Enterprise-level multi-dimensional information integration and data retrieval system

Abstract

The invention discloses an enterprise-level multi-dimensional information integration and data retrieval system, which belongs to the technical field of data processing, and is characterized in that structural data and non-structural data in enterprise-level multi-dimensional information are enabled to have relatively uniform data features through integration and feature extraction of the enterprise-level multi-dimensional information. According to the method, the integration degree of multi-dimensional information can be effectively improved, meanwhile, the data feature index database is constructed according to the data features corresponding to each piece of information, matching can be conducted in the data feature index database according to the enterprise data retrieval request, target data features are obtained, and therefore target information corresponding to the target data features is determined; according to the method, the retrieval efficiency of enterprise-level multi-dimensional information can be effectively improved, the method is not limited to the types of enterprise data retrieval requests, retrieval of information of various dimensions can be achieved, and the data retrieval flexibility is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and particularly relates to an enterprise-level multi-dimensional information integration and data retrieval system. Background Art

[0002] Enterprise-level multi-dimensional information refers to the data and knowledge collected, integrated, and analyzed from multiple perspectives and levels during the operation of an enterprise. It covers multiple fields such as market, finance, human resources, production, and customer feedback, including structured data, unstructured data, and semi-structured data. Through the integration and analysis of multi-dimensional information, an enterprise can more comprehensively understand its own operation status, market environment, and customer needs, thereby making more accurate decisions and enhancing its competitiveness. This information management method emphasizes the relevance, real-time nature, and visualization of data, providing strong data support and decision-making basis for the enterprise. With the rapid development of information technology, enterprises are facing a vast amount of data and information. How to effectively integrate this multi-source and heterogeneous data and provide efficient and accurate data retrieval services has become an important means for enterprises to enhance their competitiveness. Existing information integration and data retrieval systems often suffer from problems such as low integration degree, low retrieval efficiency, and inability to meet multi-dimensional query requirements. Summary of the Invention

[0003] The present invention provides an enterprise-level multi-dimensional information integration and data retrieval system to solve the problems of low data integration degree and low retrieval efficiency in the prior art.

[0004] An enterprise-level multi-dimensional information integration and data retrieval system includes: a multi-dimensional information acquisition module, a data feature extraction module, a data storage module, a data request module, a data retrieval module, and a data feedback module;

[0005] The multi-dimensional information acquisition module is used to acquire enterprise-level multi-dimensional information; among them, the enterprise-level multi-dimensional information includes the text information, audio information, video information, and image information of the enterprise;

[0006] The data feature extraction module is used to extract the data features of each item of information in the enterprise-level multi-dimensional information, including text information, audio information, video information, and image information, to obtain the data features corresponding to each item of information;

[0007] The data storage module is used to construct a data feature index library according to the data features corresponding to each item of information, and establish an association relationship between the data feature index library and the storage address of the corresponding information;

[0008] The data request module is used to obtain an enterprise data retrieval request input by the user through human-computer interaction; among them, the enterprise data retrieval request includes at least one vocabulary;

[0009] The data retrieval module is used to match in the data feature index library according to the enterprise data retrieval request, obtain target data features, and determine the association relationships related to the target data features;

[0010] The data feedback module is used to schedule the target information in the corresponding storage address based on the association relationships related to the target data features, and display the target information to the user, completing enterprise-level multi-dimensional information integration and data retrieval.

[0011] Further, obtaining enterprise-level multi-dimensional information includes: obtaining text information, audio information, video information, and image information from a specified data source or input by staff to obtain enterprise-level multi-dimensional information.

