort, and analyze the question-and-answer data, ensuring that answers can be provided quickly and accurately when faced with questions of different types and complexities. The cooperation of the data display terminal and the storage module can not only clearly display the question-and-answer data results, but also effectively store, record and verify the results to ensure the reliability and traceability of the data. The setting of the invalid data module further improves the stability of the system and the user experience, and can promptly process invalid question-and-answer data to avoid interference with the normal question-and-answer process. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0029] Figure 1 The present invention is a flowchart of a question-and-answer knowledge base and a method for establishing the same, and an intelligent question-and-answer method and system. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] See also Figure 1 As shown, the present invention focuses on the construction of a question-and-answer knowledge base and an intelligent question-and-answer process, and specifically includes the following embodiments.

[0032] Example 1

[0033] The source data acquisition unit obtains data from the computer system by connecting to various data interfaces of the computer system, such as database connection interface, file reading interface, network data receiving interface, etc. These data come from a wide range of sources, such as system logs, user operation records, document materials, web page information, etc. After the data is collected, it is immediately given a unique identification tag as source data for subsequent processing. For example: when collecting data from the internal computer system of an enterprise, business operation record data, such as order processing data, customer information data, etc., are obtained from the office software database used by employees, and similar identifications such as "source data-001" and "source data-002" are added to each record.

[0034] The feature extraction and classification unit conducts in-depth analysis on the marked source data. First, the text semantic structure of the source data is extracted using natural language processing technology and data format parsing algorithms. For example, for a text describing the instructions for use of a product, its sentence structure, subject-predicate-object relationship, etc. are analyzed to understand the core semantics of the text. Next, the keyword type is identified to distinguish different types of keywords such as professional terms, general vocabulary, and functional description vocabulary. Based on the extracted feature data, the source data is classified into different categories using pre-set classification rules and machine learning classification models (such as classification models trained based on decision trees, neural networks, etc.), and the distinguishing data QB is obtained. A classification identifier is added to each QB data, such as "product knowledge category QB-01" and "operation process category QB-02". For the associated data in the distinguishing data QB, a semantic similarity calculation algorithm (such as a cosine similarity algorithm, etc.) is used to find text data with the same semantic theme or related keywords. For example, in the product knowledge category QB, all different description data about a certain product model are associated. Then these related data are integrated and preprocessed to generate preprocessed data Yc, such as organizing the related data into a unified data structure, extracting key information and performing standardized processing.

[0035] Example 2

[0036] The data integration and transmission unit integrates the preprocessed data Yc obtained from different classifications according to a certain logic and structure. Create a preprocessed data bundle containing multiple sets of question-answer knowledge related databases. Among them, the database consists of data chains of various lengths covering question text, answer text and knowledge-related labels. Group processing is performed according to the data chain type, and the question text data chain, answer text data chain and knowledge-related label data chain are classified separately. Then, with the help of the data recognition channel, the data chain identification information and channel rules are used to transmit the same type of data chain to the storage unit. For example, all question text data chains are transmitted to the corresponding area of ​​the storage unit through a special question text data recognition channel.

[0037] The storage unit is pre-set with question classification templates and answer matching templates. After receiving the data chain, an initial match is first performed. For example, according to the question classification template, the question text data chain is matched with the existing question classification to determine the problem field to which it belongs. Then, the data chain is sorted from large to small according to factors such as the number of question and answer knowledge or the length of the text contained in the data chain and stored. For example, the question data chain containing more knowledge points or longer texts is placed in front for priority processing and retrieval. The storage unit is pre-set with question classification templates and answer matching templates. After receiving the data chain, an initial match is first performed. For example, according to the question classification template, the question text data chain is matched with the existing question classification to determine the problem field to which it belongs. Then, the data chain is sorted from large to small according to factors such as the number of question and answer knowledge or the length of the text contained in the data chain and stored. For example, the question data chain containing more knowledge points or longer texts is placed in front for priority processing and retrieval. The matched data chains are calculated in sequence, starting from the data chain with the smallest capacity. The calculation method includes the integration and association analysis of question and answer knowledge. For example, for a simple question data chain with a small capacity, directly perform answer matching and associated knowledge integration, such as associating relevant product manual fragments, operating instructions, etc. into the answer. The calculation results are stored in the result template. If the subsequent calculation encounters the same type of question and answer knowledge data group, the calculation is skipped and the result template data is read. At the same time, a matching template is generated and stored in the memory storage unit for easy call. For example, if a question and answer data group about the basic operation of a certain product has been calculated, when a similar question is encountered again, the data in the result template is directly read to improve processing efficiency.

