Methods and systems for constructing local knowledge bases using artificial intelligence

Abstract

This invention relates to the field of knowledge base construction technology, and particularly to a method and system for constructing a local knowledge base using artificial intelligence. The method includes: defining the usage area of ​​the local knowledge base; selecting several source points; using pre-integrated audio pickup devices at the source points to collect audio data and segment it into several segments; comparing the data with a pre-set voiceprint database to determine the weight value of each segment; transcribing the segments into text and inputting the text into a pre-built semantic analysis model to output keywords; and writing the keywords into a pre-set template to obtain new knowledge points. This invention, by constructing a multi-level version tree, can adaptively update the local knowledge base based on new knowledge points, effectively achieving parallel management of numerous knowledge points, while intuitively displaying the change process of knowledge points, greatly improving the efficiency of knowledge point retrieval and user satisfaction.

Description

Technical Field

[0001] This invention relates to the field of knowledge base construction technology, and in particular to a method and system for constructing a local knowledge base using artificial intelligence. Background Technology

[0002] A knowledge base is a collection of knowledge information used for systematic storage, organization, management, and retrieval. It typically exists in a structured or semi-structured format, aiming to provide users with convenient access to and use of professional knowledge, including internal enterprise decision-making, technology, and business information. A local knowledge base is deployed on a local device or local area network and is suitable for scenarios with high requirements for data privacy, security, and real-time performance.

[0003] Because the knowledge base is deployed locally, it lacks efficient linkage with external real-time information sources, resulting in a lag in knowledge content and difficulty in reflecting new decisions, new technologies or the latest business dynamics in a timely manner. The common practice in the existing technology is to use manual maintenance of the local knowledge base, including knowledge extraction, data cleaning, classification and organization and access control, but this has a high technical threshold and high human resource cost.

[0004] Therefore, "how to adaptively update the local knowledge base" is the technical problem that this invention aims to solve. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for constructing a local knowledge base using artificial intelligence, in order to solve the problem of "how to adaptively update the local knowledge base" mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for constructing a local knowledge base using artificial intelligence, the method comprising:

[0008] The local knowledge base usage area is defined, several source points are selected, and audio data is collected using the audio pickup devices pre-integrated in the source points. The data is then divided into several segments, compared with the preset voiceprint database, and the weight value of each segment is determined.

[0009] The segments are transcribed and input into a pre-built semantic analysis model to output keywords. The keywords are written into a preset template to obtain new knowledge points. The new knowledge points are compared with the local knowledge base to obtain target items. It is determined whether the similarity between the new knowledge points and the target items is greater than a threshold. If yes, the target items are covered. If no, the covering process is recorded, change data is generated, the target object is extracted from the new knowledge points and target items, the change data is sent to the target object's device terminal, and rollback permission is granted.

[0010] The local knowledge base is clustered into several categories, parent nodes are created that correspond one-to-one with each category, pre-built child nodes are attached to the parent nodes, target items are written into the child nodes, the parent nodes and child nodes are integrated to generate a multi-level version tree, the local knowledge base is linked, the order of the multi-level version tree is set, when the number of changed data exceeds the order, new child nodes are split off, the multi-level version tree is updated, and a linkage calling mechanism is embedded in the multi-level version tree and the local knowledge base in parallel.

[0011] Furthermore, the step of comparing with a preset voiceprint database to determine the weight value of each segment includes:

[0012] By comparing the segments with the voiceprint database, the main sound source of each segment is determined, and a lookup table is created, wherein the lookup table consists of a main sound source item and a weight value item.

[0013] The correspondence between segments and weight values ​​is established through the main body of the sound source.

[0014] Furthermore, the method also includes:

[0015] The audio data is input into the semantic analysis model, and the execution intensity of each segment is output, wherein the execution intensity includes at least: high, medium and low;

[0016] Delete the segments corresponding to execution intensities of medium and low.

[0017] Furthermore, the step of collecting audio data using a pre-integrated pickup device at the source location and dividing it into several segments includes:

[0018] Using the aforementioned sound pickup device, a microphone array is constructed, a sound recording mode is generated, and embedded into the microphone array, wherein the sound recording mode includes at least: a discussion mode and a speaking mode;

[0019] The weight values ​​are adjusted via the sound source.