[0012] Further, for the text information, audio information, video information, and image information in the enterprise-level multi-dimensional information, extracting the data features of each piece of information to obtain the data features corresponding to each piece of information, including:

[0013] For the text information in the enterprise-level multi-dimensional information, determining whether the length of the text information exceeds a preset threshold. If so, determining the text information as the first text information; otherwise, determining the text information as the second text information;

[0014] Extracting the entity data and the relationships between the entity data in each first sentence of the first text information, and obtaining the word vectors of the entity data and the relationships between the entity data corresponding to each first sentence, to obtain at least one target vector corresponding to the first text information, and obtaining the data features corresponding to the first text information;

[0015] Using a text summary extraction algorithm to extract the summary data in the second text information, and extracting the entity data and the relationships between the entity data in each second sentence of the summary data, and obtaining the word vectors of the entity data and the relationships between the entity data corresponding to each second sentence, to obtain at least one target vector corresponding to the second text information, and obtaining the data features corresponding to the second text information;

[0016] For the audio information in the enterprise-level multi-dimensional information, extracting the first speech text information in the audio information, and extracting the entity data and the relationships between the entity data in each third sentence of the first speech text information, and obtaining the relationship word vectors of the entity data and the relationships between the entity data corresponding to each third sentence, to obtain at least one target vector corresponding to the audio information, and obtaining the first data features corresponding to the audio information; at the same time, extracting the name and time information of the audio information to construct data features, and obtaining the second data features corresponding to the audio information;

[0017] For the video information in enterprise-level multi-dimensional information, perform frame extraction on the video information to obtain multiple video frames, and use a deep learning model to extract the feature vectors corresponding to the video frames to obtain the first target data features corresponding to the video information;

[0018] For the video information in enterprise-level multi-dimensional information, extract the second speech and text information of the video information, and extract the entity data of each fourth sentence in the second speech and text information and the relationships between the entity data, and obtain the word vectors of the entity data and the relationships between the entity data corresponding to each fourth sentence, to obtain at least one target vector corresponding to the video information, and obtain the second data features corresponding to the video information; at the same time, extract the name and time information of the video information to construct data features, and obtain the third data features corresponding to the video information;

[0019] For the image information in enterprise-level multi-dimensional information, use a deep learning model to extract the feature vectors corresponding to the image information to obtain the first data features corresponding to the image information; at the same time, extract the name and time information of the image information to construct data features, and obtain the second data features corresponding to the image information.

[0020] Further, the deep learning model is set as a Transformer model or a CNN model, and the softmax layer is removed from the Transformer model or the CNN model.

[0021] Further, according to the data features corresponding to each piece of information, construct a data feature index library, and establish an association relationship between the data feature index library and the storage address of the corresponding information, including:

[0022] Store the text information, audio information, video information, and image information in the enterprise-level multi-dimensional information into the database, and obtain the storage address corresponding to each piece of information;

[0023] Construct a data feature index library according to the data features corresponding to each piece of information;

[0024] For any one of the data features in the data feature index library, associate the data feature with the storage address of the information corresponding to the data feature to obtain the association relationship between the data feature index library and the storage address of the corresponding information.

[0025] Further, according to the enterprise data retrieval request, perform matching in the data feature index library to obtain the target data features, including:

[0026] Obtain the similarity between the enterprise data retrieval request and the data features in the data feature index library, and take the top N data features from largest to smallest similarity to obtain the target data features.

[0027] Further, obtaining the similarity between the enterprise data retrieval request and the data features in the data feature index library includes:

[0028] When the enterprise data retrieval request is text data, the entity data in the enterprise data retrieval request and the word vectors corresponding to the relationships between the entity data are obtained, and the word vector corresponding to the enterprise data retrieval request is obtained;

[0029] The cosine similarity between the word vector corresponding to the enterprise data retrieval request and the data features in the data feature index library is obtained, and the similarity between the enterprise data retrieval request and the data features in the data feature index library is obtained.

[0030] Further, obtaining the similarity between the enterprise data retrieval request and the data features in the data feature index library includes:

[0031] When the enterprise data retrieval request is voice data, the voice text information in the voice data is extracted, and the entity data and the relationships between the entity data in each sentence of the voice text information in the voice data are extracted, and word vectors are constructed to obtain the word vector corresponding to the enterprise data retrieval request;

[0032] The cosine similarity between the word vector corresponding to the enterprise data retrieval request and the data features in the data feature index library is obtained, and the similarity between the enterprise data retrieval request and the data features in the data feature index library is obtained.

