Question and answer method, device and equipment for big data platform, medium and product

Abstract

The invention discloses a question and answer method, device and equipment for a big data platform, a medium and a product, and relates to the field of big data, artificial intelligence and financial science and technology, and the method comprises the steps: receiving a target user question through a target big data platform, and extracting semantic features of the target user question, positioning a target first-level classification label matched with the target big data platform in a question and answer library, and obtaining a standard question and answer pair list under all second-level classification labels corresponding to the target first-level classification label; matching the target semantic feature with each standard question in a standard question and answer pair list to obtain a plurality of standard question and answer pairs meeting a similarity condition; and combining the standard answers in the plurality of standard question and answer pairs to obtain a feedback result. According to the automatic question and answer method of the big data platform, the response efficiency of the platform is improved, the labor cost is reduced, the platform can generate the comprehensive answer by intelligently combining the standard answers, more comprehensive and more accurate information support is provided for the user, and the user experience is optimized.

Description

Technical Field

[0001] The present invention relates to the fields of big data, artificial intelligence and financial technology, and in particular to a question-answering method, device, equipment, medium and product for a big data platform. Background Art

[0002] With the rapid development of big data technology, more and more business models are transforming online, such as education and training, online shopping, and financial services. With the surge in the number of online users, the demand for online user consultation and Q&A is increasing.

[0003] With the existing online Q&A method, users need to report these problems to the platform by submitting work orders or sending emails, and then need to wait for a long time to respond and solve the problem. This method not only prolongs the feedback cycle, reduces the response efficiency of the platform, but also brings inconvenience to users. In addition, during the problem handling process, platform personnel usually need to respond to a large number of basic and repetitive questions, which not only consumes a lot of human resources, but also slows down the speed of problem solving, thus affecting the user experience. Summary of the invention

[0004] The present invention provides a question-and-answer method, device, equipment, medium and product for a big data platform to solve the problems of low response efficiency, high labor cost and poor user experience when asking and answering users through a big data platform.

[0005] According to one aspect of an embodiment of the present invention, a question-answering method for a big data platform is provided, comprising:

[0006] Receive the target user's questions through the target big data platform, extract the semantic features of the target user's questions, and obtain the target semantic features;

[0007] Locate the target first-level classification label that matches the target big data platform in the question-answer library, and obtain a list of standard question-answer pairs under all second-level classification labels corresponding to the target first-level classification label;

[0008] Matching the target semantic feature with each standard question in the list of standard question-answer pairs to obtain a plurality of standard question-answer pairs that meet similarity conditions;

[0009] The standard answers in the multiple standard question-answer pairs are combined to obtain a comprehensive answer, and the comprehensive answer is used as a feedback result for the target user's question.

[0010] According to another aspect of an embodiment of the present invention, a question-answering device for a big data platform is provided, comprising:

[0011] A feature extraction module is used to receive the target user's questions through the target big data platform, and extract the semantic features of the target user's questions to obtain the target semantic features;

[0012] A list acquisition module is used to locate the target first-level classification label that matches the target big data platform in the question and answer library, and obtain the list of standard question and answer pairs under all second-level classification labels corresponding to the target first-level classification label;

[0013] A question-answer pair acquisition module, used to match the target semantic feature with each standard question in the standard question-answer pair list to obtain multiple standard question-answer pairs that meet the similarity condition;

[0014] The answer acquisition module is used to combine the standard answers in the multiple standard question-answer pairs to obtain a comprehensive answer, and use the comprehensive answer as the feedback result for the target user's question.

[0015] According to another aspect of an embodiment of the present invention, an electronic device is provided, the electronic device comprising:

[0016] at least one processor; and

[0017] a memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the question and answer method of the big data platform described in any embodiment of the present invention.

[0019] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the question-answering method for a big data platform described in any embodiment of the present invention when executed.

[0020] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program / instruction, wherein when the computer program / instruction is executed by a processor, the steps of the method described in any embodiment of the present invention are implemented.

[0021] The technical solution of the embodiment of the present invention receives the target user's question through the target big data platform, and extracts the semantic features of the target user's question to obtain the target semantic features; locates the target first-level classification label that matches the target big data platform in the question and answer library, and obtains the list of standard question and answer pairs under all the second-level classification labels corresponding to the target first-level classification label; matches the target semantic features with each standard question in the list of standard question and answer pairs to obtain multiple standard question and answer pairs that meet the similarity conditions; combines the standard answers in the multiple standard question and answer pairs to obtain a comprehensive answer, and uses the comprehensive answer as the feedback result for the target user's question. The automated question and answer method of the big data platform improves the response efficiency of the platform and reduces the labor cost. The platform can generate a comprehensive answer by intelligently combining standard answers, providing users with more comprehensive and accurate information support, and optimizing the user experience.

[0022] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 is a flow chart of a question-answering method for a big data platform provided according to Embodiment 1 of the present invention;

[0025] Figure 2 is a flowchart of another question-answering method for a big data platform provided according to Embodiment 2 of the present invention;

[0026] Figure 3 It is a schematic diagram of a module composition applicable to an embodiment of the present invention;

[0027] Figure 4 is a flow chart of a question receiving module applicable to an embodiment of the present invention;

[0028] Figure 5 is a flow chart of a classification control module applicable to an embodiment of the present invention;

[0029] Figure 6 is a flow chart of an intelligent reply module applicable to an embodiment of the present invention;

[0030] Figure 7is a flow chart of a manual intervention module applicable to an embodiment of the present invention;

[0031] Figure 8 is a structural diagram of a question-answering device for a big data platform provided according to Embodiment 3 of the present invention;

[0032] Fig. 9 It is a structural diagram of an electronic device for implementing the question-answering method of a big data platform according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0034] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0035] Embodiment 1

[0036] Figure 1 This is a flowchart of a question-answering method of a big data platform provided in the first embodiment of the present invention. This embodiment is applicable to the case of answering user business questions through a big data platform. The method can be executed by a question-answering device of the big data platform. The question-answering device of the big data platform can be implemented in the form of hardware and / or software and can generally be configured in an electronic device. Figure 1 As shown, the method includes:

[0037] S110, receiving target user questions through a target big data platform, and extracting semantic features of the target user questions to obtain target semantic features.

