Method and device for carrying out supply and demand content feature similarity calculation based on enterprise dictionary system, computer equipment and readable storage medium

Abstract

The invention discloses a supply and demand content feature similarity calculation method and device based on an enterprise dictionary system, computer equipment and a readable storage medium, and the method comprises the steps: firstly obtaining and extracting supply and demand basic data, integrating a feature vector obtained from a vector database with the basic data, constructing a document index, and storing the document index in an Elastic Search; when a target user clicks a supply and demand operation, a filtering strategy is adopted for preprocessing according to a clicked object to obtain an undetermined feature vector, then a candidate feature vector set is obtained through Elastic Search retrieval, and finally a preset distance formula is utilized for calculation and sorting to obtain a similarity calculation result. According to the method, accurate supply and demand content feature similarity calculation is realized by means of an enterprise dictionary system and integration of multi-dimensional information, and supply and demand matching efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method, device, computer equipment and readable storage medium for calculating the similarity of supply and demand content features based on an enterprise dictionary system. Background Art

[0002] In a company's supply and demand operations, the vast amount of information makes it difficult to accurately match the needs of both parties. Existing supply and demand matching technologies often lack in-depth data analysis and integration, making it difficult to accurately measure the similarity between supply and demand content. While some methods can perform simple information retrieval, they fail to fully utilize the company's internal standardized terminology system and lack comprehensive consideration of user needs and supply and demand characteristics. This results in low matching efficiency and accuracy, and fails to effectively meet the actual business needs of enterprises. Summary of the Invention

[0003] The object of the present invention is to provide a method, apparatus, computer equipment and readable storage medium for calculating the similarity of supply and demand content features based on an enterprise dictionary system.

[0004] In a first aspect, an embodiment of the present invention provides a method for calculating the similarity of supply and demand content features based on an enterprise dictionary system, comprising:

[0005] Obtaining and extracting basic supply and demand data, including supply and demand tags, supply and demand regions, supply and demand industries, publishers, and basic information of the publishers;

[0006] Acquire a characteristic vector of supply and demand from a vector database, and integrate the characteristic vector with the basic supply and demand data to obtain integrated supply and demand data; the vector database extracts the characteristic vector based on an enterprise dictionary system;

[0007] Constructing a document index for the integrated supply and demand data and storing it in Elastic Search;

[0008] In response to a target user's click-to-supply / demand operation, pre-processing the supply and demand information determined by the click-to-supply / demand operation using a corresponding filtering strategy based on the target object of the click-to-supply / demand operation to obtain a feature vector of the pending supply and demand content;

[0009] Performing a vector search based on tags, regions, and industries on the pending supply and demand content feature vectors through Elastic Search to obtain a set of candidate supply and demand content feature vectors;

[0010] A preset distance formula is used to calculate the feature similarity between each candidate supply and demand content feature vector included in the candidate supply and demand content feature vector set and the undetermined supply and demand content feature vector, and each candidate supply and demand content feature vector is sorted according to the feature similarity to obtain a supply and demand content feature similarity calculation result.

[0011] In a possible implementation, extracting the supply and demand basic data further includes:

[0012] Extract transaction scale and transaction time limit information from the supply and demand basic data.

[0013] In a possible implementation, obtaining the characteristic vector of supply and demand from the vector database and integrating the characteristic vector with the basic supply and demand data further includes:

[0014] The basic characteristics of the target user are integrated into the supply and demand basic data, wherein the basic characteristics of the user include the industry to which the user belongs, years of employment, and company size.

[0015] In a possible implementation, the target object based on the click supply and demand operation adopts a corresponding filtering strategy, including:

[0016] If the target object of the click supply and demand operation is the supply and demand information published by the target user himself, a first filtering strategy is adopted to remove duplicate or low-value self-related information;

[0017] If the target object of the click supply and demand operation is the supply and demand information published by other users, a second filtering strategy is adopted to filter out information that has a high degree of matching with the target user's needs.

[0018] In a possible implementation, before performing a vector search based on tags, regions, and industries on the feature vector of the pending supply and demand content through Elastic Search, the method further includes:

[0019] The undetermined supply and demand content feature vector is normalized to eliminate the dimensional effects of different feature dimensions.

[0020] In a possible implementation, after calculating the feature similarity using a preset distance formula, the method further includes:

[0021] The feature similarity calculation result is weighted according to the target user's historical click behavior data, wherein the historical click behavior data includes information such as click frequency, dwell time after clicking, whether a transaction is generated, etc.

[0022] In a possible implementation, the preset distance formula is a cosine distance formula.