[0033] Further, obtaining the similarity between the enterprise data retrieval request and the data features in the data feature index library includes:

[0034] When the enterprise data retrieval request is image data, a deep learning model is used to extract the feature vector corresponding to the image data, and the feature vector corresponding to the enterprise data retrieval request is obtained;

[0035] The cosine similarity between the feature vector corresponding to the enterprise data retrieval request and the data features in the data feature index library is obtained, and the similarity between the enterprise data retrieval request and the data features in the data feature index library is obtained.

[0036] Further, based on the association relationship related to the target data feature, the target information in the corresponding storage address is scheduled and the target information is displayed to the user, including:

[0037] Based on the association relationship related to the target data feature, the target storage address is determined;

[0038] Based on the target storage address, the target information is scheduled, and after the target information is encrypted, it is securely transmitted to the user for display.

[0039] An enterprise-level multi-dimensional information integration and data retrieval system provided by the present invention integrates and extracts features from enterprise-level multi-dimensional information, enabling structured and unstructured data in the enterprise-level multi-dimensional information to have relatively unified data features, effectively improving the integration degree of multi-dimensional information. At the same time, a data feature index library is constructed based on the data features corresponding to each piece of information, and the target data features can be obtained by matching in the data feature index library according to the enterprise data retrieval request, thereby determining the target information corresponding to the target data features, effectively improving the retrieval efficiency of enterprise-level multi-dimensional information, and not limited to the type of enterprise data retrieval request, enabling the retrieval of various dimensional information and improving the flexibility of data retrieval. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings herein are incorporated into and constitute a part of this specification, showing embodiments consistent with the present invention and, together with the specification, are used to explain the principles of the present invention.

[0041] Figure 1 FIG. is a schematic structural diagram of an enterprise-level multi-dimensional information integration and data retrieval system provided by an embodiment of the present invention.

[0042] Figure 2 FIG. is a schematic flow diagram of obtaining data features corresponding to each piece of information provided by an embodiment of the present invention.

[0043] Through the above accompanying drawings, specific embodiments of the present invention have been shown, and there will be more detailed descriptions hereinafter. These drawings and textual descriptions are not intended to limit the scope of the inventive concept in any way, but to illustrate the concept of the present invention to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] Here, the exemplary embodiments will be described in detail, and the examples are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0045] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0046] As Figure 1 shown, an embodiment of the present invention provides an enterprise-level multi-dimensional information integration and data retrieval system, including: a multi-dimensional information acquisition module 11, a data feature extraction module 12, a data storage module 13, a data request module 14, a data retrieval module 15, and a data feedback module 16;

[0047] The multi-dimensional information acquisition module 11 is used to acquire enterprise-level multi-dimensional information; among them, the enterprise-level multi-dimensional information includes the text information, audio information, video information, and image information of the enterprise;

[0048] Enterprise-level multi-dimensional information refers to various types of data and knowledge collected, integrated, and analyzed from multiple perspectives, levels, and dimensions in the internal and external environments of an enterprise. It includes, but is not limited to, information in multiple aspects such as market dynamics, financial status, human resources, production operations, customer relationships, and supply chain management. This information exists in structured, unstructured, and semi-structured forms, and is integrated and analyzed through advanced information technology means to provide comprehensive, in-depth, and real-time data insights for the enterprise, support enterprise strategic planning, operation management, and decision-making, and enhance the competitiveness and market response ability of the enterprise.

[0049] Therefore, the embodiments of the present invention mainly collect text information, audio information, video information, and image information to integrate enterprise data and improve the retrieval efficiency of enterprise multi-dimensional data.

[0050] The data feature extraction module 12 is used to extract the data features of each item of information, namely text information, audio information, video information, and image information, in the enterprise-level multi-dimensional information, and obtain the data features corresponding to each item of information;

[0051] Since the forms of text information, audio information, video information, and image information are not unified and may have structured data and unstructured data, it is difficult to store and retrieve the data uniformly. Therefore, the embodiments of the present invention extract the data features of text information, audio information, video information, and image information respectively, so that all information can be retrieved, and the integration degree and retrievability of enterprise-level multi-dimensional information are improved.