[0038] In the embodiment of the present invention, the target big data platform can be specifically understood as: a platform used to process and analyze large amounts of data, which can store, process and analyze massive amounts of data from different sources, and these data can be structured, semi-structured or unstructured. Among them, structured data refers to data stored and organized in a fixed format, and these data usually contain clear fields and records, such as question title, question content, answer content and user ratings; semi-structured data refers to data with a certain structure but not completely conforming to a fixed pattern, and these data may contain key-value pairs or tags; unstructured data refers to data without a fixed format, such as text files, pictures, videos or audio uploaded by users. Target user problems can be specifically understood as: business problems or technical requirements submitted by users to the target big data platform.

[0039] The target semantic feature can be specifically understood as a feature that is extracted from the target user question and can represent the core meaning of the text. Specifically, the target user question is preprocessed, which may include operations such as removing stop words, removing punctuation marks, and extracting stems to reduce noise, and then the text is vectorized as the target semantic feature, for example, each word is converted into a vector form using a text-to-vector model.

[0040] S120. Locate the target primary classification label that matches the target big data platform in the question and answer library, and obtain a list of standard question and answer pairs under all secondary classification labels corresponding to the target primary classification label.

[0041] In the embodiment of the present invention, the question-answer database can be specifically understood as: a database containing a large number of questions and corresponding answers to help users quickly find solutions to problems or obtain specific information. In the question-answer database, classification tags are usually used to organize and classify questions.

[0042] The first-level classification label can be specifically understood as: the top-level classification in the question-and-answer library, representing the largest classification unit in the question-and-answer library. For example, in the business requirements question-and-answer library, the first-level classification label can be financial business or educational services, etc.; in the technical support question-and-answer library, the first-level classification label can be divided according to the technology type, which can specifically include: the set distributed system infrastructure, batch processing technology, MPP (Massively Parallel Processing) cluster, Gaussian distribution, and the set open source columnar database, etc. When users submit questions to the target big data platform, they need to submit them to the corresponding section in the target big data platform. Each section corresponds to a topic, that is, the first-level classification label. When receiving the target user's questions through the target big data platform, the section source of the target user's questions is obtained to obtain the first-level classification label, and then the matching target first-level classification label is traversed and located in the question-and-answer library.

[0043] Secondary classification labels can be specifically understood as labels that are further subdivided under primary classification labels. Each primary classification label can contain one or more secondary classification labels, which provide more specific classifications to help users find answers to relevant questions more accurately. For example, financial services can include: account management, loan services, deposit products, credit card services, and investment consulting, etc.; educational services can include: online courses, test preparation, vocational training, academic research, and educational consulting, etc.; in the technical support Q&A library, it can include: real-time data analysis, performance tuning, and data model design, etc.

[0044] The standard question-answer pair list can be specifically understood as a collection of questions and answers organized according to certain standards in the question-answer database. These questions and answers are organized into pairs to facilitate user query and retrieval. Each question-answer pair can contain a question and one or more corresponding answers.

[0045] S130: Match the target semantic feature with each standard question in the list of standard question-answer pairs to obtain a plurality of standard question-answer pairs that meet similarity conditions.

[0046] In the embodiment of the present invention, the standard question-answer pair can be specifically understood as: pre-defined or verified questions in the standard question-answer pair list, i.e., standard questions, and their corresponding correct answers. The standard question-answer pair can be a complete question and its corresponding answer, or the semantic features of these question-answer pairs, presented in the form of text vectorization. The standard question-answer pair can be used as a reference standard so that when a user asks a question, the platform can quickly match and provide an accurate answer.

[0047] The similarity condition can be specifically understood as: in the matching process, the standard used to determine whether the target user question represented by the target semantic feature is similar or identical to the standard question in the list of standard question-answer pairs. For example, a similarity threshold can be set. When the similarity exceeds this threshold, the two questions are considered to be similar. Specifically, a similarity calculation algorithm, such as cosine similarity or Euclidean distance, is used to calculate the similarity between the target semantic feature and the semantic feature vectors of each standard question, and all questions and answers whose similarity exceeds the preset similarity threshold are obtained as standard question-answer pairs. If the calculated similarities are all less than or equal to the preset similarity threshold, manual processing is performed and the corresponding answer is output.

[0048] S140: Combine the standard answers in the multiple standard question-answer pairs to obtain a comprehensive answer, and use the comprehensive answer as a feedback result for the target user's question.

[0049] In an embodiment of the present invention, the comprehensive answer can be specifically understood as: an answer obtained by summarizing, organizing and optimizing the standard answers in multiple standard question and answer pairs. The answer not only contains the information in the original standard answer, but also can provide a deeper explanation or more details by integrating and analyzing this information.

[0050] Generally speaking, the standard answers in multiple standard question-and-answer pairs can be sorted according to their similarity. According to the question-and-answer requirements, the number of answers to be displayed can be determined, for example, 5. The first 5 standard answers with higher similarity can be selected as comprehensive answers, and the results can be fed back to the target users.

