Question answering method, device, equipment and product

Abstract

The invention discloses a question answering method, device, equipment and product, and relates to the technical field of man-machine interaction, and the method comprises the steps: carrying out the type analysis of a received user question through a question classification model, and obtaining a user question type; a service interface is determined according to the user question type, a pre-constructed business knowledge base is called through the service interface to generate question reply, and the business knowledge base is obtained by performing data fusion and data updating through an integration model. Therefore, the type of the question of the user is analyzed to obtain the corresponding type of the question of the user, then the service interface for replying is determined according to the type of the question, and finally the knowledge base is called through the service interface to generate the question reply, so that the question reply result is obtained. The problem that corresponding knowledge bases cannot be used to meet user requirements when users ask questions is solved, and the accuracy of question answering is improved.

Description

Technical Field

[0001] The present application relates to the field of human-computer interaction technology, and in particular to a question-answering method, apparatus, device, and product. Background Art

[0002] Currently, the banking industry generally uses manual methods to enter and maintain knowledge related to wholesale business analysis (such as industry policy interpretation, risk models, business processes, etc.) into dedicated application systems. Users enter keywords and perform fuzzy searches to obtain the required knowledge. However, this technical solution has significant drawbacks in practical applications:

[0003] First, the existing system lacks the ability to intelligently identify user search intent. Since the search keywords entered by users are often ambiguous, diverse, and unstructured, the system only relies on simple keyword matching algorithms and cannot effectively understand the deep intentions of user queries or business scenario requirements, which seriously reduces the efficiency of knowledge acquisition.

[0004] Secondly, the existing system's knowledge base has limited content coverage and is updated with a lag. Currently, the entry of knowledge content relies entirely on manual organization and maintenance, which not only requires a large amount of human resources, but is also prone to missing, obsolete or repeated content due to changes in business scenarios or untimely knowledge updates. Furthermore, manual entry methods make it difficult to efficiently integrate multi-source heterogeneous data (such as real-time market trends, corporate financial reports, regulatory policies, etc.), resulting in insufficient breadth and depth of the knowledge base, which cannot meet users' analysis needs for complex business scenarios (such as cross-border trade financing and supply chain financial risk warnings).

[0005] The above content is only used to assist in understanding the technical solution of this application and does not constitute an admission that the above content is prior art. Summary of the Invention

[0006] The main purpose of this application is to provide a question-answering method, device, equipment and product, aiming to solve the technical problem of not being able to meet user needs when facing user questions.

[0007] To achieve the above objectives, the present application proposes a question-answering method, which is applied to a business analysis platform and includes:

[0008] Analyze the types of user questions received through the question classification model to obtain the user question types;

[0009] A service interface is determined according to the type of question asked by the user, and a pre-built business knowledge base is called through the service interface to generate a response to the question. The business knowledge base is obtained by performing data fusion and data update through an integration model.

[0010] In one embodiment, before the step of analyzing the type of the received user questions using the question classification model to obtain the type of the user questions, the method further includes:

[0011] Collecting historical access data of the business analysis platform, and analyzing the historical access data to obtain dimension classification;

[0012] Establishing an intent classification framework based on the dimensional classification, and dividing the intent into subcategories based on the intent classification framework;

[0013] Building a training data set based on the historical access data;

[0014] According to the training data set, the intent classification framework and the intent sub-classification are trained on question types through a machine learning algorithm to obtain a question classification model.

[0015] In one embodiment, the step of performing type analysis on the received user questions using the question classification model to obtain the type of the user questions includes:

[0016] Preprocessing the user question, and performing word segmentation processing on the preprocessed user question to obtain a word segmentation result;

[0017] Performing word frequency inverse document frequency analysis on the word segmentation results to obtain word segmentation weights;

[0018] Vectorizing the word segmentation result to obtain a word segmentation vector;

[0019] Intent recognition is performed based on the word segmentation vector to obtain user intent, and type analysis is performed on the user intent through the question classification model according to the word segmentation weight to obtain the user question type, which includes platform type, tool type and data type.

[0020] In one embodiment, the step of determining a service interface according to the type of the user's question and generating a question answer by calling a pre-built business knowledge base through the service interface includes:

[0021] In the case where the user's question type is platform type, calling the retrieval and generation RAG service interface to query the platform data in the business knowledge base to obtain the answer document;

[0022] Information is extracted from the answer document to obtain key information of the answer, and a question answer is generated based on the key information of the answer.

[0023] In one embodiment, the step of determining a service interface according to the type of the user's question and generating a question answer by calling a pre-built business knowledge base through the service interface further includes:

[0024] In the case where the user question type is a tool type, extracting a tool keyword according to the word segmentation vector;

[0025] Determine a tool service interface based on the tool keyword, and call the tool knowledge in the business knowledge base through the tool service interface to retrieve a reply to the user's question, thereby obtaining a reply retrieval result;

[0026] If the answer search result is empty, the user question is sent to the manual service end, and the manual service end generates a question answer;

[0027] If the answer retrieval result is not empty, a question answer is generated based on the answer retrieval result.

