Enterprise knowledge question answering method and device, computer equipment and storage medium

Abstract

The invention relates to an enterprise knowledge question answering method and device, computer equipment and a storage medium. The method comprises the following steps: in response to obtained enterprise knowledge question content, vectorizing the enterprise knowledge question content to obtain an enterprise knowledge question vector; according to the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set, determining a target sub-segment matched with the enterprise knowledge question content from the sub-segment set; determining a target parent segment associated with the target child segment from parent segments associated with the child segments in the child segment set; and according to the enterprise knowledge question content, the target child segment and the target parent segment, generating and outputting enterprise knowledge answer content for the enterprise knowledge question content. By adopting the method, the query efficiency of enterprise knowledge can be improved.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a method, apparatus, computer equipment, and storage medium for answering enterprise knowledge questions. Background Technology

[0002] In the current era of accelerated digital transformation, enterprises have accumulated massive amounts of knowledge resources over their long-term operations and development, which are an important component of their core competitiveness. However, how to quickly and accurately extract the necessary knowledge content from these vast enterprise knowledge resources has become a critical issue that urgently needs to be addressed in the field of enterprise knowledge management.

[0003] In related technologies, enterprise knowledge content is often obtained through the enterprise's office automation system. For example, the office automation system performs queries based on keyword matching retrieval technology. When a user enters query keywords, the office automation system performs knowledge matching in the stored knowledge base and returns enterprise knowledge content containing the keywords.

[0004] However, this keyword-based method of acquiring enterprise knowledge content in related technologies often retrieves a large amount of knowledge content, which then requires filtering out the necessary knowledge content, resulting in low efficiency in enterprise knowledge retrieval. Summary of the Invention

[0005] Based on this, this application provides a method, apparatus, computer equipment, and storage medium for enterprise knowledge question answering, which can improve the efficiency of enterprise knowledge retrieval.

[0006] Firstly, this application provides a method for enterprise knowledge question answering, which is applied to an enterprise knowledge question answering system. The method includes:

[0007] In response to the acquired enterprise knowledge questions, the enterprise knowledge questions are vectorized to obtain enterprise knowledge question vectors.

[0008] Based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment set; the sub-segment set includes at least one sub-segment associated with each parent segment in the parent segment set, and at least one sub-segment associated with each parent segment is obtained by segmenting each enterprise knowledge text in the enterprise knowledge text library.

[0009] From the parent segments associated with each sub-segment in the sub-segment set, determine the target parent segment associated with the target sub-segment;

[0010] Based on the enterprise knowledge question content, target sub-segment, and target parent segment, generate and output enterprise knowledge answer content for the enterprise knowledge question content.

[0011] In some embodiments, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment set based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set, including:

[0012] Based on the similarity between the feature vectors of each subsegment in the subsegment set and the enterprise knowledge question vector, feature vectors of multiple candidate subsegments are determined.

[0013] The matching degree between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector is determined based on the similarity between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector, and based on the validity weight of each candidate sub-segment. The validity weight of each candidate sub-segment is included in the validity weight of each sub-segment in the sub-segment set. The validity weight of each sub-segment in the sub-segment set is determined based on the data source, publication time, and version number of each sub-segment in the sub-segment set.

[0014] Based on the matching degree between the feature vectors of each candidate subsegment and the enterprise knowledge question vector, the target subsegment matching the enterprise knowledge question content is determined from the subsegment set.

[0015] In some embodiments, the method further includes:

[0016] Obtain accuracy feedback information for the answers provided to enterprise knowledge-based questions;

[0017] Based on the accuracy feedback information, determine the current feedback accuracy of the target sub-segment;

[0018] The effectiveness weight of the target sub-segment is adjusted based on the current feedback accuracy of the target sub-segment, the current feedback provider's years of experience for the target sub-segment, the historical feedback accuracy of the target sub-segment, and the historical feedback providers' years of experience for the target sub-segment.

[0019] In some embodiments, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment set based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set, including:

[0020] The content of enterprise knowledge questions is classified by attributes to obtain the attribute categories of the enterprise knowledge question content;

[0021] When the attribute category of the enterprise knowledge question content includes the knowledge query category, the target sub-segment for matching the enterprise knowledge question content is determined from the sub-segment corresponding to the enterprise knowledge text under the knowledge query category, based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the enterprise knowledge text under the knowledge query category.

[0022] When the attribute category of the enterprise knowledge question content includes the business query category, the candidate enterprise knowledge texts that the target questioner can access are determined from the enterprise knowledge texts under the business query category, based on the permissions of the target questioner.

[0023] Based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the candidate enterprise knowledge text, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment corresponding to the candidate enterprise knowledge text.

[0024] In some embodiments, based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the candidate enterprise knowledge text, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment corresponding to the candidate enterprise knowledge text, including:

[0025] The content of enterprise knowledge questions is categorized to obtain the question knowledge categories;

[0026] Based on the segmentation category of the sub-segment corresponding to the candidate enterprise knowledge text, multiple undetermined sub-segments matching the question knowledge category are determined; wherein, the segmentation category of the sub-segment corresponding to the candidate enterprise knowledge text is included in the segmentation category of each sub-segment in the sub-segment set, and the segmentation category of each sub-segment in the sub-segment set is obtained by classifying each sub-segment associated with each parent segment according to at least one knowledge category of each parent segment in the parent segment set;

[0027] Based on the enterprise knowledge question vector and the feature vectors of multiple undetermined sub-segments matching the question knowledge category, the target sub-segment matching the enterprise knowledge question content is determined from the multiple undetermined sub-segments matching the question knowledge category.

[0028] In some embodiments, enterprise knowledge answer content is generated and output based on the enterprise knowledge question content, the target sub-segment, and the target parent segment, including:

[0029] Determine the target answering strategy based on the length of service of the target questioner in the enterprise knowledge question content;

[0030] Based on the target answering strategy, extract target knowledge content from the target sub-segments and the target parent segment;

[0031] Based on the enterprise knowledge questions and target knowledge content, generate and output enterprise knowledge answers for the enterprise knowledge questions.

[0032] In some embodiments, the method further includes:

[0033] Access the enterprise knowledge document library, the enterprise communication data content library, and the enterprise knowledge multimedia content library;

[0034] The images in the enterprise knowledge document library are converted into text descriptions to obtain the first knowledge text library;

[0035] Extract enterprise knowledge content related to enterprise knowledge from the enterprise communication data content library, and convert the multimedia content in the enterprise knowledge content into text description information to obtain a second knowledge text library;

[0036] The enterprise knowledge multimedia content in the enterprise knowledge multimedia content library is converted into text description information to obtain the third knowledge text library;

[0037] The first, second, and third knowledge text libraries are merged to obtain the enterprise knowledge text library.

[0038] Secondly, this application provides an enterprise knowledge question-answering device, which includes:

[0039] The vector acquisition module is used to vectorize the acquired enterprise knowledge question content in response to the enterprise knowledge question content, and obtain the enterprise knowledge question vector.