[0052] The data storage module 13 is used to construct a data feature index library according to the data features corresponding to each item of information, and establish an association relationship between the data feature index library and the storage address of the corresponding information;

[0053] The data feature index library refers to a database that stores the data features of all information. At the same time, there is an association relationship between the data features in this library and the storage addresses of the corresponding information (for example, a data association table can be established, and the storage address corresponding to the data feature can be determined by looking up the table), so that enterprise-level multi-dimensional information can be easily retrieved.

[0054] The data request module 14 is used to obtain an enterprise data retrieval request input by the user through human-computer interaction; among them, the enterprise data retrieval request includes at least one vocabulary;

[0055] An enterprise data retrieval request refers to some conditions given by users. These conditions often have entities and entity relationships, thus forming query information, or only a certain entity data exists, so that retrieval can be performed based on this data.

[0056] The data retrieval module 15 is used to match in the data feature index library according to the enterprise data retrieval request, obtain target data features, and determine the association relationships related to the target data features;

[0057] The method of similarity matching can be used to match the enterprise data retrieval request with the data features in the data feature index library, so as to determine the target information closest to the enterprise data retrieval request.

[0058] The data feedback module 16 is used to schedule the target information in the corresponding storage address based on the association relationships related to the target data features, and display the target information to the user, completing enterprise-level multi-dimensional information integration and data retrieval.

[0059] An enterprise-level multi-dimensional information integration and data retrieval system provided by the present invention integrates and extracts features through enterprise-level multi-dimensional information, so that the structured data and unstructured data in the enterprise-level multi-dimensional information both have relatively unified data features, which can effectively improve the integration degree of multi-dimensional information. At the same time, a data feature index library is constructed according to the data features corresponding to each piece of information. It can match in the data feature index library according to the enterprise data retrieval request, obtain target data features, and thus determine the target information corresponding to the target data features, which can effectively improve the retrieval efficiency of enterprise-level multi-dimensional information, and is not limited to the type of enterprise data retrieval request, and can realize the retrieval of various dimensional information, improving the flexibility of data retrieval.

[0060] In the embodiment of the present invention, obtaining enterprise-level multi-dimensional information includes: obtaining specified data sources or text information, audio information, video information, and image information input by staff to obtain enterprise-level multi-dimensional information.

[0061] As Figure 2 shown, for the text information, audio information, video information, and image information in the enterprise-level multi-dimensional information, extracting the data features of each piece of information to obtain the data features corresponding to each piece of information includes:

[0062] S21. For the text information in the enterprise-level multi-dimensional information, determine whether the length of the text information exceeds a preset threshold. If so, determine the text information as the first text information; otherwise, determine the text information as the second text information;

[0063] S22. Extract the entity data of each first sentence in the first text information and the relationships between the entity data, and obtain the word vectors of the entity data and the relationships between the entity data corresponding to each first sentence, so as to obtain at least one target vector corresponding to the first text information, and obtain the data features corresponding to the first text information;

[0064] Generally, a sentence is composed of entities, the relationships between entities, and entities. Then, the word vectors corresponding to the entities, the relationships between entities, and the sentences composed of entities can be obtained, so that the data features corresponding to the text data can be obtained;

[0065] S23. Use the text summarization extraction algorithm to extract the summary data in the second text information, extract the entity data of each second sentence in the summary data and the relationships between the entity data, and obtain the word vectors of the entity data and the relationships between the entity data corresponding to each second sentence, so as to obtain at least one target vector corresponding to the second text information, and obtain the data features corresponding to the second text information;

[0066] For long texts, it makes data retrieval more complex. Therefore, in the embodiments of the present invention, the text summarization extraction algorithm is first used to extract the summary data in the second text information, and then the corresponding data features are extracted, which can effectively reduce the data dimension and enable long texts to be retrieved at the same time.

[0067] S24. For the audio information in the enterprise-level multi-dimensional information, extract the first speech text information in the audio information, extract the entity data of each third sentence in the first speech text information and the relationships between the entity data, and obtain the word vectors of the entity data and the relationships between the entity data corresponding to each third sentence, so as to obtain at least one target vector corresponding to the audio information, and obtain the first data features corresponding to the audio information;

[0068] At the same time, extract the name and time information of the audio information to construct data features, and obtain the second data features corresponding to the audio information;

[0069] In the embodiments of the present invention, extracting the name and time information of the audio information to construct the second data features corresponding to the audio information can not only realize the retrieval of the specific content in the audio information, but also realize the retrieval of the audio information on a certain date or in a certain scenario, greatly improving the flexibility of data retrieval.