[0028] In one embodiment, the step of determining a service interface according to the type of the user's question and generating a question answer by calling a pre-built business knowledge base through the service interface further includes:

[0029] In the case where the user question type is a data type, extracting a data source set and a report information set corresponding to the user question in the business knowledge base through a data service interface;

[0030] Screening the data source set and the report information set to obtain a response data source and report information;

[0031] The answer data source and the report information are integrated to obtain the answer to the question.

[0032] In one embodiment, before the step of determining a service interface based on the user's question type and invoking a pre-built business knowledge base through the service interface to generate a question answer, the method further includes:

[0033] Receiving a business development request, parsing the business development request to obtain a business complexity and a business type identifier;

[0034] Obtaining multi-source heterogeneous data based on the business type identifier query;

[0035] According to the complexity of the business, extracting the data to be integrated and key information from the multi-source heterogeneous data through the integration model;

[0036] Calculating the cosine similarity between the data to be integrated and the initial knowledge base, and generating a deduplication mark if the similarity exceeds a threshold;

[0037] Adding the data to be integrated to the initial knowledge base;

[0038] The initial knowledge base is deduplicated based on the deduplication identifier, and the content of the initial database is optimized using the key information to obtain a business knowledge base.

[0039] In addition, to achieve the above-mentioned purpose, the present application also proposes a question-answering device, which is applied to a business analysis platform and includes:

[0040] An analysis module is used to analyze the types of user questions received through a question classification model to obtain the types of user questions;

[0041] The generation module is used to determine the service interface according to the user question type, and generate a question answer by calling a pre-built business knowledge base through the service interface. The business knowledge base is obtained by data fusion and data update through the integration model.

[0042] In addition, to achieve the above-mentioned purpose, the present application also proposes a question-answering device, which includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program is configured to implement the steps of the question-answering method described above.

[0043] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the question-answering method described above.

[0044] One or more technical solutions proposed in this application have at least the following technical effects:

[0045] The embodiment of the present application proposes a question-answering method, device, equipment, and product. The method analyzes the type of user questions received through a question classification model to obtain the type of user questions; determines a service interface based on the user question type, and generates a question answer by calling a pre-built business knowledge base through the service interface. The business knowledge base is obtained by performing data fusion and data update through an integration model. Thus, the user's question is analyzed by type to obtain the corresponding type of the user's question, and then the service interface for answering is determined based on the type of the question. Finally, the knowledge base is called through the service interface to generate a question answer, and the question answer result is obtained. This solves the problem of not being able to use the corresponding knowledge base to meet user needs when facing user questions, and improves the accuracy of question answers. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0048] Figure 1 A flowchart illustrating the first embodiment of the question-answering method of this application is provided;

[0049] Figure 2 The question-answering method for this application involves a schematic diagram of performing type analysis on user questions;

[0050] Figure 3 A flowchart of Example 2 of the question-answering method of this application is provided;

[0051] Figure 4 Generate a schematic diagram for responses to platform-type questions related to the question-answering method of this application;

[0052] Figure 5 Generate a schematic diagram for the responses to questions about the tool type involved in the question-answering method for this application;

[0053] Figure 6 Generate a schematic diagram for responses to questions about data types related to the question-answering method of this application;

[0054] Figure 7 This is a schematic diagram of the module structure of the question-answering device according to an embodiment of the present application;

[0055] Figure 8 This is a schematic diagram of the device structure of the hardware operating environment involved in the question-answering method in the embodiment of the present application.

[0056] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0057] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0058] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0059] The main solution of the embodiment of the present application is: collecting the historical access data of the business analysis platform and parsing the historical access data to obtain dimension classification; establishing an intent classification framework based on the dimension classification and dividing the intent sub-classification according to the intent classification framework; constructing a training data set based on the historical access data; using the training data set, the intent classification framework and the intent sub-classification are trained on the question type by a machine learning algorithm to obtain a question classification model. The user questions are pre-processed and the pre-processed user questions are segmented to obtain a segmentation result; the segmentation result is subjected to word frequency inverse document frequency analysis to obtain a segmentation weight; the segmentation result is vectorized to obtain a segmentation vector; based on the segmentation vector, intent recognition is performed to obtain user intent, and based on the segmentation weight, the user intent is type analyzed by the question classification model to obtain the user question type, which includes platform type, tool type and data type. When the user question type is platform type, the retrieval and generation RAG service interface is called to query the platform data in the business knowledge base to obtain an answer document; information is extracted from the answer document to obtain key information for the answer, and a question answer is generated based on the key information for the answer. In the case where the user question type is a tool type, a tool keyword is extracted based on the word segmentation vector; a tool service interface is determined based on the tool keyword, and the tool knowledge in the business knowledge base is called through the tool service interface to perform a reply search for the user question to obtain a reply search result; if the reply search result is empty, the user question is sent to the manual service end, and the manual service end generates a question reply; if the reply search result is not empty, a question reply is generated based on the reply search result. In the case where the user question type is a data type, a data source set and a report information set corresponding to the user question in the business knowledge base are extracted through a data service interface; the data source set and report information set are screened to obtain a reply data source and report information; and the reply data source and report information are integrated to obtain a question reply. The system receives a business development request, parses the request to obtain business complexity and a business type identifier; queries multi-source heterogeneous data based on the business type identifier; extracts data to be integrated and key information from the multi-source heterogeneous data using the integration model based on the business complexity; calculates the cosine similarity between the data to be integrated and the initial knowledge base, and generates a deduplication identifier if the similarity exceeds a threshold; adds the data to be integrated to the initial knowledge base; deduplicates the initial knowledge base based on the deduplication identifier, and optimizes the content of the initial database using the key information to obtain a business knowledge base. This solves the problem of being unable to use the corresponding knowledge base to meet user needs when facing user questions, achieves responses to questions, and improves the accuracy of responses to questions.Based on the solution of the present invention, in view of the fact that the knowledge base content coverage of the existing system is limited and the update is delayed, and the current entry of knowledge content relies entirely on manual organization and maintenance, which not only requires a large amount of human resources, but is also prone to content loss due to changes in business scenarios or untimely knowledge updates, and cannot guarantee the accuracy of the answer results, thus leading to low accuracy, a question-answering method is designed, and the effectiveness of the question-answering method of the present invention is verified when answering questions. Finally, the accuracy of question-answering using the method of the present invention is significantly improved.