[0040] The matching module is used to determine the target sub-segment for matching the enterprise knowledge question content from the sub-segment set based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set. The sub-segment set includes at least one sub-segment associated with each parent segment in the parent segment set. The at least one sub-segment associated with each parent segment is obtained by segmenting each enterprise knowledge text in the enterprise knowledge text library.

[0041] The parent segment determination module is used to determine the target parent segment associated with the target sub-segment from the parent segments associated with each sub-segment in the sub-segment set;

[0042] The generation and output module is used to generate and output enterprise knowledge answers based on the enterprise knowledge question content, target sub-segments, and target parent segments.

[0043] Thirdly, this application provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods in the first aspect.

[0044] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method of any one of the first aspects.

[0045] In the technical solution provided by this application embodiment, the enterprise knowledge question content is first vectorized to obtain a question vector. Then, the target sub-segment is determined based on the matching of the question vector with the feature vectors of each sub-segment in the sub-segment set. Subsequently, the target parent segment is located. Finally, the enterprise knowledge answer content is generated by combining the question content, the target sub-segment, and the target parent segment. Compared with the keyword-based retrieval method in related technologies, which easily outputs a large amount of irrelevant knowledge content and requires users to filter it, this application embodiment, through the hierarchical knowledge division and vector matching mechanism of sub-segments and parent segments, can accurately lock the core knowledge fragments related to the question content, avoid redundant output of irrelevant knowledge, reduce the time spent by users in filtering, realize the rapid location and acquisition of enterprise knowledge, effectively solve the problem of low enterprise knowledge query efficiency in related technologies, and improve the query efficiency of enterprise knowledge. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 A flowchart illustrating an enterprise knowledge question-answering method provided in some embodiments;

[0048] Figure 2 A flowchart illustrating the method for determining target sub-segments provided in the first embodiment;

[0049] Figure 3 A flowchart illustrating the method for determining the target sub-segment provided in the second embodiment;

[0050] Figure 4 A flowchart illustrating a method for generating and outputting enterprise knowledge answer content as provided in some embodiments;

[0051] Figure 5 A flowchart illustrating a method for obtaining an enterprise knowledge text base as provided in some embodiments;

[0052] Figure 6 A schematic diagram of the structure of an enterprise knowledge question-and-answer device provided in some embodiments;

[0053] Figure 7 A schematic diagram of the structure of a computer device provided for some embodiments. Detailed Implementation

[0054] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0056] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, "multiple groups" means two or more, and "each" means each of the multiple, unless otherwise explicitly defined.

[0057] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0058] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0059] Unless otherwise specified, the order of execution steps in the embodiments of this application is not limited. It should also be noted that any step in the embodiments of this application can be executed independently, that is, the execution of any step in the above embodiments can be performed without depending on the execution of other steps.

[0060] To at least address the problem of low efficiency in enterprise knowledge retrieval in related technologies, this application provides an enterprise knowledge question-and-answer method, which can improve the efficiency of enterprise knowledge retrieval.

[0061] The method described in this application embodiment can be applied to an enterprise knowledge question-answering system, which can be deployed in a computer device. The computer device may include one or more of the following: a server, a mobile phone, a tablet computer, a computer with transceiver capabilities, a handheld computer, a desktop computer, a personal digital assistant, a portable media player, a smart speaker, a navigation device, a smartwatch, smart glasses, a smart necklace, and other wearable devices, a pedometer, a digital TV, a virtual reality (VR) device, an augmented reality (AR) device, devices in industrial control, devices in self-driving, devices in remote medical surgery, devices in a smart grid, devices in transportation safety, devices in a smart city, devices in a smart home, vehicles, in-vehicle devices, in-vehicle modules, etc.

[0062] Figure 1 A flowchart illustrating an enterprise knowledge question-answering method provided in some embodiments, such as Figure 1 As shown, the method includes the following steps:

[0063] S101. In response to the obtained enterprise knowledge question content, the enterprise knowledge question content is vectorized to obtain the enterprise knowledge question vector.

[0064] For example, an enterprise knowledge Q&A system can be deployed on a computer device in the form of an application or application component. When the enterprise knowledge Q&A system is triggered, its interface can be displayed. In some embodiments, the enterprise knowledge Q&A system interface may include a question input box; the user can enter a question in the question input box, so that the enterprise knowledge Q&A system determines the user's input question as enterprise knowledge Q&A content. In other embodiments, the enterprise knowledge Q&A system interface may include a question input box and a list of prompt statements. The user can select a target prompt statement from the prompt statement list and enter a question in the question input box, so that the enterprise knowledge Q&A system determines the user's input question and the target prompt statement as enterprise knowledge Q&A content. Different prompt statements in the prompt statement list represent prompt statements for different question types. For example, the prompt statement list includes at least one of the following: question statements for knowledge acquisition during product development, question statements for knowledge acquisition during product testing, and question statements for knowledge acquisition by various business departments. For example, question statements for knowledge acquisition by various business departments include question statements for database knowledge acquisition, question statements for communication knowledge acquisition, question statements for audio-visual knowledge, question statements for charging and discharging knowledge, etc.

[0065] In some embodiments, a preset vector dimension can be configured in the vector transformation model. The enterprise knowledge question content is vectorized by the vector transformation model to obtain an enterprise knowledge question vector with the preset vector dimension. Exemplarily, the vector transformation model can be any model in the related art that can convert text into vectors, and this application embodiment will not elaborate on this.

[0066] S102. Based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set, determine the target sub-segment that matches the enterprise knowledge question content from the sub-segment set.

[0067] The sub-segment set includes at least one sub-segment associated with each parent segment in the parent segment set. The at least one sub-segment associated with each parent segment is obtained by segmenting each enterprise knowledge text in the enterprise knowledge text library.

[0068] The set of subsegments can be pre-acquired. The set of subsegments includes multiple subsegments. For example, the feature vectors of the subsegments can be determined by vectorizing each subsegment using a vector transformation model to obtain the feature vectors of each subsegment.

[0069] In some embodiments, the similarity between the enterprise knowledge question vector and the feature vectors of each sub-segment in the sub-segment set can be determined; the sub-segments corresponding to similarities greater than a first similarity threshold are determined as target sub-segments for matching the enterprise knowledge question content. In other embodiments, the similarity between the enterprise knowledge question vector and the feature vectors of each sub-segment in the sub-segment set can be determined; the sub-segments corresponding to the first preset number of similarities with the highest similarity are determined as target sub-segments for matching the enterprise knowledge question content.

[0070] The enterprise knowledge text repository can include multiple enterprise knowledge texts. Each enterprise knowledge text includes electronic texts that embody various types of knowledge within the enterprise across all its activities, including production and operation, management and operation, research and development innovation, and service delivery.