[0070] S25. For the video information in the enterprise-level multi-dimensional information, perform frame extraction on the video information to obtain multiple video frames, and use a deep learning model to extract the feature vectors corresponding to the video frames, so as to obtain the first target data features corresponding to the video information;

[0071] For the video information in enterprise-level multi-dimensional information, extract the second speech text information of the video information, extract the entity data of each fourth sentence in the second speech text information and the relationships between the entity data, and obtain the word vectors of the entity data corresponding to each fourth sentence and the relationships between the entity data, to obtain at least one target vector corresponding to the video information, and obtain the second data feature corresponding to the video information;

[0072] At the same time, extract the name and time information of the video information to construct data features, and obtain the third data feature corresponding to the video information;

[0073] The embodiments of the present invention extract three types of feature data of the video information in enterprise-level multi-dimensional information, greatly improving the retrieval flexibility. Users can retrieve video data from multiple aspects, thereby achieving rapid positioning of data.

[0074] S26. For the image information in enterprise-level multi-dimensional information, use a deep learning model to extract the feature vector corresponding to the image information, and obtain the first data feature corresponding to the image information;

[0075] At the same time, extract the name and time information of the image information to construct data features, and obtain the second data feature corresponding to the image information.

[0076] The embodiments of the present invention extract the name and time information of the image information to construct data features, and obtain the second data feature corresponding to the image information, which can not only realize the retrieval of the specific content in the image information, but also realize the retrieval of the image information on a certain date or in a certain scenario, greatly improving the data retrieval flexibility.

[0077] In the embodiments of the present invention, the deep learning model is set as a Transformer model or a CNN (Convolutional Neural Network) model, and the softmax layer is removed from the Transformer model or the CNN model.

[0078] Both the Transformer model and the CNN model have the ability to process image data. Therefore, the Transformer model or the CNN model can be used to extract data features, making the image data easy to be retrieved, enhancing the data positioning ability, and improving the user's ability to trace the data source. For example, after storing the data set, users can hold a certain image to retrieve the corresponding data set, making the data easier to be retrieved and traced, and improving the data retrieval ability.

[0079] Optionally, the Transformer model and the CNN model can also be trained first to improve the data feature extraction ability. For example, the Transformer model and the CNN model can be trained first using the image information and the artificial labels corresponding to the image information, and then the softmax layer of the Transformer model or the CNN model can be removed, so that when these two models process the image data, they can output data features, which is convenient for data retrieval.

[0080] In the embodiments of the present invention, according to the data features corresponding to each piece of information, a data feature index library is constructed, and an association relationship is established between the data feature index library and the storage address of the corresponding information, including:

[0081] The text information, audio information, video information, and image information in the enterprise-level multi-dimensional information are stored in the database, and the storage address corresponding to each piece of information is obtained;

[0082] According to the data features corresponding to each piece of information, a data feature index library is constructed;

[0083] For any one of the data features in the data feature index library, the data feature is associated with the storage address corresponding to the information corresponding to the data feature, so as to obtain the association relationship between the data feature index library and the storage address of the corresponding information.

[0084] In the embodiments of the present invention, according to the enterprise data retrieval request, a match is made in the data feature index library to obtain the target data feature, including:

[0085] Obtain the similarity between the enterprise data retrieval request and the data features in the data feature index library, and take the top N data features from largest to smallest similarity to obtain the target data feature.

[0086] In the embodiments of the present invention, obtaining the similarity between the enterprise data retrieval request and the data features in the data feature index library includes:

[0087] When the enterprise data retrieval request is text data, the entity data in the enterprise data retrieval request and the word vectors corresponding to the relationships between the entity data are obtained to obtain the word vector corresponding to the enterprise data retrieval request;

[0088] Obtain the cosine similarity between the word vector corresponding to the enterprise data retrieval request and the data features in the data feature index library to obtain the similarity between the enterprise data retrieval request and the data features in the data feature index library.