[0051] Specifically, for each standard answer semantic feature vector, an aggregation method (such as average pooling or weighted average, etc.) can be used to integrate the semantic information of multiple answers, and the similarity between the aggregated vector and each standard answer vector can be calculated to determine which answers contribute more to the comprehensive answer. Based on the similarity results, key information can be extracted, and the amount of key information can be determined based on the answer combination requirements. For example, the answer semantic features corresponding to the top 5 higher similarity results can be extracted as key information, and the key information can be filled into the predefined standard answer template corresponding to the standard question to construct a personalized answer as a comprehensive answer, and the result can be fed back to the target user.

[0052] The technical solution of the embodiment of the present invention receives the target user's question through the target big data platform, and extracts the semantic features of the target user's question to obtain the target semantic features; locates the target first-level classification label that matches the target big data platform in the question and answer library, and obtains the list of standard question and answer pairs under all the second-level classification labels corresponding to the target first-level classification label; matches the target semantic features with each standard question in the list of standard question and answer pairs to obtain multiple standard question and answer pairs that meet the similarity conditions; combines the standard answers in the multiple standard question and answer pairs to obtain a comprehensive answer, and uses the comprehensive answer as the feedback result for the target user's question. The automated question and answer method of the big data platform improves the response efficiency of the platform and reduces the labor cost. The platform can generate a comprehensive answer by intelligently combining standard answers, providing users with more comprehensive and accurate information support, and optimizing the user experience.

[0053] Optionally, based on the above embodiments, the standard answers in the multiple standard question-answer pairs are combined to obtain a comprehensive answer, including:

[0054] Extracting key description information from each standard answer in the plurality of standard question-answer pairs;

[0055] The key description information and the target semantic features are input into a pre-trained answer generation model to obtain a comprehensive answer.

[0056] In the embodiment of the present invention, the key description information can be specifically understood as: the core information extracted from multiple standard question-answer pairs, representing the main content of each answer, such as proper nouns, professional terms, and important operation processes or steps. Specifically, the answers in the question-answer pairs can be preprocessed by removing stop words, punctuation marks, and stem extraction, and a text vectorization algorithm, such as the TF-IDF (Term Frequency-Inverse Document Frequency) algorithm, is used to vectorize the preprocessed text information as candidate keywords, evaluate the importance of each candidate word and sort them, and sort them according to the TF-IDF value, the higher the TF-IDF value, the more important it is, or sort them according to the frequency of occurrence of each candidate word in each standard question-answer pair, the higher the frequency, the more important it is, and set the number of required keywords according to actual needs, for example, 5, and select the top 5 words with higher TF-IDF values ​​or higher frequencies as key description information.

[0057] The pre-trained answer generation model can be specifically understood as: based on the question and answer pair data corresponding to each secondary classification label in the question and answer library, the trained model can understand the semantic features of the question and generate the corresponding answer. Generally speaking, the pre-training method of the model can be: pre-processing and vectorizing the questions and answers corresponding to each secondary classification label in the question and answer library, and inputting them into a preset model, such as the BERT (Bidirectional Encoder Representations from Transformers, a bidirectional encoder based on Transformers) model, using the cross entropy loss function to calculate the difference between the predicted answer and the true answer, and using the back propagation algorithm to update the model weights to minimize the loss function. After the training is completed, save the model as a pre-trained answer generation model. After the model is deployed, personalized answers can be generated based on key descriptive information and target semantic features to obtain comprehensive answers.

[0058] In a specific example, the BERT model can predict the start and end positions of a text segment that contains key descriptive information and target semantic features. When processing each word in the text, the model inputs the final output semantic feature vector into a start word classifier. The classifier can use a softmax (normalized exponential function) activation function to generate a probability distribution for all candidate words, thereby identifying the word with the highest probability as the start word of the segment. Subsequently, the identified keywords are inserted into the corresponding text segment to construct a complete answer. The big data question and answer platform combines the features of existing answers and the semantic features of the input questions to generate personalized answers, which not only optimizes the use of existing question and answer resources, improves the relevance and accuracy of answers, and enhances user experience, but also improves the flexibility and intelligence level of the platform's question and answer, reduces the repetitiveness of output content, and makes user interaction more natural and smooth.

[0059] Further, based on the above embodiments, after receiving the target user's question through the target big data platform, it also includes:

[0060] If the target primary classification label matching the target big data platform is not located in the question and answer database, after the target primary classification label is established in the question and answer database, the target user question is sent to the manual processing platform, and the target answer returned by the manual processing platform is used as the feedback result for the target user question;

[0061] Establishing a new standard question-answer pair according to the target answer, and establishing a secondary classification label matching the standard question-answer pair;

[0062] Add the established secondary classification labels and the matching standard question-answer pairs to the target primary classification labels;

[0063] and / or

[0064] Further, based on the above embodiments, after the comprehensive answer is used as the feedback result of the target user's question, the method further includes:

[0065] Obtaining satisfaction feedback from the party asking the target user's question on the comprehensive answer;

[0066] If the satisfaction feedback is negative feedback, the target user question is sent to the manual processing platform, and the target answer returned by the manual processing platform is used as the feedback result for the target user question;

[0067] According to the target answer, the list of standard question-answer pairs under all secondary classification labels under the target primary classification label in the question-answer library and the preset similarity threshold are updated.

[0068] Specifically, when there is no target first-level classification label matching the target user's question in the question-answering database, it means that there is no preset answer to display for this type of question. It is necessary to create a target first-level classification label representing the topic or field corresponding to the question. The corresponding section topic in the target big data platform can be used as the newly created target first-level classification label. Then, the target user's question is sent to the manual processing platform, and manual intervention is performed to process the target user's questions that the machine cannot automatically answer, and the newly generated target answer is fed back to the user. The currently processed target user question and the corresponding target answer form a new standard question-answer pair.