[0071] In some embodiments, a parent-child segmentation model can be used to segment each enterprise knowledge text, resulting in at least one parent segment and at least one child segment associated with each parent segment. Thus, all the resulting parent segments are defined as a set of parent segments, and all the resulting child segments are defined as a set of child segments, with each parent segment in the set of parent segments associated with at least one child segment.

[0072] For example, the parent-child segmentation model can refer to any model in the related art that can identify parent segments and child segments, and the embodiments of this application will not elaborate on this.

[0073] The parent-child segmentation model is a hierarchical document processing strategy that divides a document into parent and child segments. The parent segment contains more characters than the child segments. A parent segment is a complete paragraph or a collection of related paragraphs, providing contextual information. Child segments are smaller blocks of text further subdivided from the parent segment and are the actual units used for retrieval.

[0074] In some embodiments, the maximum number of characters occupied by the parent segment can be a default value, and the maximum number of characters occupied by the child segment can be a default value.

[0075] In other embodiments, the maximum number of characters occupied by the parent segment and / or the maximum number of characters occupied by the child segment can be determined based on at least one of the following: the domain to which the knowledge belongs, the category to which the knowledge belongs, the importance of the knowledge, etc. The category to which the knowledge belongs can include at least one of the following: knowledge of the development process, knowledge of the testing process, knowledge of various business units, etc. The knowledge of various business units can include at least one of the following: database knowledge, communication knowledge, audio-visual knowledge, charging and discharging knowledge, etc.

[0076] S103. From the parent segments associated with each sub-segment in the sub-segment set, determine the target parent segment associated with the target sub-segment.

[0077] In this embodiment of the application, each parent segment can be associated with at least one child segment, and each child segment can be associated with one parent segment.

[0078] If the target subsegment includes one subsegment, the parent segment associated with that one subsegment is determined as the target parent segment. If the target subsegment includes multiple subsegments, the parent segment associated with each of those multiple subsegments is determined as the target parent segment.

[0079] S104. Based on the enterprise knowledge question content, target sub-segment, and target parent segment, generate and output enterprise knowledge answer content for the enterprise knowledge question content.

[0080] In some embodiments, S104 may include: inputting enterprise knowledge question content, target sub-segment, and target parent segment into a large model, so that the large model outputs enterprise knowledge answer content for the enterprise knowledge question content.

[0081] In some embodiments, the enterprise knowledge response content may include text response content. In other embodiments, the enterprise knowledge response content may include text response content, as well as the files to which the target sub-segment and target parent segment belong. In some embodiments, the output file may be the original file. In other embodiments, the output file may include a file with the content of the target sub-segment and target parent segment tagged in the original file.

[0082] In the technical solution provided by this application embodiment, the enterprise knowledge question content is first vectorized to obtain a question vector. Then, the target sub-segment is determined based on the matching of the question vector with the feature vectors of each sub-segment in the sub-segment set. Subsequently, the target parent segment is located. Finally, the enterprise knowledge answer content is generated by combining the question content, the target sub-segment, and the target parent segment. Compared with the keyword-based retrieval method in related technologies, which easily outputs a large amount of irrelevant knowledge content and requires users to filter it, this application embodiment, through the hierarchical knowledge division and vector matching mechanism of sub-segments and parent segments, can accurately lock the core knowledge fragments related to the question content, avoid redundant output of irrelevant knowledge, reduce the time spent by users in filtering, realize the rapid location and acquisition of enterprise knowledge, effectively solve the problem of low enterprise knowledge query efficiency in related technologies, and improve the query efficiency of enterprise knowledge.

[0083] Figure 2 A flowchart illustrating the method for determining the target sub-segment provided in the first embodiment is shown below. Figure 2 As shown, this method describes S102, and includes the following steps:

[0084] S201. Based on the similarity between the feature vectors of each sub-segment in the sub-segment set and the enterprise knowledge question vector, determine the feature vectors of multiple candidate sub-segments.

[0085] In some embodiments, the similarity between the feature vector of each sub-segment in the sub-segment set and the enterprise knowledge question vector can be obtained; the feature vector of the sub-segment corresponding to the similarity greater than the second similarity threshold is determined as the feature vector of multiple candidate sub-segments.

[0086] In the embodiments of this application, the values ​​of different similarity thresholds can be the same or different.

[0087] In other embodiments, the similarity between the feature vector of each sub-segment in the sub-segment set and the enterprise knowledge question vector can be obtained; the feature vectors of the sub-segments with the highest similarity of the second preset number of similarities are determined as the feature vectors of multiple candidate sub-segments.

[0088] In this embodiment of the application, the values ​​of different preset quantities can be the same or different.

[0089] S202. Based on the similarity between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector, and based on the validity weight of each candidate sub-segment, determine the matching degree between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector.

[0090] The validity weight of each candidate sub-segment is included in the validity weight of each sub-segment in the sub-segment set. The validity weight of each sub-segment in the sub-segment set is determined based on the data source of each sub-segment in the sub-segment set, the publication time of each sub-segment in the sub-segment set, and the version number of each sub-segment in the sub-segment set.

[0091] In some implementations, the source weight of each sub-segment is determined based on its data source; the publication time weight of each sub-segment is determined based on the time difference between its publication time and the current time; the version weight of each sub-segment is determined based on its version number; and the matching degree between the feature vector of each candidate sub-segment and the enterprise knowledge question vector is determined based on the source weight, publication time weight, and version weight of each sub-segment. For example, the more authoritative the data source of a sub-segment, the higher its source weight. For example, the smaller the time difference between the publication time and the current time of a sub-segment, the higher its publication time weight. And for example, the higher the version number of a sub-segment, the higher its version weight.

[0092] For example, the source weight, publication time weight, and version weight of each sub-segment can be multiplied or added together to obtain the matching degree between the feature vector of each candidate sub-segment and the enterprise knowledge question vector.

[0093] For example, the source weight, publication time weight, and version weight of each sub-segment can be weighted and summed or weighted averaged to obtain the matching degree between the feature vector of each candidate sub-segment and the enterprise knowledge question vector.

[0094] In other embodiments, a validity weight determination model can be obtained. The data source of each sub-segment in the sub-segment set, the publication time of each sub-segment in the sub-segment set, and the version number of each sub-segment in the sub-segment set are input into the validity weight determination model, so that the validity weight determination model outputs the validity weight of each sub-segment in the sub-segment set. Exemplarily, the validity weight determination model can be any model capable of determining weights in related technologies. For example, the validity weight determination model may include a neural network model.

[0095] In some embodiments, the product of the similarity between the feature vector of each candidate sub-segment and the enterprise knowledge question vector and the validity weight of each candidate sub-segment can be determined as the matching degree between the feature vector of each candidate sub-segment and the enterprise knowledge question vector.

[0096] In other embodiments, the similarity between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector, and the sum of the validity weights of each candidate sub-segment, can be determined as the matching degree between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector.