[0020] Furthermore, the step of comparing the newly added knowledge points with the local knowledge base to obtain the target item includes:

[0021] The local knowledge base is divided into several existing knowledge points, and a keyword matching algorithm is used to match and calculate the new knowledge points and existing knowledge points.

[0022] Based on the newly added knowledge points, the existing knowledge points with the highest matching degree are found and defined as target items.

[0023] Furthermore, the steps of extracting the target object from the newly added knowledge points and target items, sending the changed data to the target object's device terminal, and granting rollback permissions include:

[0024] Construct a selection window and write the changed data into the selection window;

[0025] Embed "Accept," "Modify Suggestion," and "Reject" buttons in the selection window.

[0026] Furthermore, the step of splitting out new child nodes and updating the multi-level version tree includes:

[0027] The target items in the new child nodes are synchronized to the parent node, and the summary is extracted to build the index mechanism;

[0028] Integrate all version trees to generate a version forest.

[0029] The present invention also provides a local knowledge base construction system applying artificial intelligence, the system comprising:

[0030] The determination module is used to define the usage area of ​​the local knowledge base, select several source points, use the sound pickup devices pre-integrated in the source points to collect audio data, divide it into several segments, compare it with the preset voiceprint library, and determine the weight value of each segment.

[0031] The rollback module is used to transcribe the segmented text and input it into a pre-built semantic analysis model to output keywords. The keywords are written into a preset template to obtain new knowledge points. The new knowledge points are compared with the local knowledge base to obtain target items. It is determined whether the similarity between the new knowledge points and the target items is greater than a threshold. If so, the target items are overwritten. If not, the overwriting process is recorded, change data is generated, the target object is extracted from the new knowledge points and target items, the change data is sent to the target object's device terminal, and rollback permission is granted.

[0032] The calling module is used to cluster the local knowledge base into several categories, create parent nodes corresponding one-to-one with the categories, attach pre-built child nodes to the parent nodes, write target items to the child nodes, integrate the parent nodes and child nodes to generate a multi-level version tree, link the local knowledge base, set the order of the multi-level version tree, split off new child nodes when the number of changed data exceeds the order, update the multi-level version tree, and embed a linkage calling mechanism in parallel with the multi-level version tree and the local knowledge base.

[0033] Furthermore, the determining module includes:

[0034] The comparison unit is used to compare the segments with the voiceprint library, determine the main sound source of each segment, and create a comparison table, wherein the comparison table consists of a sound source main item and a weight value item.

[0035] Establish a unit to establish the correspondence between segments and weight values ​​via the sound source body;

[0036] A construction unit is used to construct a microphone array using the sound pickup device, generate a sound recording mode, and embed it into the microphone array, wherein the sound recording mode includes at least: a discussion mode and a speaking mode;

[0037] An adjustment unit is used to adjust the weight value via the sound source body.

[0038] Furthermore, the rollback module includes:

[0039] The calculation unit is used to divide the local knowledge base into several existing knowledge points and use a keyword matching algorithm to perform matching calculations on the new knowledge points and existing knowledge points.

[0040] A definition unit is used to find the existing knowledge point with the highest matching degree based on the newly added knowledge point, and define it as the target item;

[0041] The writing unit is used to construct a selection window and write the changed data into the selection window;

[0042] An embedding unit is used to embed accept, modification suggestion, and rejection buttons into the selection window.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] This invention, by defining segments, can significantly broaden the data sources of the knowledge base, prevent knowledge loss, and retain important information conveyed orally. By identifying new knowledge points and coverage targets, it can automatically update the local knowledge base without manual recording, greatly improving the data accuracy. By generating change data, it can organically combine artificial intelligence with human supervision, improving the accuracy and reliability of the knowledge base content, effectively ensuring the traceability of the knowledge base content, and enhancing users' trust and satisfaction with the local knowledge base. By constructing a multi-level version tree, it can adaptively update the local knowledge base based on new knowledge points, effectively achieving parallel management of numerous knowledge points, while intuitively displaying the change process of knowledge points, greatly improving the efficiency of knowledge point retrieval and user satisfaction. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0046] Figure 1 This is an example diagram illustrating a multi-level version tree in the local knowledge base construction method using artificial intelligence provided in this embodiment of the invention.