[0069] In order to organize and manage questions more meticulously, it is also necessary to set secondary classification labels under the newly created target primary classification labels. These secondary labels can be specific subtopics of the questions or classifications of related questions. They can be determined based on the semantic features of the newly generated standard question-answer pairs and the target primary classification labels. For example, if the primary classification label is "health", and the semantic features of the standard question-answer pairs include: "diet", "exercise" or "psychology", etc., then the secondary classification labels can be "dietary health", "exercise health" or "psychology", etc. Then add the secondary classification labels and the matching standard question-answer pairs to the target primary classification labels.

[0070] Specifically, in order to continuously improve the efficiency and user satisfaction of the question-answering system, after feeding back the results to the user, the platform will collect user feedback. When the user gives negative feedback to the automatically generated answer, such as feedback of dissatisfaction or unresolved problems, the target user question will be sent to the manual processing platform, and the target user questions that cannot be automatically answered by the machine will be processed manually, and the newly generated target answer will be fed back to the user. According to the manually processed target answer, the standard question-answer pair list of all secondary classification labels under the corresponding first-level classification label in the question-answering library is updated, where the target answer can correspond to one or more secondary classification labels to ensure that more accurate answers can be automatically provided in the future, and the corresponding preset similarity thresholds under all secondary classification labels under the target first-level classification label in the question-answering library are updated. In addition, the corresponding number of received user negative feedback can also be updated. When the user's negative feedback accumulates to a certain number, it indicates that the current threshold setting may not be reasonable enough, and it needs to be adjusted based on the preset similarity threshold to optimize the relevance and accuracy of the answer.

[0071] By introducing new primary and secondary classification labels, the organization of the Q&A database is improved, the information content of the Q&A database is improved, and the subsequent platform Q&A management and question retrieval are facilitated. The standard Q&A pairs obtained by manual processing are updated to the Q&A database in an extensible way, so that the Q&A database can flexibly adapt to the emergence of new questions and meet the ever-changing information needs, which promotes the continuous learning and automatic answering capabilities of the Q&A database, provides users with more comprehensive and accurate information support, and optimizes the user experience. The accumulation of negative feedback and the optimization of thresholds promote the continuous improvement of the Q&A database and provide better Q&A services. In addition, the optimized threshold helps the big data platform provide accurate answers faster and improves the platform's response efficiency.

[0072] Embodiment 2

[0073] Figure 2 A flowchart of another question-and-answer method for a big data platform provided in Example 2 of the present invention. This embodiment is a refinement of the operation of matching the target semantic features with each standard question in the list of standard question-and-answer pairs in the above embodiment to obtain multiple standard question-and-answer pairs that meet the similarity conditions, specifically including: respectively calculating the similarity between the target semantic features and the semantic features of each standard question in the list of standard question-and-answer pairs; obtaining multiple standard question-and-answer pairs whose similarities are within a preset similarity threshold; wherein the similarity threshold is dynamically generated based on the satisfaction of multiple questioners with the answers to the feedback questions.

[0074] Correspondingly, such as Figure 2 As shown, the method includes:

[0075] S210, receiving target user questions through a target big data platform, and extracting semantic features of the target user questions to obtain target semantic features.

[0076] S220. Locate the target primary classification label that matches the target big data platform in the question and answer library, and obtain a list of standard question and answer pairs under all secondary classification labels corresponding to the target primary classification label.

[0077] S230, respectively calculating the similarity between the target semantic feature and the semantic feature of each standard question in the list of standard question-answer pairs.

[0078] Specifically, each standard question in the standard question-answer pair list is converted into a semantic feature vector, for example, a text vectorization result is obtained through a pre-trained BERT model, and a similarity is calculated between the semantic feature vector of the target user question and the semantic feature vector of each standard question, such as by calculating the dot product and modulus of the two vectors to measure the cosine similarity between them.

[0079] S240: Acquire a plurality of standard question-answer pairs whose similarities are within a preset similarity threshold.

[0080] The similarity threshold is dynamically generated according to the satisfaction of multiple questioners with the answers to the fed-back questions.

[0081] Specifically, the standard question-answer pairs are sorted according to the calculated similarity values, and the results are screened according to the preset similarity threshold. Only when the calculated similarity exceeds this threshold, the standard question is considered similar enough to the target user's question, and the standard question and the corresponding answer can be used as the standard question-answer pair. Among them, the similarity threshold is related to the user's satisfaction with the answer to the question. That is, if the user is not satisfied with the result, it is regarded as negative feedback, which may indicate that the threshold setting is unreasonable. The threshold needs to be dynamically optimized according to the accumulation of negative feedback and the answer coverage corresponding to different thresholds.

[0082] In a specific example, an initial threshold is first set as the starting point for measuring the similarity of questions, such as 20%, that is, the top 20% of the answers in the standard question and answer pairs are selected as the standard question and answer pairs. The threshold is dynamically adjusted according to the user's satisfaction feedback. If one or more users repeatedly report that they are dissatisfied with the answer, when the set number threshold is reached, the similarity threshold needs to be dynamically updated. The user is dissatisfied with the answer, which may be because the threshold is set too small, resulting in the relevant answers expected by the user not being covered. At this time, the threshold can be appropriately expanded, such as adjusted to 30%, to increase the answer coverage. On the contrary, if the user continues to express satisfaction with the answer and reaches the set number threshold, it means that the current threshold setting can better cover the user's needs. The threshold range can be appropriately narrowed, such as adjusted to 10%, to reduce the answer coverage, so as to streamline the output and improve the relevance of the answer.