[0097] In some other embodiments, the similarity between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector can be weighted and averaged or summed with the validity weights of each candidate sub-segment to obtain the matching degree between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector.

[0098] S203. Based on the matching degree between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector, determine the target sub-segment matching the enterprise knowledge question content from the sub-segment set.

[0099] In some embodiments, the sub-segments with matching scores greater than a preset matching score threshold can be selected from the sub-segment set as target sub-segments for matching enterprise knowledge question content.

[0100] In other embodiments, a specified number of sub-segments with the highest matching degree can be selected from the set of sub-segments as the target sub-segments for matching the enterprise knowledge question content.

[0101] In the technical solution provided in this application embodiment, candidate sub-segments are first screened based on the similarity between the feature vector of the sub-segment and the enterprise knowledge question vector. Then, the final matching degree is calculated based on the similarity and the validity weight determined by the data source, publication time, and version of the sub-segment, thereby determining the target sub-segment. Compared with related technologies that rely solely on vector similarity to determine the matching result, which is susceptible to interference from low-quality, outdated, or non-authoritative sub-segments and thus leads to inaccurate matching, this application embodiment quantifies the quality and timeliness of sub-segments by introducing validity weights. That is, sub-segments with authoritative data sources, recent publications, and high versions receive higher weights, which strengthens the matching priority of high-quality sub-segments, weakens the influence of low-value sub-segments, makes the target sub-segment more in line with the question requirements, reduces invalid matching, and thus improves the quality of the answer.

[0102] In some embodiments, the validity weight of each sub-segment in the sub-segment set can be adjusted. The following uses the adjustment of the validity weight of a target sub-segment as an example to illustrate the method for adjusting the validity weight of a target sub-segment. The method in this application embodiment further includes: obtaining accuracy feedback information for enterprise knowledge answer content; determining the current feedback accuracy of the target sub-segment based on the accuracy feedback information; and adjusting the validity weight of the target sub-segment based on the current feedback accuracy of the target sub-segment, the current feedback provider's years of experience corresponding to the target sub-segment, the historical feedback accuracy of the target sub-segment, and the historical feedback providers' years of experience.

[0103] When an enterprise knowledge Q&A system outputs enterprise knowledge answers, it can also simultaneously output accuracy scores. For example, it can output rating options such as Excellent, Good, Average, Poor, and Very Poor. Users can trigger one of these rating options, and the user-triggered rating option will be designated as accuracy feedback information. Another example is outputting rating options between 0-10, 11-20, ..., 91-100 points. Users can trigger one of these rating options, and the user-triggered rating option will be designated as accuracy feedback information.

[0104] For example, the accuracy of feedback can range from 0 to 1. For instance, the accuracy of rating options such as Excellent, Good, Average, Poor, and Very Poor corresponds to 0.2, 0.4, 0.6, 0.8, and 1, respectively. As another example, the accuracy of feedback for scores of 0-10, 11-20, ..., 91-100 corresponds to 0.1, 0.2, ..., 1, respectively.

[0105] In some embodiments, if the ratio of the number of feedback accuracies greater than a preset accuracy threshold to the total number of feedback accuracies of the target sub-segment is greater than or equal to a preset ratio in both the current feedback accuracy and the historical feedback accuracy of the target sub-segment, then there is no need to adjust the validity weight of the target sub-segment.

[0106] In other embodiments, if the total number of feedback accuracies for the target sub-segment is less than or equal to a set number, no adjustment to the effectiveness weight of the target sub-segment is required. In some embodiments, if the total number of feedback accuracies for the target sub-segment is greater than a set number, and the ratio of the number of feedback accuracies greater than a preset accuracy threshold among the current feedback accuracy and historical feedback accuracy of the target sub-segment to the total number of feedback accuracies for the target sub-segment is greater than or equal to a preset ratio, then no adjustment to the effectiveness weight of the target sub-segment is required.

[0107] In some embodiments, if the ratio of the number of feedback accuracies greater than a preset accuracy threshold to the total number of feedback accuracies of the target sub-segment is less than a preset ratio in both the current feedback accuracy and the historical feedback accuracy of the target sub-segment, the effectiveness weight of the target sub-segment needs to be adjusted.

[0108] In other embodiments, if the number of all feedback accuracies of the target sub-segment is greater than a set number, and the ratio of the number of feedback accuracies greater than a preset accuracy threshold to the number of all feedback accuracies of the target sub-segment is less than a preset ratio, then the effectiveness weight of the target sub-segment needs to be adjusted.

[0109] For example, the adjustment of the effectiveness weight of the target sub-segment can be determined as follows: Normalize the feedback provider's years of service to obtain normalized data; take a weighted average of the current and historical feedback accuracy of each feedback accuracy for the target sub-segment, along with the normalized data of the current and historical feedback providers corresponding to the target sub-segment, to obtain a weight adjustment factor; and adjust the effectiveness weight of the target sub-segment according to the weight adjustment factor. For example, if the effectiveness weight of the target sub-segment is A and the weight adjustment factor is B, then the adjusted effectiveness weight is AA × B.

[0110] In the technical solution provided by this application embodiment, the accuracy of the current feedback of the target sub-segment is determined by obtaining the accuracy feedback information of the enterprise knowledge answer content. The validity weight is adjusted by combining the current feedback provider's years of experience, the historical feedback accuracy of the target sub-segment, and the years of experience of historical feedback providers. This enables dynamic adaptive optimization of the validity weight, significantly improving the continuous accuracy of subsequent enterprise knowledge matching. Compared to related technologies where validity weights are mostly fixed values, failing to adapt to the dynamic changes in enterprise knowledge and the value differences in feedback information, and easily leading to a decay in matching accuracy after long-term use, this application embodiment adjusts the weight based on multi-dimensional feedback parameters. It calibrates the weight in a timely manner through current feedback, ensures the stability of the adjustment through historical feedback, and highlights the reference value of authoritative feedback by combining the feedback provider's years of experience. This ensures that the weight always matches the actual application value of the sub-segment, providing a reliable basis for the accurate selection of subsequent target sub-segments and improving the accuracy of the enterprise knowledge answer content.

[0111] Figure 3 A flowchart illustrating the method for determining the target sub-segment provided in the second embodiment is shown below. Figure 3 As shown, this method describes S102, and includes the following steps:

[0112] S301. Classify the enterprise knowledge question content by attribute to obtain the attribute categories of the enterprise knowledge question content.

[0113] For example, attribute categories can include knowledge query categories and business query categories. Knowledge query categories refer to the categories corresponding to user queries whose core need is simply to obtain predetermined static knowledge information within the enterprise. These queries only target the acquisition of basic, fixed, and non-dynamically executable knowledge information within the enterprise. Business query categories refer to the categories corresponding to user queries whose core need is to obtain practical information related to the specific business operations of the enterprise. These queries need to be answered by matching relevant enterprise knowledge content related to the corresponding business.