[0060] In this embodiment, for ease of description, the following description is made with the question-answering device as the execution entity.

[0061] Since the business processing systems in the existing technology lack the ability to intelligently identify user search intentions, and the search keywords entered by users are often ambiguous, diverse, and unstructured, the system only relies on simple keyword matching algorithms and cannot effectively understand the deep intentions of user queries or business scenario requirements, which seriously reduces the efficiency of knowledge acquisition. Furthermore, the knowledge base content coverage of the existing system is limited and the update is delayed. Currently, the entry of knowledge content relies entirely on manual organization and maintenance, which not only requires a large amount of human resources, but is also prone to missing, obsolete or repeated content due to changes in business scenarios or untimely knowledge updates. Furthermore, manual entry methods make it difficult to efficiently integrate multi-source heterogeneous data (such as real-time market dynamics, corporate financial reports, regulatory policies, etc.), resulting in insufficient breadth and depth of the knowledge base, which cannot meet users' analysis needs for complex business scenarios (such as cross-border trade financing, supply chain financial risk warnings), thereby resulting in a decrease in the accuracy of the responses.

[0062] This application provides a solution. In the business analysis platform, the question types of users' questions are first analyzed. At the same time, the service interface is determined based on the obtained question types. Finally, the corresponding knowledge base is called through the service interface to generate question answers, providing users with better services.

[0063] This application uses a question classification model to analyze the types of user questions received to obtain the user question types; determines the service interface based on the user question type, and generates a question answer by calling a pre-built business knowledge base through the service interface. The business knowledge base is obtained by performing data fusion and data update through an integration model. Thus, the user's question is analyzed by type to obtain the corresponding type of the user's question, and then the service interface for answering is determined based on the question type. Finally, the knowledge base is called through the service interface to generate a question answer, and the question answer result is obtained. This solves the problem of not being able to use the corresponding knowledge base to meet user needs when facing user questions, and improves the accuracy of question answers.

[0064] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, mobile phone, etc., or an electronic device or question-answering device capable of implementing the above functions. The following uses the question-answering device as an example to illustrate this embodiment and the following embodiments.

[0065] Based on this, the embodiment of the present application provides a question answering method, referring to Figure 1 , Figure 1 This is a flowchart of the first embodiment of the question-answering method of this application.

[0066] In this embodiment, the question answering method is applied to a business analysis platform, and the question answering method includes steps S05 to S06:

[0067] Step S05: Analyze the type of the user's question received by the question classification model to obtain the user's question type;

[0068] It should be clear that the present embodiment addresses two key issues in knowledge management systems in the banking industry:

[0069] (1) Insufficient search capabilities: The system only supports basic keyword matching and cannot intelligently identify users' vague and diverse search intentions or business scenario requirements, resulting in low knowledge acquisition efficiency;

[0070] (2) Limited quality of the knowledge base: It relies on manual entry and maintenance, resulting in delayed content updates and prone to duplication or omissions. At the same time, it is difficult to integrate multi-source heterogeneous data (such as market trends, financial reports, and policies), making it difficult to support the analytical needs of complex businesses such as cross-border trade financing and supply chain finance.

[0071] Therefore, in this embodiment, the received user questions are first analyzed by type to determine the type of user question. The solution of this embodiment is primarily applied to a business analysis platform. That is, within an enterprise, business personnel who need to obtain customer information and corresponding technical support can use the business analysis platform to ask questions and obtain solutions to their problems. Therefore, user questions in this embodiment can be understood as questions about the platform (e.g., I need to retrieve customer information for the second quarter and don't know how to do so on the platform), tool (e.g., I want to create a visual report on sales for the second quarter and how can I use TABLEAU and Guanyuan?), and data (e.g., how is department data for the second quarter calculated?). Therefore, the business analysis platform of this embodiment analyzes received questions to determine the specific question type for subsequent response generation based on the question type in different situations. In the subsequent explanation, this embodiment uses the Huoyan platform (i.e., the Huoyan wholesale business analysis platform) as an example. This platform provides low-threshold data analysis services to 40% of the bank's business personnel, and is deeply applied in business scenarios such as operational analysis, business thematic analysis, customer profiling, user profiling, performance assessment statements, daily deposit notifications, and marketing acquisition.