[0083] By calculating the cumulative distribution function (CDF), we analyze the coverage of answers at different similarity thresholds. The coverage is equal to the cumulative frequency of each similarity score divided by the total number of samples. Based on the results of the CDF analysis, we determine the similarity threshold corresponding to the expected coverage. We select a threshold that ensures both answer coverage and answer quality to balance the relevance and quantity of answers.

[0084] S250: combining the standard answers in the plurality of standard question-answer pairs to obtain a comprehensive answer, and using the comprehensive answer as a feedback result for the target user's question.

[0085] The technical solution of the embodiment of the present invention receives the target user question through the target big data platform, extracts the semantic features of the target user question, and obtains the target semantic features; locates the target first-level classification label matching the target big data platform in the question and answer library, and obtains the standard question and answer pair list under all the second-level classification labels corresponding to the target first-level classification label; respectively calculates the similarity between the target semantic feature and the semantic features of each standard question in the standard question and answer pair list; obtains multiple standard question and answer pairs whose similarity is within a preset similarity threshold; combines the standard answers in the multiple standard question and answer pairs to obtain a comprehensive answer, and uses the comprehensive answer as the feedback result for the target user question. The automated question and answer method of the big data platform is used to improve the response efficiency of the platform and reduce the labor cost. The platform can generate a comprehensive answer by intelligently combining the standard answers to provide users with more comprehensive and more accurate information support. Through this method of dynamically adjusting the similarity threshold, it can ensure that the big data platform can meet the personalized needs of users when providing answers, while maintaining high efficiency and accuracy. This method makes the question and answer system more flexible and intelligent, and can self-adjust and optimize according to actual usage, thereby providing a better user experience.

[0086] Optionally, on the basis of the above embodiments, the target semantic feature is matched with each standard question in the list of standard question-answer pairs to obtain multiple standard question-answer pairs that meet the similarity condition, including:

[0087] If it is determined that the similarity threshold between the target semantic feature and any standard question does not exceed the preset threshold, then obtaining the questioner identification of the target user question;

[0088] Querying historical user questions that match the questioner identifier, and obtaining historical standard questions that have positive feedback satisfaction with the historical user questions;

[0089] Generating a question correction solution matching the questioner identifier according to the grammatical differences between the target user's question and the historical standard question;

[0090] After the target user question is corrected according to the question correction plan, a new target semantic feature is recalculated, and the target semantic feature is matched with each standard question in the standard question-answer pair list again.

[0091] In the embodiment of the present invention, the questioner identifier can be specifically understood as: a unique identifier that can be used to identify the user who asks the question, which can be a user ID (identification), user name, email address or any other information that can uniquely determine the user's identity, and can be obtained from the data of the question submitted by the user to the platform.

[0092] Historical user questions can be specifically understood as records of questions asked by the user in the past. These records are stored in a question-and-answer library or database and can be used to analyze the user's questioning habits or common question types.

[0093] Historical standard questions can be specifically understood as questions in standard question-answer pairs that match historical user questions and are marked as positive feedback such as user satisfaction or solved.

[0094] Grammatical differences can be specifically understood as differences in grammatical structure between the questions raised by users and existing historical standard questions, which may include sentence patterns, word order, tense or grammatical errors, etc. By analyzing these differences, the system can understand the unique way of expressing the user's question.

[0095] The question correction scheme can be specifically understood as: based on the grammatical differences between the target user's question and the historical standard question, the generated suggestions or steps for adjusting the target user's question so that the user's question is closer to the form of the standard question, thereby improving the matching accuracy and the possibility of obtaining a satisfactory answer.

[0096] Specifically, if it is determined that the similarity threshold between the target semantic feature and any standard question does not exceed the preset threshold, it means that the preset answer in the question and answer library cannot answer the target user question. It may be because the user's question is too simple or the wording is incorrect, resulting in a matching failure. Therefore, additional contextual information can be provided for the target user question to supplement the necessary information and improve the target user question to improve the matching success rate. Specifically, according to the identification of the questioner, historical question records are obtained and historical standard questions that the user has given positive feedback are screened out. Based on grammatical differences, the user's question is corrected and adjusted to improve the matching degree, and its semantic features are recalculated to match the questions in the standard question and answer list to find a more accurate answer. By utilizing the user's historical data and feedback, the system can adapt to the user's questioning habits, which helps the big data platform to more accurately understand the user's questioning intentions and provide more accurate answers, thereby improving the platform's question and answer response efficiency and user satisfaction.

[0097] Optionally, on the basis of the above embodiments, the target semantic feature is matched with each standard question in the list of standard question-answer pairs to obtain multiple standard question-answer pairs that meet the similarity condition, including:

[0098] If it is determined that the similarity threshold between the target semantic feature and any standard question does not exceed the preset threshold, then obtaining the common grammatical features of each standard question in the list of standard question-answer pairs;

[0099] After the target user question is corrected according to the common grammatical features, new target semantic features are recalculated, and the target semantic features are matched with each standard question in the list of standard question-answer pairs again.

[0100] In the embodiment of the present invention, the common grammatical features can be specifically understood as: grammatical patterns or structural features that are prevalent in a set of standard questions, which may include commonly used part-of-speech patterns, sentence structures, specific phrases or vocabulary collocations, etc. For example, if most of the questions in the standard question list follow the structure of "how [verb] [noun]", then this structure is a common grammatical feature. Similarly, if many questions contain specific proprietary words, such as technical terms or business types, these proprietary words can also be regarded as common grammatical features.