[0114] For example, questions in the knowledge query category could include "How to troubleshoot charging abnormalities in device A?". Similarly, questions in the business query category could include "What are the sales data for each person in our sales department in December 2025, and what does this reflect?"

[0115] S302. When the attribute category of the enterprise knowledge question content includes the knowledge query category, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment corresponding to the enterprise knowledge text under the knowledge query category, based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the enterprise knowledge text under the knowledge query category.

[0116] In some embodiments, the scheme of S302 can be combined with... Figure 2 The embodiments are combined. For example, based on the similarity between the feature vector of the enterprise knowledge question and the feature vector of the sub-segment corresponding to the enterprise knowledge text under the knowledge query category, multiple candidate sub-segment feature vectors are determined; based on the similarity between the feature vector of each candidate sub-segment and the enterprise knowledge question vector, and based on the validity weight of each candidate sub-segment, the matching degree between the feature vector of each candidate sub-segment and the enterprise knowledge question vector is determined; based on the matching degree between the feature vector of each candidate sub-segment and the enterprise knowledge question vector, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment set.

[0117] S303. When the attribute category of the enterprise knowledge question content includes the business query category, determine the candidate enterprise knowledge texts that the target questioner can access from the enterprise knowledge texts under the business query category, based on the permissions of the target questioner of the enterprise knowledge question content.

[0118] For example, if the target questioner is from the sales department, the candidate company knowledge texts accessible to the target questioner include those accessible to the sales department. Similarly, if the target questioner is from the R&D department, the candidate company knowledge texts accessible to the target questioner include those accessible to the R&D department.

[0119] S304. Based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the candidate enterprise knowledge text, determine the target sub-segment matching the enterprise knowledge question content from the sub-segment corresponding to the candidate enterprise knowledge text.

[0120] In the technical solution provided in this application embodiment, the enterprise knowledge query content is first classified by attribute to obtain corresponding attribute categories. For the knowledge query category, the target sub-segment is directly matched by combining the enterprise knowledge query vector with the feature vector of the sub-segment under that category. For the business query category, the accessible candidate enterprise knowledge text is first filtered out from that category according to the permissions of the target questioner. Then, the target sub-segment is matched by combining the enterprise knowledge query vector with the feature vector of the corresponding sub-segment of the candidate enterprise knowledge text. This can greatly improve the targeting and compliance of enterprise knowledge matching, while ensuring the efficiency of knowledge query. This classification matching method can realize the differentiated processing logic adaptation of query content with different attributes. The permission filtering step added for the business query category can accurately prevent the target questioner from accessing enterprise knowledge text without permission, strictly control the access boundary of business knowledge. At the same time, both types of attributes rely on vector matching to lock the target sub-segment. On the basis of meeting the permission control requirements, it can ensure that all types of query content can be efficiently matched with the corresponding target sub-segment, effectively solving the problems of insufficient adaptability and lack of business knowledge access control under the unified matching mode.

[0121] In some embodiments, determining the target sub-segment matching the enterprise knowledge question content from the sub-segments corresponding to the candidate enterprise knowledge text, based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the candidate enterprise knowledge text, includes: classifying the enterprise knowledge question content to obtain a question knowledge category; determining multiple undetermined sub-segments matching the question knowledge category based on the segment category of the sub-segment corresponding to the candidate enterprise knowledge text; wherein, the segment category of the sub-segment corresponding to the candidate enterprise knowledge text is included in the segment category of each sub-segment in the sub-segment set, and the segment category of each sub-segment in the sub-segment set is obtained by classifying each sub-segment associated with each parent segment based on at least one knowledge category of each parent segment in the parent segment set; and determining the target sub-segment matching the enterprise knowledge question content from the multiple undetermined sub-segments matching the question knowledge category based on the enterprise knowledge question vector and the feature vector of the multiple undetermined sub-segments matching the question knowledge category.

[0122] For example, knowledge categories may include at least one of the following: knowledge of the development process, knowledge of the testing process, knowledge of various business units, etc. Knowledge of various business units may include at least one of the following: database knowledge, communication knowledge, audio-visual knowledge, charging and discharging knowledge, etc.

[0123] In the technical solution provided in this application embodiment, for candidate enterprise knowledge text under a business query category, the enterprise knowledge question content is first classified to obtain the question knowledge category. Then, based on the segment category of the corresponding sub-segment of the candidate enterprise knowledge text, matching undetermined sub-segments are selected. Finally, the target sub-segment is determined by combining the enterprise knowledge question vector and the feature vector of the undetermined sub-segment. This can effectively narrow the sub-segment matching range in the business query scenario and significantly improve the matching accuracy and efficiency of the target sub-segment. In this way, by relying on the question knowledge category and segment category formed by knowledge classification for pre-screening, sub-segments in the candidate enterprise knowledge text that are unrelated to the question knowledge category can be directly eliminated. There is no need to perform vector matching operations on all sub-segments. This reduces invalid vector matching operations and allows subsequent vector matching to focus on sub-segments of the same knowledge category, making the matching results more in line with the core requirements of the enterprise knowledge question content. At the same time, it avoids the interference of irrelevant sub-segments on the matching results, achieving a dual improvement in the efficiency and accuracy of enterprise knowledge matching in the business query scenario.

[0124] Figure 4 A flowchart illustrating a method for generating and outputting enterprise knowledge answer content in some embodiments, such as... Figure 4 As shown, this method describes S104 and includes the following steps:

[0125] S401. Determine the target answering strategy based on the length of service of the target questioner in the enterprise knowledge question content.

[0126] For example, multiple work experience ranges can be preset, and the work experience of the target questioner can be matched to the corresponding work experience range. According to the preset correspondence between each work experience range and the target answer strategy, the target answer strategy for the target questioner can be determined. For example, the multiple work experience ranges can be divided into at least a junior work experience range, an intermediate work experience range, and an advanced work experience range, with different work experience ranges corresponding to different target answer strategies.

[0127] In some embodiments, the target answering strategy corresponding to the primary age range is a basic knowledge output strategy, the target answering strategy corresponding to the intermediate age range is a refined knowledge output strategy, and the target answering strategy corresponding to the advanced age range is a deep knowledge output strategy.

[0128] S402. Based on the target answering strategy, extract target knowledge content from the target sub-segment and the target parent segment.

[0129] For example, the knowledge extraction rules corresponding to the target answer strategy are identified. Based on these rules, core knowledge information is extracted from the target sub-segment, and related supplementary knowledge information is extracted from the target parent segment. The core knowledge information and related supplementary knowledge information are then integrated to obtain the target knowledge content. For instance, the knowledge extraction rules include knowledge extraction granularity rules and knowledge extraction dimension rules. The knowledge extraction granularity rules define the level of detail in extracting knowledge from the target sub-segment and target parent segment, while the knowledge extraction dimension rules define the coverage of knowledge extracted from the target sub-segment and target parent segment.