[0023] In a second aspect, an embodiment of the present invention provides a device for calculating similarity between supply and demand content features based on an enterprise dictionary system, comprising:

[0024] An acquisition module is configured to acquire basic supply and demand data and extract the basic supply and demand data, including supply and demand tags, supply and demand regions, supply and demand industries, publishers, and basic publisher information; acquire a supply and demand feature vector from a vector database, and integrate the feature vector with the basic supply and demand data to obtain integrated supply and demand data; the vector database extracts feature vectors based on an enterprise dictionary system; constructs a document index from the integrated supply and demand data and stores it in Elastic Search; in response to a target user's click-on supply and demand operation, pre-processes the supply and demand information determined by the click-on supply and demand operation using a corresponding filtering strategy based on the target object of the click-on supply and demand operation to obtain a pending supply and demand content feature vector; and performs a vector search based on tags, regions, and industries on the pending supply and demand content feature vector through Elastic Search to obtain a set of candidate supply and demand content feature vectors.

[0025] a calculation module, configured to calculate, using a preset distance formula, a feature similarity between each candidate supply and demand content feature vector included in the candidate supply and demand content feature vector set and the undetermined supply and demand content feature vector, and to sort each of the candidate supply and demand content feature vectors according to the feature similarity to obtain a supply and demand content feature similarity calculation result.

[0026] In a third aspect, an embodiment of the present invention provides a computer device, comprising a processor and a non-volatile memory storing computer instructions, wherein when the computer instructions are executed by the processor, the computer device executes the method described in the first aspect.

[0027] In a fourth aspect, an embodiment of the present invention provides a readable storage medium, wherein the readable storage medium includes a computer program, and when the computer program is executed, the computer device where the readable storage medium is located is controlled to execute the method described in the first aspect.

[0028] Compared with the existing technology, the beneficial effects provided by the present invention include: the present invention discloses a method, device, computer equipment and readable storage medium for calculating the similarity of supply and demand content features based on the enterprise dictionary system, which obtains and extracts the basic data of supply and demand, integrates the feature vectors obtained from the vector database with the basic data, and constructs a document index stored in Elastic Search. When the target user clicks on the supply and demand operation, a filtering strategy is pre-processed based on the click object to obtain a pending feature vector, and then a set of candidate feature vectors is retrieved through Elastic Search. Finally, a preset distance formula is used to calculate and sort to obtain the similarity calculation result. With the help of the enterprise dictionary system, this method integrates multi-dimensional information to achieve accurate supply and demand content feature similarity calculation, thereby improving the efficiency and accuracy of supply and demand matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly describes the drawings required for use in the embodiments. It should be understood that the following drawings illustrate only certain embodiments of the present invention and should not be construed as limiting the scope of the present invention. Those skilled in the art can, without inventive effort, derive other relevant drawings from these drawings.

[0030] Figure 1 A schematic flow chart of the steps of a method for calculating the similarity of supply and demand content features based on an enterprise dictionary system provided by an embodiment of the present invention;

[0031] Figure 2 A schematic diagram of a framework for calculating the similarity of supply and demand content features based on an enterprise dictionary system provided by an embodiment of the present invention;

[0032] Figure 3 A schematic block diagram of the structure of an apparatus for calculating the similarity of supply and demand content features based on an enterprise dictionary system provided by an embodiment of the present invention;

[0033] Figure 4 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the embodiments of the present invention. It should be understood that the described embodiments are only a portion of the embodiments of the present invention, not all of them. Generally, the components of the embodiments of the present invention described and illustrated in the drawings herein may be arranged and designed in a variety of different configurations.

[0035] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0036] In order to solve the technical problems in the above background technology, Figure 1 This is a flow chart of a method for calculating the similarity of supply and demand content features based on an enterprise dictionary system provided in an embodiment of the present disclosure. The method for calculating the similarity of supply and demand content features based on an enterprise dictionary system is introduced in detail below.

[0037] Step S201: Obtain supply and demand basic data and extract the supply and demand basic data, wherein the supply and demand basic data includes supply and demand tags, supply and demand regions, supply and demand industries, publishers, and basic information of the publishers;

[0038] Step S202: Acquire a characteristic vector of supply and demand from a vector database, and integrate the characteristic vector with the basic supply and demand data to obtain integrated supply and demand data; the vector database extracts the characteristic vector based on an enterprise dictionary system;

[0039] Step S203: constructing a document index of the integrated supply and demand data and storing it in Elastic Search;

[0040] Step S204, in response to the target user's click-through supply and demand operation, pre-processing the supply and demand information determined by the click-through supply and demand operation using a corresponding filtering strategy based on the target object of the click-through supply and demand operation to obtain a feature vector of the pending supply and demand content;

[0041] Step S205 , performing a vector search based on tags, regions, and industries on the pending supply and demand content feature vectors through Elastic Search to obtain a set of candidate supply and demand content feature vectors;

[0042] Step S206 , using a preset distance formula to calculate the feature similarity between each candidate supply and demand content feature vector included in the candidate supply and demand content feature vector set and the undetermined supply and demand content feature vector, and sorting each candidate supply and demand content feature vector according to the feature similarity to obtain a supply and demand content feature similarity calculation result.