[0101] Specifically, when the similarity between the target question and any standard question does not exceed the preset threshold, it means that the target user's question needs to be corrected or supplemented with contextual information for improvement. When the target user has no historical questions, all standard questions in the standard question and answer pair list can be analyzed to extract their common grammatical features. Based on the extracted common grammatical features, the target user's question is corrected to make it closer to the features of the standard question in terms of grammatical structure, thereby increasing the possibility of matching. The corrected question will be reanalyzed to calculate new target semantic features and matched with the questions in the standard question and answer pair list. By constructing common grammatical features and correcting the target user's question for matching answers, the matching accuracy can be improved, user satisfaction can be enhanced, trends and patterns in the user group can be identified, and data support can be provided for the improvement of the question and answer library, thereby helping to improve the platform's question and answer response efficiency and enhance adaptability.

[0102] Specific application scenarios

[0103] With the continuous development and application of big data technology, big data platforms have become an important tool for business processing and analysis of massive data. However, in the process of using big data platforms, users often encounter various business problems, operational problems, and incompatible query results, such as data import failure, slow query speed, and poor data visualization. For these problems, users usually need to provide feedback to platform staff or technical support personnel by submitting work orders or emails, etc., which has problems such as long feedback cycles and low problem-solving efficiency, causing inconvenience to users. Secondly, in the process of problem handling, there are often a large number of basic and repetitive question responses, which not only wastes a lot of human resources, but also reduces user problem-solving efficiency. In order to solve the above problems, an embodiment of the present invention proposes a question-and-answer method for a big data platform. Figure 3 is a schematic diagram of a module composition applicable to an embodiment of the present invention, such as Figure 3As shown, the functional modules of the method may include: a question receiving module, a classification control module, an intelligent reply module and a manual intervention module.

[0104] The question receiving module is responsible for receiving questions from users on the big data platform and manually classifying them into primary categories. The primary categories can be divided according to the fields of each secondary sub-platform. The classification control module uses natural language processing technology to classify the received questions. The intelligent reply module selects questions with high similarity from the existing question and answer library for automatic reply. The manual intervention module submits professional replies to complex questions.

[0105] Figure 4 is a flow chart of a question receiving module applicable to an embodiment of the present invention, such as Figure 4 As shown, the technical solution of the question receiving module may include:

[0106] 1. Post a question and specify the first-level category of the question

[0107] Business personnel submit questions on the big data support platform and select the field to which the questions belong for primary classification.

[0108] 2. Determine whether it matches a known category

[0109] Determine whether the business selected domain exists in the preset system of the big data platform. If so, proceed to step 4; otherwise, proceed to step 3.

[0110] 3. Add a new first-level category

[0111] Added a new first-level classification and updated the preset system.

[0112] 4. Problem Preprocessing

[0113] Perform preprocessing on the problem, such as semantic refinement and feature extraction.

[0114] 5. Submit the problem classification module for processing

[0115] Figure 5 is a flow chart of a classification control module applicable to an embodiment of the present invention, such as Figure 5 As shown, the technical solution of the classification control module may include:

[0116] 1. Receive problem information

[0117] Receive the problem information submitted by the problem receiving module.

[0118] 2. Extracting semantic features

[0119] Natural language processing technology is used to further extract semantic features.

[0120] 3. Generate secondary classification labels

[0121] Generate secondary classification labels based on preset secondary business types. Among them, secondary business types are specific subdivisions under primary classification labels. Primary classification labels represent a wide range of business areas, such as financial services, educational services, or customer support. Under each primary classification label, define specific secondary business types, corresponding to specific business processes or services. For example, financial services may include account management, loan services, deposit products, credit card services, and investment consulting; educational services may include online courses, test preparation, vocational training, academic research, and educational consulting; customer support may be classified according to the technical types involved in the problem, including the set distributed system infrastructure, batch processing technology, MPP clusters, Gaussian distribution, and set open source columnar databases. Correspondingly, when the primary business type is customer support, the secondary classification label can be generated based on the preset problem error category and other information to determine the technical type.

[0122] 4. Matching question and answer database

[0123] Determine whether the generated secondary classification label exists in the preset system of the big data platform. If so, transfer it to the intelligent reply module; otherwise, proceed to step 5.

[0124] 5. Add secondary tags

[0125] Add secondary classification labels and update the preset system, and then transfer to the manual intervention module.

[0126] Figure 6 is a flow chart of an intelligent reply module applicable to an embodiment of the present invention, such as Figure 6 As shown, the technical solution of the intelligent reply module may include:

[0127] 1. Receive problem information

[0128] Receive problem information submitted by the classification control module.

[0129] 2. Calculation of question similarity under the same label

[0130] Traverse the list of questions under the same tag and calculate the similarity one by one.

[0131] 3. Return to historical answers

[0132] Sort the question lists under the same tag in the question and answer database according to the similarity, and return the top 5 similarity question and answer lists for application reference. The number of answers returned can be set according to usage requirements.

[0133] 4. Apply the answers

[0134] The application provides feedback on satisfaction, whether the problem has been resolved, etc. based on the above questions and answers.

[0135] 5. Judgment Feedback

[0136] If the feedback shows that the problem has been solved, the work order is closed; otherwise, it is transferred to the manual intervention module.

[0137] Figure 7 is a flow chart of a manual intervention module applicable to an embodiment of the present invention, such as Figure 7 As shown, the technical solution of the manual intervention module may include:

[0138] 1. Reception issues

[0139] Receive problem information submitted from the previous module.

[0140] 2. Manual reply

[0141] Notify professional support personnel via email or other means to answer your questions.