[0047] Figure 2 A flowchart illustrating the method for constructing a local knowledge base using artificial intelligence, as provided in an embodiment of the present invention.

[0048] Figure 3 A first sub-flow diagram of the local knowledge base construction method using artificial intelligence provided in an embodiment of the present invention;

[0049] Figure 4 This is a second sub-flow diagram of the local knowledge base construction method using artificial intelligence provided in an embodiment of the present invention;

[0050] Figure 5 A third sub-flow diagram of the local knowledge base construction method using artificial intelligence provided in an embodiment of the present invention;

[0051] Figure 6 This is a block diagram of a local knowledge base construction system using artificial intelligence, provided in an embodiment of the present invention.

[0052] Figure 7 A block diagram illustrating the composition of a module in a local knowledge base construction system using artificial intelligence, provided in an embodiment of the present invention.

[0053] Figure 8 A block diagram illustrating the composition of the rollback module in a local knowledge base construction system using artificial intelligence, as provided in an embodiment of the present invention.

[0054] Figure 9 A block diagram showing the composition of the calling module in a local knowledge base construction system using artificial intelligence, provided in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0056] In Example 1, Figure 1 and Figure 2 The implementation flow of the local knowledge base construction method using artificial intelligence provided in this embodiment of the invention is illustrated below, and is described in detail below:

[0057] S100: Define the usage area of ​​the local knowledge base, select several source points, use the pickup devices pre-integrated in the source points to collect audio data, divide it into several segments, compare it with the preset voiceprint library, and determine the weight value of each segment.

[0058] Define the usage area of ​​the local knowledge base, which refers to the office area of ​​the creator of the local knowledge base, such as a company or a floor. Within the usage area, select several source locations, such as meeting rooms and discussion areas, which are places where knowledge exchanges frequently occur. Deploy audio pickup devices at the source locations to capture the dialogue, speeches, or discussions between users. Use the audio pickup devices to collect audio data at the source locations and divide the audio data into several independent segments, i.e., segments, according to rules such as time intervals, semantic boundaries, or speech pauses. Each segment represents a speech fragment or semantic unit.

[0059] The voiceprint features in each segment are identified, and compared with the pre-built voiceprint database to determine the source object of each segment. The source object is the individual who makes the sound in the segment. Each individual makes the sound and has a different weight value, which is the weight value of the segment it corresponds to. The weight value should be determined according to the job position and job attributes of the source object. For example, a larger weight can be set for the company leader and a smaller weight can be set for the person who actually performs the work.

[0060] In this application, the voiceprint database is integrated into the edge device within the usage area. The collected audio data is sent to the edge device for preprocessing, the voiceprint feature vector is extracted, and the source object is determined by comparing it with the voiceprint database. For locally generated text data, it can also be segmented and processed.

[0061] S200: Transcribe the segmented text and input it into a pre-built semantic analysis model to output keywords. Write the keywords into a preset template to obtain new knowledge points. Compare the new knowledge points with the local knowledge base to obtain target items. Determine whether the similarity between the new knowledge points and the target items is greater than a threshold. If yes, cover the target items. If no, record the covering process, generate change data, extract the target object from the new knowledge points and target items, send the change data to the target object's device terminal, and grant rollback permission.

[0062] Using speech recognition technology, the speech content corresponding to each segment is converted into text data. The transcribed text is then input into a pre-built semantic analysis model. This model is a processing model based on natural language processing technology, used to understand the semantic relationships, context, and potential meanings in the text, and to extract valuable keywords. The keywords are specific content words. The extracted keywords are written into a preset template to obtain structured new knowledge points. The preset template is pre-defined by the local knowledge base administrators. The new knowledge points can be new decisions, new technology content, or new business descriptions, etc.