[0072] Step S06: determining a service interface according to the type of question asked by the user, and generating a response to the question by calling a pre-built business knowledge base through the service interface.

[0073] After understanding the type of question the user is asking, we can determine the service interface to use based on the type of question the user is asking. That is, we can use the corresponding service interface to call the knowledge base related to the user's question to generate the reply content, so as to obtain the reply to the user's question.

[0074] Through the above method, the corresponding knowledge base can be called to generate answer content for different types of user questions, thereby improving the user's actual experience.

[0075] Specifically, before performing type analysis on user questions, a question classification model needs to be pre-built for type analysis. Therefore, in step S05 of this embodiment, before performing type analysis on the received user questions using the question classification model to obtain the type of the user questions, the method further includes:

[0076] Step S01: collecting historical access data of the business analysis platform, parsing the historical access data, and obtaining dimension classification;

[0077] Step S02: establishing an intent classification framework based on the dimensional classification, and dividing the intent into subcategories based on the intent classification framework;

[0078] Step S03, constructing a training data set based on the historical access data;

[0079] Step S04: Based on the training data set, the intent classification framework and the intent sub-classification are trained on question types by a machine learning algorithm to obtain a question classification model.

[0080] For the analysis of the types of user questions, in this embodiment, a question classification model is used for analysis, and an exemplary question classification model construction process is as follows:

[0081] (1) Collecting and parsing historical access data: In this embodiment, the data mainly comes from the business analysis platform. Therefore, when building the model, it is necessary to first obtain the user access logs of the business analysis platform, including structured and unstructured data such as search records, click behaviors, session paths, and length of stay. Then, the data is parsed for dimension classification. Natural language processing (NLP) technology (word segmentation, entity recognition, semantic clustering) is used to extract keywords, high-frequency phrases, and user behavior tags, etc., to obtain the core dimensions of type classification: business dimension (risk model, industry policy, supply chain finance), user role dimension (risk control personnel, customer manager, management), scenario dimension (real-time warning, process optimization, compliance review), and time dimension (quarterly analysis, emergency response), etc.

[0082] (2) Establish an intent classification framework: The framework design principle can follow a hierarchical classification, i.e., first-level classification (business category) → second-level classification (scenario subcategory) → third-level classification (specific intent). For example, the first-level classification corresponds to risk warning, policy interpretation, and process query; the second-level classification corresponds to cross-border trade financing risks and supply chain financial compliance policies, etc. The intent subclassification (third-level classification) rules correspond to defining subclass boundaries based on user behavior patterns in historical data (the correlation between high-frequency search terms and final operations).

[0083] (3) Constructing a training dataset: This includes data annotation and processing. First, manual or semi-automatic annotation is used (preliminary annotation based on rule matching, manual verification and correction). Then, feature engineering (TF-IDF, word vectors (Word2Vec / BERT), contextual semantic features) is used to extract features. Finally, a unique code is assigned to each intent subclass. It should be clear that the training dataset in this embodiment needs to cover all intent subclasses to avoid class imbalance (which can be optimized through oversampling or undersampling).

[0084] (4) Machine learning model training and optimization: The final step is to train a specific model. In this embodiment, algorithms can be selected based on actual business needs, that is, the size and complexity of the data, including but not limited to SVM, random forest (suitable for small-scale data), BERT, TextCNN, LSTM (suitable for complex semantic understanding) and Stacking (combining the advantages of multiple models);

[0085] The specific training process is to vectorize the text and behavioral features in the training data, then define the loss function (cross entropy loss) and optimizer (such as Adam), and adjust the hyperparameters (learning rate, batch size) through grid search or Bayesian optimization. During the training and use of the model, it is also necessary to evaluate the accuracy, recall rate, F1 value, and confusion matrix. The model is iteratively updated with the above accuracy, recall rate, F1 value, and confusion matrix to adapt to business changes. Finally, the model is deployed as an API service and integrated into the business analysis platform to support real-time intent recognition and return matching knowledge base content or operation suggestions.

[0086] The above solution is used to build a question classification model. It not only takes into account the data set construction of specific business scenarios, but also selects different models for different business volumes. This allows for more accurate analysis results when analyzing user types.

[0087] More specifically, after obtaining the question classification model, the type of the received user question can be analyzed. Therefore, the above step S05, in which the type of the received user question is analyzed by the question classification model, the step of obtaining the type of the user question includes:

[0088] Step S051, pre-processing the user question, and performing word segmentation processing on the pre-processed user question to obtain a word segmentation result;

[0089] Step S052, performing word frequency inverse document frequency analysis on the word segmentation result to obtain a word segmentation weight;

[0090] Step S053, vectorizing the word segmentation result to obtain a word segmentation vector;

[0091] In step S054, intent recognition is performed based on the word segmentation vector to obtain user intent, and type analysis is performed on the user intent through the question classification model according to the word segmentation weight to obtain the user question type, which includes platform type, tool type and data type.

[0092] After obtaining the question classification model, the user's question can be preprocessed. The purpose is to convert the original text input by the user into a form suitable for computer processing. The operations of preprocessing include removing noise (removing irrelevant symbols, punctuation marks, special characters, etc.), standardization (unifying case, unifying number representation, etc.), removing stop words (stop words are words that do not help much in understanding sentences, such as "of", "is", etc.), and correcting spelling mistakes (that is, in the case of spelling mistakes, they will also be corrected during the preprocessing process).