[0089] It should be noted that date data should also be recognized as entity data in the embodiments of the present invention. Generally, image naming adopts entity naming, so that users can retrieve according to the date (i.e., entity data) plus the image name (i.e., entity data). Since some data features are obtained through names and time information, after date data appears, the word vectors of enterprise data retrieval requests can be directly used for matching, so as to realize data retrieval of name plus time, and further improve data flexibility.

[0090] In the actual implementation process, problems generally only include entities and entity relationships (for example, the address of Company A is in Place B, so the problem is generally where is the address of Company A. Therefore, the embodiments of the present invention only consider the case where the length of the problem is shorter than the length of the data, and do not consider other cases). It is often possible that the length of the problem is shorter than the length of the data, and it is difficult to directly obtain the cosine similarity. Therefore, it is necessary to first supplement the word vector corresponding to the problem. The specific methods may include:

[0091] The word vector corresponding to the enterprise data retrieval request can be head-to-head aligned with the data features in the data feature index library, and then the word vector corresponding to the enterprise data retrieval request can be shifted bit by bit to the left or right to find the position with the largest overlap between the word vector corresponding to the enterprise data retrieval request and the data features in the data feature index library. Then, the positions with vacancies on both sides of the word vector corresponding to the enterprise data retrieval request are filled with zeros, so that the lengths of the word vector of the problem and the word vector of the data feature are the same. For example, assuming a data feature is [2, 3, 1, 1, 1, 5, 2], and the word vector corresponding to an enterprise data retrieval request is [1, 1, 1], then in the process of matching this data feature with the word vector corresponding to the enterprise data retrieval request, the word vector corresponding to the enterprise data retrieval request can be processed as [0, 0, 1, 1, 1, 0, 0], so as to obtain the cosine similarity.

[0092] Optionally, when extracting data features, a delimiter can be used to separate entity data from relationship data between entities. Therefore, in the matching process, the target entity carried in the enterprise data retrieval request can be queried, and then the data features carrying the target entity can be found. Finally, the cosine similarity corresponding to the entity relationship in the enterprise data retrieval request and the relationship between entities in the data feature can be obtained, so as to determine the similarity between the problem and the data feature.

[0093] For example, in this embodiment, the enterprise data retrieval request is converted into an internal representation that can capture semantics. For example, for the input natural language question "What is the capital of the United States", first, the central entity in the query statement can be found as the United States through the entity recognition model; then, the category corresponding to the entity (i.e., country) can be queried through the entity conceptualization mechanism; finally, the entity in the question is replaced with its corresponding category, so that the internal representation of the question "&country&'s capital" can be obtained, where "&country&" represents a wildcard, and as long as the entity is a country, it can be matched, thus enabling entity matching. After matching the data features with the target entity, the similarity can be obtained.

[0094] In the embodiment of the present invention, obtaining the similarity between the enterprise data retrieval request and the data features in the data feature index library includes:

[0095] When the enterprise data retrieval request is voice data, the voice text information in the voice data is extracted, and the entity data in each sentence of the voice text information in the voice data and the relationship between the entity data are extracted, and a word vector is constructed to obtain the word vector corresponding to the enterprise data retrieval request;

[0096] The cosine similarity between the word vector corresponding to the enterprise data retrieval request and the data features in the data feature index library is obtained to obtain the similarity between the enterprise data retrieval request and the data features in the data feature index library.

[0097] The voice data needs to be converted into text information for recognition and matching. After being converted into text information, the subsequent processing is the same as the above text information matching process, and will not be elaborated here.

[0098] In the embodiment of the present invention, obtaining the similarity between the enterprise data retrieval request and the data features in the data feature index library includes:

[0099] When the enterprise data retrieval request is image data, a deep learning model is used to extract the feature vector corresponding to the image data to obtain the feature vector corresponding to the enterprise data retrieval request;

[0100] The cosine similarity between the feature vector corresponding to the enterprise data retrieval request and the data features in the data feature index library is obtained to obtain the similarity between the enterprise data retrieval request and the data features in the data feature index library.