[0063] The process compares newly added knowledge points with existing knowledge points in the local knowledge base. Using methods such as keyword matching and cosine similarity calculation, it determines the degree of association between the new knowledge points and existing knowledge points in the local knowledge base. The existing knowledge point with the highest matching degree is identified and defined as the target item. It then checks if the similarity between the new knowledge point and the target item exceeds a threshold. If it does, the new knowledge point and the target item are considered to have high semantic consistency, and a coverage operation is performed, replacing the content corresponding to the target item with the new knowledge point. If the similarity does not exceed the threshold, the target item is also covered using the new knowledge point. The coverage process is recorded, including the covered content. The original content (target item), newly extracted content (new knowledge point), the time of coverage, and the source of the operation are used to generate a complete coverage log, i.e., change data. The source objects corresponding to the new knowledge point and the source objects corresponding to the target item are collectively referred to as target objects. The change data is sent in parallel to the device terminal corresponding to the target object, and rollback permissions are enabled, so that the target object can review, confirm, or roll back the coverage process within a specific time window. It should be noted that the target object may be one person or multiple people. If any target object chooses to roll back, the change data will continue to be sent to the local knowledge base management personnel terminal for manual review.

[0064] The above transcription process occurs in the edge device, and the rollback operation is performed through the terminal device. After receiving the changed data, the edge device, through the access control mechanism, allows the target object or other users to review, confirm or revoke the changed content within a specified time.

[0065] S300: Cluster the local knowledge base into several categories, create parent nodes corresponding one-to-one with each category, attach pre-built child nodes to the parent nodes, write target items to the child nodes, integrate the parent nodes and child nodes to generate a multi-level version tree, link the local knowledge base, set the order of the multi-level version tree, when the number of changed data is greater than the order, split off new child nodes, update the multi-level version tree, and embed a linkage call mechanism in parallel with the multi-level version tree and the local knowledge base.

[0066] Based on the types of existing knowledge points in the local knowledge base and their corresponding source objects, the existing knowledge points are clustered into several categories. Parent nodes are created, each corresponding to a category. Both parent and child nodes are logical nodes, not physical computing devices, but only used to represent the existing knowledge points of each category. Child nodes are created, each corresponding to a target item, and these child nodes are attached to the parent nodes of their corresponding categories. The parent and child nodes are then integrated to generate a multi-level version tree. This multi-level version tree is a tree-like data structure, similar to a B-tree in existing technologies. The order of the child nodes is determined by the local knowledge base administrator and refers to the number of nodes under the same root node, as shown in the appendix to the manual. Figure 1The order of child node A is 4; for example, if financial decision data (in this application, decision data refers to policy data) is stored in the local knowledge base, the financial policy data is divided into multiple categories, such as accounting measurement policy, depreciation and amortization policy, revenue recognition policy, tax policy, and tax incentive policy. A parent node is created for each category, and a child node (or target item) is created for each knowledge point. The difference between the two is that the target item is mainly used to show data changes. The parent node and child node are integrated to generate a version tree. If the tax policy knowledge point has been modified 4 times (as shown in the appendix of the specification), the version tree is updated accordingly. Figure 1 If the order is 3, then the child node (A) corresponding to the tax policy class is promoted to a parent node, and new child nodes (A1, A2, A3, and A4) are split off, generating a new version tree (see attached manual). Figure 1 (This is the state before the split). At the same time, a linkage call mechanism is embedded into the multi-level version tree and the local knowledge base. The linkage call mechanism is as follows: after the required knowledge point is determined using the multi-level version tree, the local knowledge base will also output the original data of the knowledge point in a synchronous manner.

[0067] As per the instruction manual Figure 1 As shown, A1 represents existing knowledge points related to tax policies. After a certain meeting, a new policy is formulated, covering A1 and identifying new knowledge points A2. After other meetings, A3 and A4 are also covered. Since the knowledge points corresponding to A1 have changed multiple times, in order to record the change process and avoid policy deviations, A is promoted to a parent node. The advantage of doing this is that it can improve the query efficiency of the data (tax policy category) corresponding to A. If multiple coverages have not occurred, the newly added knowledge points can be displayed using child nodes.

[0068] In Example 2, Figure 3 The implementation flow of the local knowledge base construction method using artificial intelligence provided by an embodiment of the present invention is illustrated. The following details the steps of comparing with a preset voiceprint database to determine the weight value of each segment:

[0069] S101: Compare the segments with the voiceprint library to determine the main sound source of each segment and create a lookup table, wherein the lookup table consists of a main sound source item and a weight value item.