[0038] Example 3

[0039] The storage module receives the analysis and calculation result template data from the data processing module, and matches and verifies it with the result template. Data verification algorithms, such as hash verification and data structure comparison, are used to ensure the accuracy and integrity of the stored data. For example, when storing a question and answer result data, its data hash value is calculated and compared with the hash value of the corresponding data in the result template. If they are consistent, the storage is successful, otherwise an error prompt and data repair processing is performed. The stored data is backed up and sorted regularly, and the data is stored in different storage media or storage areas according to factors such as the frequency of use and importance of the data. For example, frequently queried popular question and answer data is stored in the cache area, and historical data is backed up in a large-capacity disk array.

[0040] The data display terminal obtains the processed question and answer data results from the storage module and displays them to the operator in a user-friendly interface. For example, questions and answers are displayed in clear text format, charts (such as knowledge association diagrams) in the form of web interfaces, desktop application interfaces, etc. At the same time, the invalid data module processes the invalid question and answer data input. When question and answer data that does not meet system requirements or cannot be recognized is received, such as garbled data, data unrelated to the knowledge base topic, etc., it is marked and recorded, and transmitted to the data display terminal for the operator to observe. The operator can adjust and optimize the data collection source or data processing process based on this invalid data information.

[0041] Example 4

[0042] According to the implementation steps of the above source data processing module, data organization module and data storage and display module, the question-answering knowledge base is gradually constructed. First, the source data acquisition unit is started to collect a large amount of original data, and then through a series of processes such as feature extraction and classification, data integration and transmission, and storage unit processing, the question-answering knowledge data in the question-answering knowledge base is continuously enriched and improved. During the establishment process, parameters such as classification rules, matching templates, and calculation algorithms can be adjusted according to actual needs and data characteristics to improve the quality and applicability of the knowledge base.

[0043] Data acquisition operation: The computer obtains question-and-answer related data information through various input methods (such as user input text, voice recognition input, etc.). For example, a user enters the question "How to check the status of a product order?" in the input box of the intelligent customer service system.

[0044] Data processing operation: Data is transmitted to the data processing center. The data grouping unit of the data processing center groups the data according to the data link type and identifies the question data as the question text data link type. The data sorting unit sorts the grouped data links by capacity. Since this is a newly input question, there is no sorting comparison for the time being. The data analysis and calculation unit calculates the sorted and matched data links from the smallest capacity. The calculation method includes the integration and association analysis of question and answer knowledge. Search the knowledge base for question and answer knowledge data groups related to the question. If the same type of question and answer knowledge data group is found, the calculation is skipped and the result template data is read. If there is already a standard answer about querying the product order status, it is directly read and returned to the user. The memory storage unit stores the matching template for the next call, and records the relevant information of this question and answer, such as the keywords of the question, the question time, etc., for subsequent data analysis and knowledge base optimization.

[0045] Result display operation: Data results are presented to operators through the data display terminal. For example, the intelligent customer service interface displays "You can log in to our official website and enter the order number and verification code on the order inquiry page to query the order status." The storage module stores records of data results and matches and checks with the result template to ensure the correct storage and traceability of data. The invalid data module processes invalid question and answer data and transmits it to the data display terminal for operators to observe. If the user enters garbled characters or questions unrelated to the product business, the corresponding prompt information will be displayed on the interface, such as "The question you entered cannot be recognized, please re-enter it."

[0046] In the description of the specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0047] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they shall all fall within the protection scope of the present invention.