[0130] In some embodiments, if the target answer strategy is a basic knowledge output strategy, then only the core knowledge information is extracted from the target sub-segment as the target knowledge content; if the target answer strategy is a refined or in-depth knowledge output strategy, then the core knowledge information of the target sub-segment is integrated with the associated supplementary knowledge information of the target parent segment to obtain the target knowledge content.

[0131] S403. Generate and output enterprise knowledge answers based on the enterprise knowledge questions and target knowledge content.

[0132] For example, the sentence structure and expression logic of enterprise knowledge questions can be parsed, and the target knowledge content can be reorganized and the word order optimized according to the sentence structure and expression logic to generate enterprise knowledge answer content that conforms to the expression habits of enterprise knowledge questions. The enterprise knowledge answer content can then be pushed to the terminal of the target questioner to complete the output.

[0133] In some embodiments, after generating enterprise knowledge answer content, the method further includes: standardizing the format of the enterprise knowledge answer content to match a preset knowledge output format, and then performing an output operation.

[0134] In some embodiments, the output enterprise knowledge answer content may also be marked with the source identifiers of the target sub-segment and the target parent segment corresponding to the target knowledge content.

[0135] In the technical solution provided in this application embodiment, a target answer strategy is determined based on the work experience of the target questioner in the enterprise knowledge question. Then, target knowledge content is extracted from the target sub-segment and target parent segment according to the strategy. Finally, the enterprise knowledge question content and target knowledge content are combined to generate and output enterprise knowledge answer content. This can achieve differentiated and adapted output of enterprise knowledge answer content, greatly improve the adaptability of answer content to the target questioner, and break the unified knowledge output model. By determining a matching answer strategy based on the work experience of the target questioner, the extracted target knowledge content can be tailored to the knowledge reception ability and actual needs of questioners with different work experience. Combined with the enterprise knowledge question content, targeted answer content is generated, so that the output enterprise knowledge answer content accurately responds to the core needs of the question and is adapted to the cognitive level of the questioner. This effectively avoids the problem that the answer content is too superficial and cannot meet the needs of senior questioners, or too profound and causes comprehension obstacles for newly hired questioners, thus improving the effectiveness of the output enterprise knowledge answer content for enterprise knowledge questions.

[0136] In some embodiments, a method for obtaining an enterprise knowledge text base may also be provided. Figure 5 A flowchart illustrating a method for acquiring an enterprise knowledge text base is provided for some embodiments, such as... Figure 5 As shown, the method includes the following steps:

[0137] S501. Obtain the enterprise knowledge document library, the enterprise communication data content library, and the enterprise knowledge multimedia content library.

[0138] An enterprise knowledge document repository is a collection of knowledge documents built by an enterprise and includes various types of enterprise knowledge text materials.

[0139] The enterprise communication data content repository is a collection of knowledge-based data content generated by various internal enterprise communication scenarios. As an important supplement to the enterprise knowledge document repository, the content it contains all comes from data content generated by various internal communication channels of the enterprise.

[0140] An enterprise knowledge multimedia content library is a collection of non-textual enterprise knowledge multimedia materials built by the enterprise, mainly in the form of audio, video, images, charts and other multimedia formats. It can be converted into segmentable text-based knowledge content through parsing and translation.

[0141] In some embodiments, the latest enterprise knowledge document library, enterprise communication data content library, and enterprise knowledge multimedia content library can be obtained periodically, and the following processes can be performed to improve the effectiveness of knowledge.

[0142] S502. Convert the images in the enterprise knowledge document library into text description information to obtain the first knowledge text library.

[0143] Converting an image into text description information can include: if the number of characters in the image exceeds a preset number of characters, performing Optical Character Recognition (OCR) on the image to obtain text description information; if the number of characters in the image is less than or equal to the preset number of characters, inputting the image into a large model so that the large model outputs the text description information of the image.

[0144] S503. Extract enterprise knowledge content related to enterprise knowledge from the enterprise communication data content library, and convert the multimedia content in the enterprise knowledge content into text description information to obtain the second knowledge text library.

[0145] In some embodiments, all communication data content in the enterprise communication data content library is retrieved, and the communication data content is searched and matched according to a preset enterprise knowledge keyword library. Communication data content containing keywords in the enterprise knowledge keyword library is then selected, and the selected communication data content is determined as enterprise knowledge content related to enterprise knowledge. For example, the preset enterprise knowledge keyword library may include at least one of the following: enterprise business terms, technical professional terms, management standard terms, job title names, and project identification information.

[0146] Multimedia content can include at least one of audio content, video content, and image content. For audio content within multimedia content, text recognition can be performed on the audio content to obtain a textual description of the audio content. For video content within multimedia content, text recognition can be performed on the audio within the video content, and the images within the video content can be converted into textual descriptions to obtain textual descriptions of the video content. For image content within multimedia content, the image content can be converted into textual descriptions to obtain textual descriptions of the image content.

[0147] S504. Convert the multimedia content of each enterprise knowledge in the enterprise knowledge multimedia content library into text description information to obtain the third knowledge text library.

[0148] S505. Merge the first knowledge text library, the second knowledge text library, and the third knowledge text library to obtain the enterprise knowledge text library.

[0149] In the technical solution provided in this application embodiment, by integrating three types of data sources—enterprise knowledge document library, enterprise communication data content library, and enterprise knowledge multimedia content library—and converting the images and multimedia content in each library into text description information, three types of knowledge text libraries are generated accordingly. These three text libraries are then merged to obtain the enterprise knowledge text library. This greatly expands the knowledge coverage dimension of the enterprise knowledge text library, breaking through the knowledge limitations of a single text library. The enterprise knowledge text library includes existing enterprise document knowledge, knowledge extracted from communication data, and knowledge from the enterprise knowledge multimedia content library. The merging of these three types of knowledge text libraries allows the enterprise knowledge text library to simultaneously carry knowledge content from multiple sources—documents, communications, and multimedia—avoiding omissions in knowledge data sources and improving the completeness of knowledge coverage in enterprise knowledge Q&A.

[0150] In some embodiments, by deploying an enterprise knowledge Q&A system on an enterprise server, localized and intelligent management of enterprise knowledge is achieved; all sensitive data is processed locally without uploading to the cloud, fully meeting enterprise data security compliance requirements; the system can be used without a network connection and still provides professional support in offline environments, significantly improving work efficiency; the enterprise knowledge text library used is the enterprise's own knowledge, eliminating the need for external knowledge, avoiding application programming interface (API) call fees, and significantly reducing long-term operating costs; by deeply integrating enterprise professional knowledge into the reasoning process, it provides in-depth and valuable insights; the structured storage of enterprise documents, product specifications, historical reports, professional terminology, and other knowledge effectively avoids repetitive work; enterprise employees can quickly query the knowledge base to solve common problems, greatly reducing waiting time, and new employees can quickly understand the enterprise knowledge system, significantly reducing knowledge gaps.