[0101] Since the lengths of the image feature vectors extracted by the deep learning model are all the same, when the enterprise data retrieval request is image data, the data features in the image information or video information can be directly matched, so as to achieve fast data positioning and query, and thus achieve multifunctional data retrieval.

[0102] In an embodiment of the present invention, based on the association relationship related to the target data feature, the target information in the corresponding storage address is scheduled and the target information is presented to the user, including:

[0103] Based on the association relationship related to the target data feature, determine the target storage address;

[0104] Based on the target storage address, schedule the target information, and after encrypting the target information, securely transmit it to the user for display. For example, the public-private key encryption algorithm can be used to encrypt the target information.

[0105] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0106] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the specified functions in one Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0107] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the specified functions in one Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0108] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are performed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the specified functions in one Figure 1One process or multiple processes and / or boxes Figure 1 Steps of functions specified in one box or multiple boxes.

[0109] Those of ordinary skill in the art can understand that all or part of the steps in implementing the above facts and methods can be completed by instructing relevant hardware through a program. The involved program or the described program can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: At this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disc, etc.

[0110] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An enterprise-level multi-dimensional information integration and data retrieval system, characterized in that: include: Multi-dimensional information acquisition module, data feature extraction module, data storage module, data request module, data retrieval module and data feedback module; The multi-dimensional information acquisition module is used to acquire enterprise-level multi-dimensional information; wherein the enterprise-level multi-dimensional information includes text information, audio information, video information and image information of the enterprise; The data feature extraction module is used to extract data features of each item of information, including text information, audio information, video information and image information in the enterprise-level multi-dimensional information, and obtain data features corresponding to each item of information; The data storage module is used to construct a data feature index library according to the data features corresponding to each piece of information, and to establish an association relationship between the data feature index library and the storage address of the corresponding information; The data request module is used to obtain an enterprise data search request input by a user through human-computer interaction; wherein the enterprise data search request includes at least one vocabulary; The data retrieval module is used to match the data feature index library according to the enterprise data retrieval request, obtain the target data features, and determine the association relationship related to the target data features; The data feedback module is used to schedule the target information in the corresponding storage address based on the association relationship related to the target data characteristics, and display the target information to the user, thereby completing enterprise-level multi-dimensional information integration and data retrieval.

2. The enterprise-level multi-dimensional information integration and data retrieval system according to claim 1, characterized in that: Acquiring enterprise-level multi-dimensional information includes: acquiring text information, audio information, video information, and image information from a designated data source or input by a staff member to obtain enterprise-level multi-dimensional information.

3. The enterprise-level multi-dimensional information integration and data retrieval system according to claim 1, characterized in that: For the text information, audio information, video information and image information in the enterprise-level multi-dimensional information, extract the data features of each information to obtain the data features corresponding to each information, including: For the text information in the enterprise-level multi-dimensional information, determine whether the length of the text information exceeds a preset threshold, and if so, determine that the text information is the first text information, otherwise, determine that the text information is the second text information; Extracting entity data and the relationship between the entity data of each first sentence in the first text information, and obtaining word vectors of the entity data and the relationship between the entity data corresponding to each first sentence, obtaining at least one target vector corresponding to the first text information, and obtaining data features corresponding to the first text information; Extract summary data from the second text information using a text summary extraction algorithm, extract entity data of each second sentence in the summary data and the relationship between the entity data, obtain word vectors of the entity data corresponding to each second sentence and the relationship between the entity data, obtain at least one target vector corresponding to the second text information, and obtain data features corresponding to the second text information; For the audio information in the enterprise-level multi-dimensional information, extract the first voice and text information in the audio information, extract the entity data of each third sentence in the first voice and text information and the relationship between the entity data, and obtain the entity data corresponding to each third sentence and the relationship word vector between the entity data, obtain at least one target vector corresponding to the audio information, and obtain the first data feature corresponding to the audio information; at the same time, extract the name and time information of the audio information to construct the data feature, and obtain the second data feature corresponding to the audio information; For the video information in the enterprise-level multi-dimensional information, the video information is subjected to frame extraction processing to obtain multiple video frames, and a deep learning model is used to extract feature vectors corresponding to the video frames to obtain first target data features corresponding to the video information; For the video information in the enterprise-level multi-dimensional information, extract the second voice and text information of the video information, extract the entity data of each fourth sentence in the second voice and text information and the relationship between the entity data, and obtain the word vectors of the entity data corresponding to each fourth sentence and the relationship between the entity data, obtain at least one target vector corresponding to the video information, and obtain the second data feature corresponding to the video information; at the same time, extract the name and time information of the video information to construct the data feature, and obtain the third data feature corresponding to the video information; For the image information in the enterprise-level multi-dimensional information, a deep learning model is used to extract the feature vector corresponding to the image information to obtain the first data feature corresponding to the image information; at the same time, the name and time information of the image information are extracted to construct data features to obtain the second data feature corresponding to the image information.