[0070] The collected audio data is divided into several segments, and each segment is compared with multiple recorded sound sources in the voiceprint database. Based on the matching results of the voiceprint features, the main body of the sound source corresponding to each segment is determined, i.e., the speaker's identity. A lookup table consisting of a sound source main body item and a weight value item is created. The weight value is determined by the position, work attributes, etc. of the sound source main body.

[0071] S102: Establish the correspondence between segments and weight values ​​through the main body of the sound source.

[0072] Each segment corresponds to a sound source subject, and each sound source subject corresponds to a weight value. Using the sound source subjects, the correspondence between segments and weight values ​​is established.

[0073] In Example 3, Figure 3 This paper illustrates the implementation flow of a local knowledge base construction method using artificial intelligence provided in an embodiment of the present invention. The following details the steps of collecting audio data using a pre-integrated pickup device at the source location and dividing it into several segments:

[0074] S103: Using the sound pickup device, construct a microphone array, generate a sound recording mode, and embed it into the microphone array, wherein the sound recording mode includes at least: discussion mode and speaker mode.

[0075] The spatial arrangement of the sound pickup devices was adjusted to construct a microphone array, and multiple sound recording modes were created, including a discussion mode and a presentation mode. Specifically, the discussion mode is suitable for scenarios where multiple people speak simultaneously. The microphone array uses omnidirectional multi-focus beamforming technology to evenly pick up multiple sound sources, ensuring that the voice signals of each participant can be clearly captured. The presentation mode is suitable for single-person presentations or lectures. The microphone array uses directional beamforming technology to focus the sound pickup on the direction of the presenter, suppressing background noise from non-target directions and highlighting the presenter's voice.

[0076] S104: The weight value is adjusted via the sound source body.

[0077] The weight values ​​should be adjusted according to the different entities of the sound source; for example, different personnel may correspond to different roles in different meetings. In a production task scheduling meeting, the weight value corresponding to the financial personnel should be reduced.

[0078] In Example 4, Figure 4 The implementation flow of the local knowledge base construction method using artificial intelligence provided in this embodiment of the invention is illustrated below. The steps of comparing the newly added knowledge points with the local knowledge base to obtain the target item are described in detail below:

[0079] S201: Divide the local knowledge base into several existing knowledge points, and use a keyword matching algorithm to perform matching calculations on the new knowledge points and existing knowledge points.

[0080] The local knowledge base is divided into several existing knowledge points. The matching degree between the new knowledge points and the existing knowledge points is calculated using a keyword matching algorithm. It should be noted that this example only uses the keyword matching algorithm. The cosine similarity formula can also be used to calculate the similarity and find the existing knowledge points with the highest similarity.

[0081] S202: Based on the newly added knowledge points, find the existing knowledge points with the highest matching degree and define them as target items.

[0082] The existing knowledge point with the highest matching degree is defined as the target item. Each new knowledge point corresponds to a target item. The target item can be simply understood as the data that needs to be replaced using the new knowledge point.

[0083] In Example 5, Figure 4 This paper illustrates the implementation flow of a local knowledge base construction method using artificial intelligence provided in an embodiment of the present invention. The following details the steps of extracting the target object from newly added knowledge points, sending the changed data to the target object's device terminal, and granting rollback permissions:

[0084] S203: Construct a selection window and write the changed data into the selection window.

[0085] A selection window is constructed to display the change data of the local knowledge base. The selection window uses interface components to visualize the change data and supports user interaction.

[0086] S204: Embed "Accept", "Modify Suggestion", and "Reject" buttons in the selection window.

[0087] Embed buttons for accepting, modifying suggestions, and rejecting in the selection window. Accepting means confirming the overwriting of the target item, while rejecting means rolling back the overwriting operation and restoring the target item.

[0088] In Example 6, Figure 5 The implementation flow of the local knowledge base construction method using artificial intelligence provided in this embodiment of the invention is illustrated below. The steps of splitting out new child nodes and updating the multi-level version tree are described in detail below:

[0089] S301: Synchronize the target items in the new child node to the parent node, extract the summary, and build the index mechanism.