[0142] 3. Update the Q&A database

[0143] Update the manually answered question and answer information to the question and answer database.

[0144] A question-and-answer method for a big data platform proposed in an embodiment of the present invention solves the problem of answering a large number of basic and repeated questions in a big data support platform through intelligent means, can realize self-feedback and answering of questions, improves the response efficiency of the platform, reduces labor costs, and optimizes user experience.

[0145] Embodiment 3

[0146] Figure 8 This is a schematic diagram of the structure of a question-answering device for a big data platform provided in Embodiment 3 of the present invention. Figure 8 As shown, the device includes: a feature extraction module 810, a list acquisition module 820, a question-answer pair acquisition module 830 and an answer acquisition module 840, wherein:

[0147] The feature extraction module 810 is used to receive the target user's question through the target big data platform, and extract the semantic features of the target user's question to obtain the target semantic features;

[0148] The list acquisition module 820 is used to locate the target primary classification label that matches the target big data platform in the question and answer library, and obtain the list of standard question and answer pairs under all secondary classification labels corresponding to the target primary classification label;

[0149] A question-answer pair acquisition module 830 is used to match the target semantic feature with each standard question in the standard question-answer pair list to obtain a plurality of standard question-answer pairs that meet the similarity condition;

[0150] The answer acquisition module 840 is used to combine the standard answers in the multiple standard question-answer pairs to obtain a comprehensive answer, and use the comprehensive answer as a feedback result for the target user's question.

[0151] The technical solution of the embodiment of the present invention receives the target user's question through the target big data platform, and extracts the semantic features of the target user's question to obtain the target semantic features; locates the target first-level classification label that matches the target big data platform in the question and answer library, and obtains the list of standard question and answer pairs under all the second-level classification labels corresponding to the target first-level classification label; matches the target semantic features with each standard question in the list of standard question and answer pairs to obtain multiple standard question and answer pairs that meet the similarity conditions; combines the standard answers in the multiple standard question and answer pairs to obtain a comprehensive answer, and uses the comprehensive answer as the feedback result for the target user's question. The automated question and answer method of the big data platform improves the response efficiency of the platform and reduces the labor cost. The platform can generate a comprehensive answer by intelligently combining standard answers, providing users with more comprehensive and accurate information support, and optimizing the user experience.

[0152] Based on the above embodiments, the question-answer pair acquisition module 830 is specifically used to:

[0153] Calculating the similarity between the target semantic feature and the semantic feature of each standard question in the list of standard question-answer pairs respectively;

[0154] Acquire multiple standard question-answer pairs whose similarities are within a preset similarity threshold;

[0155] The similarity threshold is dynamically generated according to the satisfaction of multiple questioners with the answers to the fed-back questions.

[0156] Further, based on the above embodiments, the question-answer pair acquisition module 830 can also be specifically used for:

[0157] If it is determined that the similarity threshold between the target semantic feature and any standard question does not exceed the preset threshold, then obtaining the questioner identification of the target user question;

[0158] Querying historical user questions that match the questioner identifier, and obtaining historical standard questions that have positive feedback satisfaction with the historical user questions;

[0159] Generating a question correction solution matching the questioner identifier according to the grammatical differences between the historical user questions and the historical standard questions;

[0160] After the target user question is corrected according to the question correction plan, a new target semantic feature is recalculated, and the target semantic feature is matched with each standard question in the standard question-answer pair list again.

[0161] Further, based on the above embodiments, the question-answer pair acquisition module 830 can also be specifically used for:

[0162] If it is determined that the similarity threshold between the target semantic feature and any standard question does not exceed the preset threshold, then obtaining the common grammatical features of each standard question in the list of standard question-answer pairs;

[0163] After the target user question is corrected according to the common grammatical features, new target semantic features are recalculated, and the target semantic features are matched with each standard question in the list of standard question-answer pairs again.

[0164] Based on the above embodiments, the answer acquisition module 840 is specifically used for:

[0165] Extracting key description information from each standard answer in the plurality of standard question-answer pairs;

[0166] The key description information and the target semantic features are input into a pre-trained answer generation model to obtain a comprehensive answer.

[0167] Optionally, based on the above embodiments, the feature extraction module 810 may include: a first label establishment unit, a second label establishment unit and a question-answer pair adding unit, wherein:

[0168] A first label establishment unit is used for, after receiving a target user question through a target big data platform, if a target primary classification label matching the target big data platform is not located in the question-and-answer database, to establish a target primary classification label in the question-and-answer database, send the target user question to the manual processing platform, and use the target answer returned by the manual processing platform as a feedback result for the target user question;

[0169] A second label establishment unit, used to establish a new standard question-answer pair according to the target answer, and establish a secondary classification label matching the standard question-answer pair;

[0170] A question-answer pair adding unit, used to add the established secondary classification label and the matching standard question-answer pair to the target primary classification label;

[0171] and / or

[0172] Optionally, based on the above embodiments, the answer acquisition module 840 may include: a feedback acquisition unit, a negative processing unit and a question and answer updating unit.

[0173] A feedback acquisition unit, configured to acquire satisfaction feedback of the questioner of the target user's question on the comprehensive answer after taking the comprehensive answer as the feedback result on the target user's question;

[0174] A negative processing unit, configured to send the target user's question to a manual processing platform if the satisfaction feedback is negative feedback, and use the target answer returned by the manual processing platform as a feedback result for the target user's question;

[0175] The question and answer updating unit is used to update the list of standard question and answer pairs under all secondary classification labels under the target primary classification label in the question and answer library and the preset similarity threshold according to the target answer.