Claims

1. A question-answering knowledge base, characterized in that: include: A source data processing module is configured to collect data in a computer system and mark it as source data, extract feature data such as text semantic structure, keyword type and information format of the source data to classify the source data, obtain distinguishing data QB and mark it, and analyze the associated data in the distinguishing data QB to generate preprocessing data Yc, wherein the associated data is text data with the same semantic theme or related keywords; The data organization module integrates the pre-processed data Yc obtained from different classifications into a pre-processed data bundle containing multiple sets of question-answering knowledge-related databases, wherein the database is composed of data chains of various lengths covering question texts, answer texts, and knowledge-related labels. The module also performs group processing according to the type of data chain, and transmits the same type of data chains to the storage unit with the help of the data identification channel. The storage unit is pre-set with question classification templates and answer matching templates, etc., which are used to match, sort, re-match and store the data chains for the second time, and calculate the matched data chains in sequence, starting from the data chain with the smallest capacity, and storing the results in the result template. If the subsequent calculation encounters the same type of question-answering knowledge data group, the calculation is skipped and the result template data is read. At the same time, the matching template is generated and stored in the memory storage unit for calling; The data storage and display module includes a storage module and a data display terminal. The storage module is used to store and record analysis and calculation result template data and match and verify it with the result template. The data display terminal is used to display the question and answer data results processed by the storage module and the invalid question and answer data transmitted by the invalid data module. The invalid data module is used to process the input invalid question and answer data.

2. A method for establishing a question-answer knowledge base as claimed in claim 1, characterized in that: The specific steps include: Collect the data information obtained by the computer and mark it as source data, extract the feature data of the text semantic structure, keyword type and information format of the source data, classify the source data according to the feature data to obtain the distinguishing data QB and mark it; Analyze and process the associated data in the distinguishing data QB, i.e., the text data of the same semantic topic or related keywords, to generate preprocessed data Yc; The preprocessed data Yc obtained by different classifications are bundled to form a preprocessed data bundle package containing multiple sets of question-answer knowledge related databases. The database consists of data links of various lengths. The data link types include question text, answer text and knowledge related tags. The data links are grouped according to the data link type and the data links of the same type are transmitted to the storage unit with the help of a data identification channel. The storage unit matches the data chain using a pre-set template, sorts and stores the data chain from large to small according to factors such as the number of question-answer knowledge or text length contained in the data chain, and then performs more accurate matching on the sorted data chain based on keyword range, semantic depth, etc. and feeds it back to the storage unit for secondary storage; The matched data chains are calculated sequentially starting from the one with the smallest capacity. The calculation method includes the integration of question and answer knowledge and association analysis. The calculation results are stored in the result template. If the subsequent calculation encounters the same type of question and answer knowledge data group, the calculation is skipped and the result template data is read. At the same time, a matching template is generated and stored in the memory storage unit.

3. An intelligent question-answering method for a question-answering knowledge base as claimed in claim 1, characterized in that: The specific steps include: Data acquisition operation: the computer acquires data information related to the question and answer; Data processing operation: data is transmitted to the data processing center. The data grouping unit of the data processing center groups the data according to the data link type. The data sorting unit sorts the grouped data links according to the capacity. The data analysis and calculation unit calculates the sorted and matched data links starting from the smallest capacity. The calculation method includes the integration and association analysis of question and answer knowledge. If the same type of question and answer knowledge data group is encountered, the calculation is skipped and the result template data is read. The memory storage unit stores the matching template for easy calling. Result display operation: The data results are presented to the operator through the data display terminal. The storage module stores and records the data results and matches and verifies them with the result template. The invalid data module processes invalid question and answer data and transmits it to the data display terminal for the operator to observe.

4. A system for establishing a question-answer knowledge base as claimed in claim 1, characterized in that: include: A source data acquisition unit, used for acquiring data in a computer system and marking it as source data; The feature extraction and classification unit extracts the feature data such as the text semantic structure, keyword type and information format of the source data to classify the source data, obtain the distinguishing data QB and mark it, and can analyze the associated data in the distinguishing data QB to generate pre-processed data Yc; A data integration and transmission unit integrates the pre-processed data Yc obtained from different classifications into a pre-processed data bundle containing multiple sets of question-answer knowledge related databases; The data chain processing unit and the storage unit are pre-set with question classification templates and answer matching templates. The unit is used to match, sort, re-match and store the data chains, and calculate the matched data chains in sequence, starting from the data chain with the smallest capacity, and storing the results in the result template. If the subsequent calculation encounters the same type of question and answer knowledge data group, the calculation is skipped and the result template data is read. At the same time, a matching template is generated and stored in the memory storage unit; The storage and display unit includes a storage module and a data display terminal. The storage module is used to store and record analysis and calculation result template data and match and verify it with the result template. The data display terminal is used to display the question and answer data results processed by the storage module and the invalid question and answer data transmitted by the invalid data module. The invalid data module is used to process the input invalid question and answer data.

Images