[0090] From each child node, segmented summaries are extracted. Using keywords, entity names, timestamps, and source objects in the summaries, an index table is built. The indexing mechanism refers to using the index table to retrieve the summaries.

[0091] S302: Integrate all version trees to generate a version forest.

[0092] By integrating the version trees corresponding to each category, a version forest is generated. The version forest is a collection of version trees that can intuitively show the change process of each knowledge point, and can also quickly retrieve knowledge points.

[0093] In Example 7, unlike Example 1, the method further includes:

[0094] The audio data is input into the semantic analysis model, and the execution intensity of each segment is output, wherein the execution intensity includes at least: high, medium and low;

[0095] Delete the segments corresponding to execution intensities of medium and low.

[0096] The audio data is input into the semantic analysis model to determine the execution intensity of each segment, which includes high, medium and low. The segments corresponding to medium and low execution intensities are deleted. In other words, the knowledge points of the segments corresponding to medium and low execution intensities are not transcribed.

[0097] For example, one segment is: "I suggest adjusting the off-get off work time to 6 pm," which has a low execution intensity. Another segment is: "I have notified the HR department that starting tomorrow, all departments will uniformly adjust their off-get off work time to 6 pm," which has a high execution intensity.

[0098] Figure 6 This diagram illustrates the structural composition of a local knowledge base construction system applying artificial intelligence, as provided in an embodiment of the present invention. The local knowledge base construction system 1 applying artificial intelligence includes:

[0099] The determination module 11 is used to delineate the usage area of ​​the local knowledge base, select several source points, use the sound pickup devices pre-integrated in the source points to collect audio data, divide it into several segments, compare it with the preset voiceprint library, and determine the weight value of each segment.

[0100] The rollback module 12 is used to transcribe the segmented text and input it into a pre-built semantic analysis model to output keywords. The keywords are written into a preset template to obtain new knowledge points. The new knowledge points are compared with the local knowledge base to obtain target items. It is determined whether the similarity between the new knowledge points and the target items is greater than a threshold. If so, the target items are overwritten. If not, the overwriting process is recorded, change data is generated, the target object is extracted from the new knowledge points and target items, the change data is sent to the target object's device terminal, and rollback permission is granted.

[0101] Module 13 is used to cluster the local knowledge base into several categories, create parent nodes corresponding one-to-one with the categories, attach pre-built child nodes to the parent nodes, write target items to the child nodes, integrate the parent nodes and child nodes, generate a multi-level version tree, link the local knowledge base, set the order of the multi-level version tree, split off new child nodes when the number of changed data is greater than the order, update the multi-level version tree, and embed a linkage calling mechanism in parallel with the multi-level version tree and the local knowledge base.

[0102] Figure 7 This diagram illustrates the structural composition of a local knowledge base construction system using artificial intelligence, as provided in an embodiment of the present invention. The determining module 11 includes:

[0103] The comparison unit 111 is used to compare the segments and the voiceprint library, determine the main sound source of each segment, and create a comparison table, wherein the comparison table consists of a main sound source item and a weight value item.

[0104] Unit 112 is established to establish the correspondence between segments and weight values ​​via the sound source body;

[0105] The construction unit 113 is used to construct a microphone array using the sound pickup device, generate a sound recording mode, and embed it into the microphone array, wherein the sound recording mode includes at least: a discussion mode and a narration mode;

[0106] The adjustment unit 114 is used to adjust the weight value via the sound source body.

[0107] Figure 8 This diagram illustrates the structural composition of a local knowledge base construction system using artificial intelligence, as provided in an embodiment of the present invention. The rollback module 12 includes:

[0108] The calculation unit 121 is used to divide the local knowledge base into several existing knowledge points and use a keyword matching algorithm to perform matching calculations on the new knowledge points and existing knowledge points.

[0109] Definition unit 122 is used to find the existing knowledge point with the highest matching degree based on the newly added knowledge point, and define it as the target item;

[0110] The writing unit 123 is used to construct a selection window and write the changed data into the selection window;

[0111] Embedding unit 124 is used to embed accept, modification suggestion and rejection buttons into the selection window.