[0176] The question-and-answer device for a big data platform provided in an embodiment of the present invention can execute the question-and-answer method for a big data platform provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0177] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0178] Embodiment 4

[0179] Fig. 9 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0180] like Fig. 9As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0181] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0182] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 11 executes the various methods and processes described above, such as the question-answering method of the big data platform, that is:

[0183] Receive the target user's questions through the target big data platform, extract the semantic features of the target user's questions, and obtain the target semantic features;

[0184] Locate the target first-level classification label that matches the target big data platform in the question-answer library, and obtain a list of standard question-answer pairs under all second-level classification labels corresponding to the target first-level classification label;

[0185] Matching the target semantic feature with each standard question in the list of standard question-answer pairs to obtain a plurality of standard question-answer pairs that meet similarity conditions;

[0186] The standard answers in the multiple standard question-answer pairs are combined to obtain a comprehensive answer, and the comprehensive answer is used as a feedback result for the target user's question.

[0187] In some embodiments, the question-answering method of the big data platform may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the question-answering method of the big data platform described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the question-answering method of the big data platform in any other appropriate manner (e.g., by means of firmware).

[0188] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0189] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0190] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, 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 of the foregoing.

[0191] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0192] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0193] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0194] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0195] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A question-answering method for a big data platform, characterized in that: include: Receive the target user's questions through the target big data platform, extract the semantic features of the target user's questions, and obtain the target semantic features; Locate the target first-level classification label that matches the target big data platform in the question-answer library, and obtain a list of standard question-answer pairs under all second-level classification labels corresponding to the target first-level classification label; Matching the target semantic feature with each standard question in the list of standard question-answer pairs to obtain a plurality of standard question-answer pairs that meet similarity conditions; The standard answers in the multiple standard question-answer pairs are combined to obtain a comprehensive answer, and the comprehensive answer is used as a feedback result for the target user's question.

2. The method according to claim 1, characterized in that Matching the target semantic feature with each standard question in the standard question-answer pair list to obtain multiple standard question-answer pairs that meet similarity conditions, including: Calculating the similarity between the target semantic feature and the semantic feature of each standard question in the list of standard question-answer pairs respectively; Acquire multiple standard question-answer pairs whose similarities are within a preset similarity threshold; The similarity threshold is dynamically generated according to the satisfaction of multiple questioners with the answers to the fed-back questions.

3. The method according to claim 2, characterized in that Matching the target semantic feature with each standard question in the standard question-answer pair list to obtain multiple standard question-answer pairs that meet similarity conditions, including: If it is determined that the similarity threshold between the target semantic feature and any standard question does not exceed the preset threshold, then obtaining the questioner identification of the target user question; Querying historical user questions that match the questioner identifier, and obtaining historical standard questions that have positive feedback satisfaction with the historical user questions; Generating a question correction solution matching the questioner identifier according to the grammatical differences between the historical user questions and the historical standard questions; After the target user question is corrected according to the question correction plan, a new target semantic feature is recalculated, and the target semantic feature is matched with each standard question in the standard question-answer pair list again.

4. The method according to claim 2, characterized in that: Matching the target semantic feature with each standard question in the standard question-answer pair list to obtain multiple standard question-answer pairs that meet similarity conditions, including: If it is determined that the similarity threshold between the target semantic feature and any standard question does not exceed the preset threshold, then obtaining the common grammatical features of each standard question in the list of standard question-answer pairs; After the target user question is corrected according to the common grammatical features, new target semantic features are recalculated, and the target semantic features are matched with each standard question in the list of standard question-answer pairs again.

5. The method according to claim 1, characterized in that The standard answers in the plurality of standard question-answer pairs are combined to obtain a comprehensive answer, including: Extracting key description information from each standard answer in the plurality of standard question-answer pairs; The key description information and the target semantic features are input into a pre-trained answer generation model to obtain a comprehensive answer.

6. The method according to any one of claims 2 to 5, characterized in that: After receiving the target user questions through the target big data platform, it also includes: If the target primary classification label matching the target big data platform is not located in the question and answer database, after the target primary classification label is established in the question and answer database, the target user question is sent to the manual processing platform, and the target answer returned by the manual processing platform is used as the feedback result for the target user question; Establishing a new standard question-answer pair according to the target answer, and establishing a secondary classification label matching the standard question-answer pair; Add the established secondary classification labels and the matching standard question-answer pairs to the target primary classification labels; and / or After the comprehensive answer is used as the feedback result to the target user's question, the method further includes: Obtaining satisfaction feedback from the party asking the target user's question on the comprehensive answer; If the satisfaction feedback is negative feedback, the target user question is sent to the manual processing platform, and the target answer returned by the manual processing platform is used as the feedback result for the target user question; According to the target answer, the list of standard question-answer pairs under all secondary classification labels under the target primary classification label in the question-answer library and the preset similarity threshold are updated.

7. A question-answering device for a big data platform, characterized in that: include: A feature extraction module is used to receive the target user's questions through the target big data platform, and extract the semantic features of the target user's questions to obtain the target semantic features; A list acquisition module is used to locate the target first-level classification label that matches the target big data platform in the question and answer library, and obtain the list of standard question and answer pairs under all second-level classification labels corresponding to the target first-level classification label; A question-answer pair acquisition module, used to match the target semantic feature with each standard question in the standard question-answer pair list to obtain multiple standard question-answer pairs that meet the similarity condition; The answer acquisition module is used to combine the standard answers in the multiple standard question-answer pairs to obtain a comprehensive answer, and use the comprehensive answer as the feedback result for the target user's question.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the question-answering method of the big data platform described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the question-answering method for a big data platform according to any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the question-answering method for a big data platform according to any one of claims 1-6.

Images