[0151] Based on the same inventive concept, this application also provides an enterprise knowledge question-answering device for implementing the enterprise knowledge question-answering method described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more enterprise knowledge question-answering device embodiments provided below can be found in the limitations of the enterprise knowledge question-answering method above, and will not be repeated here.

[0152] In one exemplary embodiment, Figure 6 A schematic diagram of the structure of an enterprise knowledge question-answering device provided in some embodiments, such as Figure 6 As shown, the enterprise knowledge Q&A device 600 includes:

[0153] The vector acquisition module 601 is used to vectorize the acquired enterprise knowledge question content in response to the enterprise knowledge question content, and obtain the enterprise knowledge question vector.

[0154] The matching module 602 is used to determine the target sub-segment for matching the enterprise knowledge question content from the sub-segment set based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set; the sub-segment set includes at least one sub-segment associated with each parent segment in the parent segment set, and the at least one sub-segment associated with each parent segment is obtained by segmenting each enterprise knowledge text in the enterprise knowledge text library.

[0155] The parent segment determination module 603 is used to determine the target parent segment associated with the target sub-segment from the parent segments associated with each sub-segment in the sub-segment set;

[0156] The generation and output module 604 is used to generate and output enterprise knowledge answer content based on the enterprise knowledge question content, target sub-segment, and target parent segment.

[0157] In some embodiments, the matching module 602 includes a candidate sub-segment vector determination unit, a matching degree determination unit, and a target sub-segment determination unit. The candidate sub-segment vector determination unit is used to determine the feature vectors of multiple candidate sub-segments based on the similarity between the feature vectors of each sub-segment in the sub-segment set and the enterprise knowledge question vector. The matching degree determination unit is used to determine the matching degree between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector based on the similarity between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector, and based on the validity weight of each candidate sub-segment. The validity weight of each candidate sub-segment is included in the validity weight of each sub-segment in the sub-segment set, and the validity weight of each sub-segment in the sub-segment set is determined based on the data source, publication time, and version number of each sub-segment in the sub-segment set. The target sub-segment determination unit is used to determine the target sub-segment matching the enterprise knowledge question content from the sub-segment set based on the matching degree between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector.

[0158] In some embodiments, the enterprise knowledge question-answering device further includes an adjustment module, configured to acquire accuracy feedback information on the enterprise knowledge answer content; determine the current feedback accuracy of the target sub-segment based on the accuracy feedback information; and adjust the validity weight of the target sub-segment based on the current feedback accuracy of the target sub-segment, the current years of service of the current feedbacker corresponding to the target sub-segment, the historical feedback accuracy of the target sub-segment, and the historical years of service of the feedbackers of the target sub-segment.

[0159] In some embodiments, the matching module includes a classification unit and a target sub-segment determination unit. The classification unit is used to classify the enterprise knowledge question content by attribute to obtain the attribute category of the enterprise knowledge question content. The target sub-segment determination unit is used to determine the target sub-segment matching the enterprise knowledge question content from the sub-segments corresponding to the enterprise knowledge text under the knowledge query category, based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the enterprise knowledge text under the knowledge query category, when the attribute category of the enterprise knowledge question content includes the business query category; and to determine the candidate enterprise knowledge text that the target questioner can access from the enterprise knowledge text under the business query category, based on the permissions of the target questioner of the enterprise knowledge question content; and to determine the target sub-segment matching the enterprise knowledge question content from the sub-segments corresponding to the candidate enterprise knowledge text, based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the candidate enterprise knowledge text.

[0160] In some embodiments, the target sub-segment determination unit is further configured to classify the enterprise knowledge question content to obtain a question knowledge category; determine multiple undetermined sub-segments matching the question knowledge category based on the segmentation category of the sub-segment corresponding to the candidate enterprise knowledge text; wherein, the segmentation category of the sub-segment corresponding to the candidate enterprise knowledge text is included in the segmentation category of each sub-segment in the sub-segment set, and the segmentation category of each sub-segment in the sub-segment set is obtained by classifying each sub-segment associated with each parent segment based on at least one knowledge category of each parent segment in the parent segment set; and determine the target sub-segment matching the enterprise knowledge question content from the multiple undetermined sub-segments matching the question knowledge category based on the enterprise knowledge question vector and the feature vector of the multiple undetermined sub-segments matching the question knowledge category.

[0161] In some embodiments, the generation and output module includes a strategy determination unit, an extraction unit, and a generation and output unit; the strategy determination unit is used to determine a target answer strategy based on the years of service of the target questioner of the enterprise knowledge question content; the extraction unit is used to extract target knowledge content from the target sub-segment and the target parent segment according to the target answer strategy; the generation and output unit is used to generate and output enterprise knowledge answer content for the enterprise knowledge question content based on the enterprise knowledge question content and the target knowledge content.

[0162] In some embodiments, the enterprise knowledge question-and-answer device further includes a text library acquisition module, used to acquire an enterprise knowledge document library, an enterprise communication data content library, and an enterprise knowledge multimedia content library; convert images in the enterprise knowledge document library into text description information to obtain a first knowledge text library; extract enterprise knowledge content related to enterprise knowledge from the enterprise communication data content library, and convert multimedia content in the enterprise knowledge content into text description information to obtain a second knowledge text library; convert each enterprise knowledge multimedia content in the enterprise knowledge multimedia content library into text description information to obtain a third knowledge text library; and merge the first knowledge text library, the second knowledge text library, and the third knowledge text library to obtain an enterprise knowledge text library.

[0163] The descriptions of the above device embodiments are similar to those of the above method embodiments, and have similar beneficial effects. For technical details not disclosed in the device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0164] Each module in the aforementioned enterprise knowledge Q&A device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0165] In one exemplary embodiment, Figure 7This is a schematic diagram of the structure of a computer device provided in some embodiments. The computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for wired or wireless communication with external terminals. Wireless communication can be implemented through Wireless Fidelity (WIFI), mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements an enterprise knowledge question-answering method. The display unit of the computer device is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0166] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0167] For example, a computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method of any of the above embodiments.

[0168] In one embodiment, a computer-readable storage medium is provided, wherein a computer program, when executed by a processor, implements the steps of the method provided in any of the above embodiments.

[0169] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method provided in any of the above embodiments.

[0170] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the methods described above.