[0093] Subsequently, word segmentation is performed on the text, that is, the process of breaking it into individual words or phrases. In Chinese processing, word segmentation is a key step because there is no clear word delimiter in Chinese. For example, the word segmentation result of "Wholesale operation analysis report" may be "Wholesale", "operation", "analysis", "report". The result obtained after word segmentation is a list composed of individual words.

[0094] After completing word segmentation, term frequency-inverse document frequency (TF-IDF) analysis is performed. It should be clear that term frequency (TF, Term Frequency) refers to the frequency of a certain word in a document, and inverse document frequency (IDF, Inverse Document Frequency) is the inverse of the frequency of a word in all documents. The purpose is to reduce the weight of common words and increase the weight of rare words. TF-IDF is the product of TF and IDF, indicating the importance of a certain word. If a word appears frequently in a certain document and rarely in other documents, then the TF-IDF value of this word is relatively high, indicating that it makes a greater contribution to this document. Therefore, in this embodiment, through TF-IDF analysis, a weight can be assigned to each word, indicating its importance in the text.

[0095] To facilitate the analysis of the question classification model, in this embodiment, the text data is converted into a digital form, that is, vectorized, so that it can be used as the input of a machine learning model. The vectorization methods adopted in this embodiment include but are not limited to the bag-of-words model (Bag of Words), that is, converting the text into a fixed-length vector, where each dimension represents a word in the vocabulary, and the value represents the frequency of the word in the text, as well as embedding vectors such as Word2Vec and GloVe, that is, mapping the vocabulary to points in a high-dimensional space, and the semantic similarity between the vocabulary can be measured by the distance between the vectors.

[0096] Finally, as Figure 2As shown, the word segmentation vector is used for intent recognition to obtain the user intent, and the question classification model is used to analyze the user intent to determine the type of user question. This model is trained to obtain a classifier, which can predict which category the question belongs to based on the input word segmentation vector. In this embodiment, the question type can be the following: platform type (that is, the user question may involve a certain platform or system, such as a specific operating system or software platform), tool type (the user question may involve a certain tool or function, such as data analysis tools, programming tools, etc.) and data type (that is, the user question may involve a certain data or data processing method, such as data format, database query, etc.).

[0097] A series of processing is performed on the user's questions to obtain the user's intention, and then the question classification model can be used to perform a specific type analysis of the user's intention to obtain the question type of the user's question, which in this embodiment includes platform type, tool type and data type, and can achieve faster calling of the service interface to generate the reply content.

[0098] In this embodiment, the business knowledge base can be updated not only for historical access data but also for newly launched businesses. Therefore, before the step S06 of determining a service interface based on the user's question type and generating a question answer by calling the pre-built business knowledge base through the service interface, the method further includes:

[0099] Step S0601: receiving a business development request, parsing the business development request to obtain a business complexity and a business type identifier;

[0100] Step S0602: obtaining multi-source heterogeneous data based on the business type identifier;

[0101] Step S0603: extracting the data to be integrated and key information from the multi-source heterogeneous data using the integration model according to the business complexity;

[0102] Step S0604, calculating the cosine similarity between the data to be integrated and the initial knowledge base, and generating a deduplication flag if the similarity exceeds a threshold;

[0103] Step S0605, adding the data to be integrated to the initial knowledge base;

[0104] Step S0606: Deduplication is performed on the initial knowledge base based on the deduplication identifier, and content optimization is performed on the initial database using the key information to obtain a business knowledge base.

[0105] When a business development request is received or the existing business needs to be expanded and updated, the business analysis platform receives and parses the business development request. The request contains various types of information related to the current business, including but not limited to the business's goals, needs, priorities, etc. The system parses this information to obtain two key parameters: business complexity and business type identification. Business complexity indicates the complexity of business needs, which may include multiple dimensions such as technical complexity, market complexity, and resource requirements. The business type identification is used to indicate the field or category to which the business belongs, which may include sales, R&D, supply chain, and other types.

[0106] Based on the business type identifier obtained, the system automatically queries relevant data sources. These data sources may come from different systems, platforms, partners or external public resources. These data sources are heterogeneous and may contain structured data (such as database tables), semi-structured data (such as JSON, XML files) or unstructured data (such as text, pictures, videos, etc.). The query results return multi-source heterogeneous data, which are the core resources to support business development.

[0107] The system uses an integration model (including data preprocessing and cleaning model, feature extraction and selection model, machine learning model and similarity calculation model) to analyze the complexity of the business and decide how to extract the most critical data and information for the current business from multi-source heterogeneous data. The integration model uses certain algorithms (such as data preprocessing, data cleaning, feature extraction, etc.) to filter out the data to be integrated and key information from complex heterogeneous data. This information is of great value for the subsequent optimization and update of the knowledge base. Then the cosine similarity is calculated, the deduplication mark is determined, and the cosine similarity is calculated for the extracted data to be integrated and the existing data in the initial knowledge base. If the similarity between the data to be integrated and the data in the initial knowledge base exceeds the set threshold, a deduplication mark is generated, indicating that the data already exists in the knowledge base and does not need to be added repeatedly.