[0112] Figure 9 This diagram illustrates the structural composition of a local knowledge base construction system using artificial intelligence, as provided in an embodiment of the present invention. The calling module 13 includes:

[0113] Index unit 131 is used to synchronize the target items in the new child node to the parent node, extract the summary, and build an indexing mechanism;

[0114] Generation unit 132 is used to integrate all version trees and generate a version forest.

[0115] The determination module 11 is mainly used to complete step S100, the rollback module 12 is mainly used to complete step S200, and the calling module 13 is mainly used to complete step S300.

[0116] The comparison unit 111 is mainly used to complete step S101, the establishment unit 112 is mainly used to complete step S102, the construction unit 113 is mainly used to complete step S103, and the adjustment unit 114 is mainly used to complete step S104.

[0117] The calculation unit 121 is mainly used to complete step S201, the definition unit 122 is mainly used to complete step S202, the writing unit 123 is mainly used to complete step S203, and the embedding unit 124 is mainly used to complete step S204.

[0118] The indexing unit 131 is mainly used to complete step S301, and the generating unit 132 is mainly used to complete step S302.

[0119] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a local knowledge base using artificial intelligence, characterized in that, The method includes: The local knowledge base usage area is defined, several source points are selected, and audio data is collected using the audio pickup devices pre-integrated in the source points. The data is then divided into several segments, compared with the preset voiceprint database, and the weight value of each segment is determined. The segments are transcribed and input into a pre-built semantic analysis model to output keywords. The keywords are written into a preset template to obtain new knowledge points. The new knowledge points are compared with the local knowledge base to obtain target items. It is determined whether the similarity between the new knowledge points and the target items is greater than a threshold. If yes, the target items are covered. If no, the covering process is recorded, change data is generated, the target object is extracted from the new knowledge points and target items, the change data is sent to the target object's device terminal, and rollback permission is granted. The local knowledge base is clustered into several categories, parent nodes corresponding one-to-one with each category are created, pre-built child nodes are attached to the parent nodes, target items are written into the child nodes, the parent nodes and child nodes are integrated to generate a multi-level version tree, the local knowledge base is linked, the order of the multi-level version tree is set, when the number of changed data is greater than the order, new child nodes are split off, the multi-level version tree is updated, and a linkage call mechanism is embedded in the multi-level version tree and the local knowledge base in parallel. The step of comparing the data with a preset voiceprint database to determine the weight value of each segment includes: By comparing the segments with the voiceprint database, the main sound source of each segment is determined, and a lookup table is created, wherein the lookup table consists of a main sound source item and a weight value item. By establishing the correspondence between segments and weight values ​​through the main sound source; The steps of collecting audio data using a pre-integrated pickup device at the source location and dividing it into several segments include: Using the aforementioned sound pickup device, a microphone array is constructed, a sound recording mode is generated, and embedded into the microphone array, wherein the sound recording mode includes at least: a discussion mode and a speaking mode; The weight values ​​are adjusted via the sound source.

2. The method for constructing a local knowledge base using artificial intelligence according to claim 1, characterized in that, The method further includes: The audio data is input into the semantic analysis model, and the execution intensity of each segment is output, wherein the execution intensity includes at least: high, medium and low; Delete the segments corresponding to execution intensities of medium and low.

3. The method for constructing a local knowledge base using artificial intelligence according to claim 1, characterized in that, The step of comparing newly added knowledge points with the local knowledge base to obtain the target item includes: The local knowledge base is divided into several existing knowledge points, and a keyword matching algorithm is used to match and calculate the new knowledge points and existing knowledge points. Based on the newly added knowledge points, the existing knowledge points with the highest matching degree are found and defined as target items.

4. The method for constructing a local knowledge base using artificial intelligence according to claim 1, characterized in that, The steps of extracting the target object from the newly added knowledge points and target items, sending the changed data to the target object's device terminal, and granting rollback permissions include: Construct a selection window and write the changed data into the selection window; Embed "Accept," "Modify Suggestion," and "Reject" buttons in the selection window.

5. The method for constructing a local knowledge base using artificial intelligence according to claim 1, characterized in that, The step of splitting out new child nodes and updating the multi-level version tree includes: The target items in the new child nodes are synchronized to the parent node, and the summary is extracted to build the index mechanism; Integrate all version trees to generate a version forest.

Images