[0171] The processor, functional modules, or functional units in any embodiment of this application may include an integration of one or more of the following: a general-purpose processor, an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field-programmable gate array (FPGA), a central processing unit (CPU), a graphics processing unit (GPU), an embedded neural network processing unit (NPU), a controller, a microcontroller, a microprocessor, a programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, a quantum computing-based data processing logic unit, an artificial intelligence (AI) processor, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0172] The memory or computer-readable storage medium in any embodiment of this application may include at least one of non-volatile memory and volatile memory. Non-volatile memory includes integration of one or more of the following: Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Ferromagnetic Random Access Memory (FRAM), Flash Memory, Magnetic Surface Memory, Optical Disc, Compact Disc Read-Only Memory (CD-ROM), Magnetic Tape, Floppy Disk, Flash Memory, Optical Memory, High-Density Embedded Non-Volatile Memory, Resistive Random Access Memory (ReRAM), Magnetoresistive Random Access Memory (MRAM), Ferroelectric Random Access Memory (FRAM), Phase Change Memory (PCM), Graphene Memory, Volatile Memory, etc. Volatile memory includes one or more of the following: Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0173] The acquisition, transmission, storage, use, and processing of data in this application comply with relevant national laws and regulations. It should be noted that certain software, components, models, and other existing industry solutions may be mentioned in the embodiments of this application. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0174] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0175] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for answering enterprise knowledge questions, characterized in that, The method is applied to an enterprise knowledge-based question-and-answer system, and the method includes: In response to the obtained enterprise knowledge question content, the enterprise knowledge question content is vectorized to obtain the enterprise knowledge question vector. Based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment set; the sub-segment set includes at least one sub-segment associated with each parent segment in the parent segment set, and the at least one sub-segment associated with each parent segment is obtained by segmenting each enterprise knowledge text in the enterprise knowledge text library; From the parent segments associated with each sub-segment in the sub-segment set, determine the target parent segment associated with the target sub-segment; Based on the enterprise knowledge question content, the target sub-segment, and the target parent segment, generate and output enterprise knowledge answer content for the enterprise knowledge question content.

2. The method according to claim 1, characterized in that, The step of determining the target sub-segment matching the enterprise knowledge question content from the sub-segment set based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set includes: Based on the similarity between the feature vectors of each sub-segment in the sub-segment set and the enterprise knowledge question vector, feature vectors of multiple candidate sub-segments are determined. The matching degree between the feature vector of each candidate sub-segment and the enterprise knowledge question vector is determined based on the similarity between the feature vector of each candidate sub-segment and the enterprise knowledge question vector, and based on the validity weight of each candidate sub-segment; wherein, the validity weight of each candidate sub-segment is included in the validity weight of each sub-segment in the sub-segment set, and the validity weight of each sub-segment in the sub-segment set is determined based on the data source of each sub-segment in the sub-segment set, the publication time of each sub-segment in the sub-segment set, and the version number of each sub-segment in the sub-segment set; Based on the matching degree between the feature vectors of each candidate sub-segment and the enterprise knowledge question vector, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment set.

3. The method according to claim 2, characterized in that, The method further includes: Obtain accuracy feedback information for the answers provided regarding the enterprise's knowledge; Based on the accuracy feedback information, determine the current feedback accuracy of the target sub-segment; The effectiveness weight of the target sub-segment is adjusted based on the current feedback accuracy of the target sub-segment, the current feedback provider's years of experience for the target sub-segment, the historical feedback accuracy of the target sub-segment, and the historical feedback providers' years of experience for the target sub-segment.

4. The method according to claim 1, characterized in that, The step of determining the target sub-segment matching the enterprise knowledge question content from the sub-segment set based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set includes: The enterprise knowledge question content is classified by attributes to obtain the attribute categories of the enterprise knowledge question content; When the attribute category of the enterprise knowledge question content includes a knowledge query category, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment corresponding to the enterprise knowledge text under the knowledge query category, based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the enterprise knowledge text under the knowledge query category. If the attribute category of the enterprise knowledge question content includes a business query category, then, based on the permissions of the target questioner, the candidate enterprise knowledge text that the target questioner can access is determined from the enterprise knowledge text under the business query category. Based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the candidate enterprise knowledge text, the target sub-segment matching the enterprise knowledge question content is determined from the sub-segment corresponding to the candidate enterprise knowledge text.

5. The method according to claim 4, characterized in that, The step of determining the target sub-segment matching the enterprise knowledge question content from the sub-segments corresponding to the candidate enterprise knowledge text based on the enterprise knowledge question vector and the feature vector of the sub-segment corresponding to the candidate enterprise knowledge text includes: The enterprise knowledge questions are categorized to obtain the question knowledge categories; Based on the segmentation category of the sub-segment corresponding to the candidate enterprise knowledge text, multiple undetermined sub-segments matching the question knowledge category are determined; wherein, the segmentation category of the sub-segment corresponding to the candidate enterprise knowledge text is included in the segmentation category of each sub-segment in the sub-segment set, and the segmentation category of each sub-segment in the sub-segment set is obtained by classifying each sub-segment associated with each parent segment according to at least one knowledge category of each parent segment in the parent segment set; Based on the enterprise knowledge question vector and the feature vectors of multiple undetermined sub-segments matching the question knowledge category, the target sub-segment matching the enterprise knowledge question content is determined from the multiple undetermined sub-segments matching the question knowledge category.

6. The method according to any one of claims 1-5, characterized in that, The step of generating and outputting enterprise knowledge answer content based on the enterprise knowledge question content, the target sub-segment, and the target parent segment includes: Based on the target questioner's years of service in the enterprise knowledge question content, determine the target answering strategy; Based on the target answering strategy, target knowledge content is extracted from the target sub-segment and the target parent segment; Based on the enterprise knowledge question and the target knowledge content, generate and output enterprise knowledge answer content for the enterprise knowledge question.

7. The method according to any one of claims 1-5, characterized in that, The method further includes: Access the enterprise knowledge document library, the enterprise communication data content library, and the enterprise knowledge multimedia content library; The images in the enterprise knowledge document library are converted into text description information to obtain the first knowledge text library; Extract enterprise knowledge content related to enterprise knowledge from the enterprise communication data content library, and convert the multimedia content in the enterprise knowledge content into text description information to obtain a second knowledge text library; The enterprise knowledge multimedia content in the enterprise knowledge multimedia content library is converted into text description information to obtain the third knowledge text library; The first knowledge text library, the second knowledge text library, and the third knowledge text library are merged to obtain the enterprise knowledge text library.

8. A corporate knowledge question-and-answer device, characterized in that, The enterprise knowledge Q&A device includes: The vector acquisition module is used to vectorize the acquired enterprise knowledge question content in response to the enterprise knowledge question content, and obtain the enterprise knowledge question vector. The matching module is used to determine the target sub-segment matching the enterprise knowledge question content from the sub-segment set based on the enterprise knowledge question vector and the feature vector of each sub-segment in the sub-segment set; the sub-segment set includes at least one sub-segment associated with each parent segment in the parent segment set, and the at least one sub-segment associated with each parent segment is obtained by segmenting each enterprise knowledge text in the enterprise knowledge text library; The parent segment determination module is used to determine the target parent segment associated with the target sub-segment from the parent segments associated with each sub-segment in the sub-segment set; The generation and output module is used to generate and output enterprise knowledge answer content for the enterprise knowledge question content based on the enterprise knowledge question content, the target sub-segment, and the target parent segment.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images