[0108] If there is no significant duplication between the data to be integrated and the data in the knowledge base (i.e., the cosine similarity does not exceed the threshold), the system will add the data to be integrated to the initial knowledge base. The initial knowledge base refers to an existing, unoptimized data set that serves as the basis for all business decisions and analysis. The system then deduplicates the initial knowledge base based on deduplication identifiers, removing duplicate data entries to ensure that the knowledge stored in the knowledge base is high-quality and non-redundant. At the same time, the system uses key information to optimize the content of the initial knowledge base. The optimization process may include supplementing, deleting, or reconstructing existing content based on business complexity and business needs to improve the accuracy and business relevance of the knowledge base.

[0109] Finally, after deduplication and optimization, the knowledge base forms a complete business knowledge base, which can efficiently support business decisions, optimize business processes, and provide comprehensive data support for subsequent business development.

[0110] This embodiment improves the integration efficiency, accuracy and real-time performance of the knowledge base through intelligent data processing and optimization methods, and can effectively support enterprises in making decision optimization in a dynamic and complex business environment.

[0111] This embodiment, through the above-mentioned solution, specifically analyzes the type of user questions received using a question classification model to obtain the user question type; determines a service interface based on the user question type, and uses the service interface to call a pre-built business knowledge base to generate a question response. The business knowledge base is obtained by integrating data and updating data using an integration model. Thus, the user's question is analyzed to obtain the corresponding type of the user question, and then the service interface for the response is determined based on the question type. Finally, the service interface calls the knowledge base to generate the question response, resulting in the question response result. This solves the problem of not being able to use the corresponding knowledge base to meet user needs when facing user questions, and improves the accuracy of the question response.

[0112] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 3 In step S06, determining a service interface according to the user's question type, and generating a question answer by calling a pre-built business knowledge base through the service interface, the question answering method further includes steps S061 to S062:

[0113] Step S061: When the user's question type is platform type, the retrieval and generation RAG service interface is called to query the platform data in the business knowledge base to obtain an answer document;

[0114] Step S062: extract information from the answer document to obtain key information of the answer, and generate a question answer based on the key information of the answer.

[0115] When it is determined that the user's question type is platform type, Figure 4As shown, first, the business question is rewritten to reformulate the original business question so that it better conforms to the input requirements and specifications of the Huoyan platform tool rag service interface. For example, if the original business question is "Our sales data has been unusual recently, what's going on?", it may be rewritten as "Analyze the reasons for the recent abnormal fluctuations in sales data." The rewritten business question is input into the Huoyan platform tool rag service interface to obtain relevant answers or solutions. The rag service interface searches and matches within its internal knowledge base, data resources, etc., attempting to find content corresponding to the question. It triggers an interface call through specific program code or operation interface, and it is necessary to ensure that the parameters of the interface call are accurate, including information such as the question text and relevant business identifiers.

[0116] Then it determines whether there is an answer. If there is an answer, the answer content is provided based on the called knowledge base. That is, the answers obtained from the rag service interface are organized and formatted, and presented to the user who asked the question or relevant business personnel in a clear and easy-to-understand manner. The answer may include specific data analysis results, explanation of the cause of the problem, recommended solutions, etc.

[0117] If there is no answer, the question will be transferred to experts in the relevant field, customer service teams, etc. through manual channels. At the same time, the demander will be informed that the problem has entered the manual processing stage, and the approximate waiting time and the subsequent method of obtaining answers will be informed. After receiving the question, the manual support staff will further analyze the problem and can use their own professional knowledge, additional research, etc. to find a solution. After finding the answer, they will provide the answer to the demander through a feedback method similar to when there is an answer.

[0118] Specifically, in the above embodiment, it is mentioned that the user question type of the present application also includes the tool type and the data type. Therefore, in step S06 of the above embodiment, the step of determining the service interface according to the user question type and generating the question answer by calling the pre-built business knowledge base through the service interface also includes:

[0119] Step S063: if the user's question type is a tool type, extract tool keywords based on the word segmentation vector;

[0120] Step S064, determining a tool service interface based on the tool keyword, and using the tool service interface to call tool knowledge in the business knowledge base to perform a reply search for the user's question, thereby obtaining a reply search result;

[0121] Step S065: If the answer search result is empty, the user question is sent to the manual service end, and the manual service end generates a question answer;

[0122] Step S066: If the answer search result is not empty, generate a question answer based on the answer search result.

[0123] When it is determined that the user's question type is a tool type, such as Figure 5 As shown, responses to tool-type user questions also need to be rewritten as business questions to make them more consistent with the specifications and requirements of subsequent analysis tool processing.

[0124] Subsequently, it is necessary to extract the tool categories and keywords. In this embodiment, relevant tool category keywords such as "tableau, chart cube, bix" are used. These keywords will be used to subsequently determine whether there are clear tools to choose from. For example, if the user wants to know how to use the above tools to produce corresponding visualization effects, he will mention the relevant tools when asking questions. At this time, the tool selection can be determined. If the specific tool can be determined through keyword extraction, the tool knowledge base retrieval link can be entered. At this time, a search will be performed in the corresponding tool knowledge base to find out whether there is a matching answer. If the answer is retrieved in the tool knowledge base, the process will directly enter the "provide answer" link and feedback the answer to the user who asked the question or the relevant party. If the answer is not retrieved in the tool knowledge base, the process will enter the "provide manual support channel" and manual intervention will be performed for further processing and support, and finally the process ends.