4. The enterprise-level multi-dimensional information integration and data retrieval system according to claim 3, characterized in that: The deep learning model is set to a Transformer model or a CNN model, and the Transformer model or the CNN model removes the softmax layer.

5. The enterprise-level multi-dimensional information integration and data retrieval system according to claim 3, characterized in that: According to the data features corresponding to each piece of information, a data feature index library is constructed, and an association relationship between the data feature index library and the storage address of the corresponding information is established, including: Store the text information, audio information, video information and image information in the enterprise-level multi-dimensional information into a database, and obtain the storage address corresponding to each piece of information; Build a data feature index library based on the data features corresponding to each piece of information; For any data feature in the data feature index library, the data feature is associated with a storage address corresponding to the information corresponding to the data feature to obtain an association relationship between the data feature index library and the storage address of the corresponding information.

6. The enterprise-level multi-dimensional information integration and data retrieval system according to claim 5, characterized in that: According to the enterprise data retrieval request, matching is performed in the data feature index library to obtain target data features, including: The similarity between the enterprise data retrieval request and the data features in the data feature index library is obtained, and the top N data features are taken from large to small according to the similarity to obtain the target data features.

7. The enterprise-level multi-dimensional information integration and data retrieval system according to claim 6, characterized in that: Obtaining the similarity between the enterprise data retrieval request and the data features in the data feature index library includes: When the enterprise data search request is text data, the word vectors corresponding to the entity data and the relationship between the entity data in the enterprise data search request are obtained to obtain the word vector corresponding to the enterprise data search request; The cosine similarity between the word vector corresponding to the enterprise data retrieval request and the data features in the data feature index library is obtained, and the similarity between the enterprise data retrieval request and the data features in the data feature index library is obtained.

8. The enterprise-level multi-dimensional information integration and data retrieval system according to claim 7, characterized in that: Obtaining the similarity between the enterprise data retrieval request and the data features in the data feature index library includes: When the enterprise data retrieval request is voice data, the voice text information in the voice data is extracted, and the entity data of each sentence in the voice text information in the voice data and the relationship between the entity data are extracted, and a word vector is constructed to obtain the word vector corresponding to the enterprise data retrieval request; The cosine similarity between the word vector corresponding to the enterprise data retrieval request and the data features in the data feature index library is obtained, and the similarity between the enterprise data retrieval request and the data features in the data feature index library is obtained.

9. The enterprise-level multi-dimensional information integration and data retrieval system according to claim 8, characterized in that: Obtaining the similarity between the enterprise data retrieval request and the data features in the data feature index library includes: When the enterprise data retrieval request is image data, a deep learning model is used to extract a feature vector corresponding to the image data to obtain a feature vector corresponding to the enterprise data retrieval request; The cosine similarity between the feature vector corresponding to the enterprise data retrieval request and the data features in the data feature index library is obtained to obtain the similarity between the enterprise data retrieval request and the data features in the data feature index library.

10. The enterprise-level multi-dimensional information integration and data retrieval system according to claim 7, characterized in that: Based on the association relationship related to the target data feature, the target information in the corresponding storage address is scheduled, and the target information is displayed to the user, including: Determining a target storage address based on an association relationship related to the target data feature; Based on the target storage address, the target information is scheduled, and after the target information is encrypted, it is securely transmitted to the user for display.

Images