[0125] If there is no clear tool to choose from after extracting keywords, the search results of the Huoyan Knowledge Base and the Tufang Knowledge Base will be merged, and a comprehensive search will be performed on the two knowledge bases. If the answer is found in the merged search, the answer will also be provided to the user. If the answer is not found in the merged search, the manual support channel will be entered, and subsequent matters will be handled manually.

[0126] More specifically, the above step S06, determining a service interface according to the type of the user's question, and generating a question answer by calling a pre-built business knowledge base through the service interface, further includes:

[0127] Step S067: If the user question type is a data type, extracting a data source set and a report information set corresponding to the user question from the business knowledge base through a data service interface;

[0128] Step S068, screening the data source set and report information set to obtain reply data source and report information;

[0129] Step S069: Integrate the response data source and report information to obtain a response to the question.

[0130] When the user's question type is determined to be a data type, such as Figure 6As shown, similarly, the original business problem needs to be rewritten to make it more in line with the specifications and requirements of subsequent processing, and then the RAG service interface (mainly including the Huoyan data operation and maintenance document knowledge base) is called to input the rewritten business problem into the RAG service interface, which is connected to the Huoyan data operation and maintenance document knowledge base. By searching for relevant information in the knowledge base, try to get the answer to the business problem.

[0131] If the answer to the business question is found in the knowledge base, the queried answer will be directly provided to the relevant demander. If the answer is not found in the knowledge base, key information related to the business question will be extracted from relevant data sources or reports. This information may be contained in various data reports, documents, etc. Based on the extracted slot information, reports or data sources closely related to the business question will be further screened out to narrow the search scope and increase the possibility of obtaining effective information.

[0132] Finally, the FireEye application interface is called according to the scenario and a large-scale summary of the acquired knowledge is made. According to the specific business scenario, different interfaces of the FireEye application are called to comprehensively organize and summarize the knowledge acquired from the filtered data sources or reports. If the answer to the business question can be obtained through this step, the answer will be fed back to the demander. If there is still no answer, the manual support channel will be entered and the problem will be transferred to manual for further processing and support.

[0133] The scenarios of reports and data sources called by the application in this embodiment are listed as follows:

[0134]

[0135]

[0136] This embodiment uses the above solution. Specifically, when the user's question type is platform-type, it calls the retrieval and generation RAG service interface to query the platform data in the business knowledge base to obtain an answer document. It then extracts information from the answer document to obtain key information for the answer, and generates a question reply based on the key information. Thus, the user's question is analyzed for type, and the corresponding type of question is obtained. The service interface for replying is then determined based on the question type. Finally, the knowledge base is called through the service interface to generate a question reply, resulting in a question reply result. This solves the problem of being unable to use the corresponding knowledge base to meet user needs when facing user questions, and improves the accuracy of question replies.

[0137] It should be noted that the above examples are only used to understand this application and do not constitute a limitation on the question-answering method of this application. More simple transformations based on this technical concept are all within the scope of protection of this application.

[0138] This application also provides a question answering device, please refer to Figure 7 The question answering device is applied to the business analysis platform, and the question answering device includes:

[0139] An analysis module 10 is used to analyze the types of user questions received through a question classification model to obtain the types of user questions;

[0140] The generating module 20 is used to determine a service interface according to the type of the user's question, and to call a pre-built business knowledge base through the service interface to generate a response to the question.

[0141] The question-answering device provided in this application utilizes the question-answering method of the aforementioned embodiment, resolving the technical issue of being unable to utilize the corresponding knowledge base to meet user needs when answering user questions. Compared to the prior art, the beneficial effects of the question-answering device provided in this application are the same as those of the question-answering method of the aforementioned embodiment. Other technical features of the question-answering device are the same as those disclosed in the aforementioned embodiment and are not further elaborated here.

[0142] The present application provides a question-answering device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the question-answering method in the above-mentioned embodiment one.

[0143] Reference below Figure 8 , which shows a schematic diagram of the structure of a question-answering device suitable for implementing embodiments of the present application. The question-answering device in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 8 The question answering device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0144] like Figure 8As shown, the question-answering device may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory 1002 or programs loaded from a storage device 1003 into a random access memory 1004. Random access memory 1004 also stores various programs and data required for the operation of the question-answering device. Processing device 1001, read-only memory 1002, and random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: an input device 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, hard disk, etc.; and a communication device 1009. Communication device 1009 can allow the question-answering device to communicate with other devices wirelessly or wired to exchange data. Although the figure shows a question-answering device with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented or have alternatively.

[0145] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a read-only memory 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are performed.

[0146] The question-answering device provided in this application utilizes the question-answering method of the aforementioned embodiment, resolving the technical issue of being unable to utilize the corresponding knowledge base to meet user needs when answering user questions. Compared to the prior art, the beneficial effects of the question-answering device provided in this application are the same as those of the question-answering method of the aforementioned embodiment. Other technical features of the question-answering device are the same as those disclosed in the aforementioned embodiment and are not further elaborated here.

[0147] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0148] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0149] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer program) stored thereon, and the computer-readable program instructions are used to execute the question-answering method in the above-mentioned embodiment.

[0150] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0151] The computer-readable storage medium may be included in the question-answering device, or may exist independently without being incorporated into the question-answering device.

[0152] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the question-answering device, the question-answering device: performs type analysis on the received user questions through a question classification model to obtain the user question type; determines the service interface according to the user question type, and calls a pre-built business knowledge base through the service interface to generate question answers.

[0153] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0154] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0155] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0156] The computer-readable storage medium provided in this application is a computer-readable storage medium storing computer-readable program instructions (i.e., a computer program) for executing the aforementioned question-answering method. This computer-readable storage medium can resolve the technical issue of being unable to utilize a corresponding knowledge base to meet user needs when answering user questions. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the question-answering method provided in the aforementioned embodiment, and are not further elaborated here.

[0157] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned question-answering method when executed by a processor.

[0158] The computer program product provided in this application can solve the technical problem of being unable to use the corresponding knowledge base to meet user needs when facing user questions. Compared with the existing technology, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the question-answering method provided in the above embodiment, and will not be repeated here.

[0159] The above description is only part of the embodiments of the present application and does not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A question-answering method, characterized in that: The question-answering method is applied to a business analysis platform, and the question-answering method includes: Analyze the types of user questions received through the question classification model to obtain the user question types; A service interface is determined according to the type of question asked by the user, and a pre-built business knowledge base is called through the service interface to generate a response to the question. The business knowledge base is obtained by performing data fusion and data update through an integration model.

2. The method according to claim 1, wherein Before the step of analyzing the type of the received user questions using the question classification model to obtain the type of the user questions, the method further includes: Collecting historical access data of the business analysis platform, and analyzing the historical access data to obtain dimension classification; Establishing an intent classification framework based on the dimensional classification, and dividing the intent into subcategories based on the intent classification framework; Building a training data set based on the historical access data; According to the training data set, the intent classification framework and the intent sub-classification are trained on question types through a machine learning algorithm to obtain a question classification model.

3. The method according to claim 2, wherein The step of performing type analysis on the received user questions by using the question classification model to obtain the type of the user questions includes: Preprocessing the user question, and performing word segmentation processing on the preprocessed user question to obtain a word segmentation result; Performing word frequency inverse document frequency analysis on the word segmentation results to obtain word segmentation weights; Vectorizing the word segmentation result to obtain a word segmentation vector; Intent recognition is performed based on the word segmentation vector to obtain user intent, and type analysis is performed on the user intent through the question classification model according to the word segmentation weight to obtain the user question type, which includes platform type, tool type and data type.

4. The method according to claim 3, wherein The steps of determining a service interface according to the type of the user's question and generating a question answer by calling a pre-built business knowledge base through the service interface include: In the case where the user's question type is platform type, calling the retrieval and generation RAG service interface to query the platform data in the business knowledge base to obtain the answer document; Information is extracted from the answer document to obtain key information of the answer, and a question answer is generated based on the key information of the answer.

5. The method according to claim 4, wherein The step of determining a service interface according to the type of the user's question and generating a question answer by calling a pre-built business knowledge base through the service interface further includes: In the case where the user question type is a tool type, extracting a tool keyword according to the word segmentation vector; Determine a tool service interface based on the tool keyword, and call the tool knowledge in the business knowledge base through the tool service interface to retrieve a reply to the user's question, thereby obtaining a reply retrieval result; If the answer search result is empty, the user question is sent to the manual service end, and the manual service end generates a question answer; If the answer retrieval result is not empty, a question answer is generated based on the answer retrieval result.

6. The method according to claim 5, wherein The step of determining a service interface according to the type of the user's question and generating a question answer by calling a pre-built business knowledge base through the service interface further includes: In the case where the user question type is a data type, extracting a data source set and a report information set corresponding to the user question in the business knowledge base through a data service interface; Screening the data source set and the report information set to obtain a response data source and report information; The answer data source and the report information are integrated to obtain the answer to the question.

7. The method according to claim 1, wherein Before the step of determining a service interface according to the user's question type and calling a pre-built business knowledge base through the service interface to generate a question answer, the method further includes: Receiving a business development request, parsing the business development request to obtain a business complexity and a business type identifier; Obtaining multi-source heterogeneous data based on the business type identifier query; According to the complexity of the business, extracting the data to be integrated and key information from the multi-source heterogeneous data through the integration model; Calculating the cosine similarity between the data to be integrated and the initial knowledge base, and generating a deduplication mark if the similarity exceeds a threshold; Adding the data to be integrated to the initial knowledge base; The initial knowledge base is deduplicated based on the deduplication identifier, and the content of the initial database is optimized using the key information to obtain a business knowledge base.

8. A question-answering device, characterized in that: The question answering device is applied to the business analysis platform, and the question answering device includes: An analysis module is used to analyze the types of user questions received through a question classification model to obtain the types of user questions; The generation module is used to determine a service interface according to the type of question asked by the user, and to generate a response to the question by calling a pre-built business knowledge base through the service interface.

9. A question-answering device, characterized in that: The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the question-answering method according to any one of claims 1 to 7.

10. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the question answering method according to any one of claims 1 to 7 are implemented.

Images