[0043] In an embodiment of the present invention, the server obtains basic supply and demand data from the enterprise's business system, data warehouse, or other data sources. This data may come from multiple channels such as the enterprise's sales department, procurement department, market research organization, etc. For example, the server of a large manufacturing enterprise obtains the following basic supply and demand data:

[0044] (1) Supply information:

[0045] 1. Supply and demand labels: "high-end CNC machine tools", "high-precision machining tools".

[0046] 2. Supply and demand area: "Shanghai".

[0047] 3. Supply and demand industry: “Manufacturing”.

[0048] 4. Publisher: "Shanghai XX Machine Tool Manufacturing Co., Ltd."

[0049] 5. Basic information of the publisher: The company was established in 2000, has 500 employees, holds multiple patented technologies, and is well-known in the field of high-end machine tool manufacturing.

[0050] (2) Demand information:

[0051] 1. Supply and demand labels: "high-precision machining center", "high-quality abrasive tools".

[0052] 2. Supply and demand area: "Beijing".

[0053] 3. Supply and demand industry: “Manufacturing”.

[0054] 4. Publisher: “Beijing XX Machinery Manufacturing Co., Ltd.”

[0055] 5. Basic information of the publisher: The company was established in 2010 and has 300 employees. It specializes in the manufacturing of mechanical parts and has a large demand for high-precision processing equipment.

[0056] The server extracts this basic supply and demand data, extracting key information. During this extraction process, the server uses natural language processing technology and data parsing algorithms to accurately identify and separate fields such as supply and demand tags, region, industry, publisher, and basic information. For example, the server can accurately separate the supply information for "high-end CNC machine tools, high-precision machining tools" into two independent supply and demand tags. From the text description of the publisher's basic information, the server can extract key data points such as founding date, number of employees, and technological advantages.

[0057] The server retrieves the characteristic vectors of supply and demand from a vector database built based on the enterprise dictionary system. The enterprise dictionary system is a set of standardized terms and concepts defined within the enterprise to uniformly describe supply and demand content. Based on the enterprise dictionary system, the vector database maps supply and demand content into vectors in a high-dimensional space, with each vector representing a characteristic of the supply and demand content.

[0058] Taking the supply information from earlier as an example, "high-end CNC machine tools" has a clear definition and classification in the enterprise dictionary system. The vector database uses this information to generate a feature vector, which contains information related to high-end CNC machine tools in various dimensions, such as technical parameters, application areas, and market positioning. Similarly, "high-precision machining tools" also have a corresponding feature vector.

[0059] The server integrates the acquired feature vectors with the previously extracted supply and demand basic data. During this integration process, the server associates each dimension in the feature vector with fields in the basic supply and demand data. For example, combining the technical parameter dimension in the feature vector for "high-end CNC machine tools" with the patented technology information in the publisher's basic information yields a more comprehensive description of the product being supplied. Ultimately, the resulting integrated supply and demand data encompasses both the basic supply and demand information and the feature vectors, creating a richer and more expressive dataset.

[0060] The server constructs the integrated supply and demand data into a document format. Each supply and demand information is converted into a document that contains all the integrated information, including supply and demand tags, region, industry, publisher, publisher basic information, and feature vectors.

[0061] Taking the supply information of "Shanghai XX Machine Tool Manufacturing Co., Ltd." as an example, the constructed document may be as follows:

[0062] |Field Name|Field Value|

[0063] |---|---|

[0064] |Supply and demand tags|"high-end CNC machine tools", "high-precision machining tools"|

[0065] |Supply and Demand Area|"Shanghai"|

[0066] |Supply and Demand Industry|"Manufacturing"|

[0067] |Publisher|"Shanghai XX Machine Tool Manufacturing Co., Ltd."|

[0068] |Publisher Information|Founded in 2000, with 500 employees, it holds numerous patents and is well-known in the high-end machine tool manufacturing sector.

[0069] |Feature vector|[technical parameter dimension value, application field dimension value, market positioning dimension value, ...]|

[0070] The server indexes these documents into ElasticSearch for storage. ElasticSearch is a distributed search engine and data analysis engine that can quickly store, retrieve, and analyze large quantities of documents. During the indexing process, the server sets appropriate indexing strategies based on the characteristics of the supply and demand data, such as partitioning the index by supply and demand region, industry, and other fields to improve the efficiency of subsequent retrieval.

[0071] Assume that the target user is a purchasing staff of "Beijing XX Machinery Manufacturing Co., Ltd.". The staff clicks on a supply information on the company's supply and demand information platform. This supply information is "high-end CNC machine tools" released by "Shanghai XX Machine Tool Manufacturing Co., Ltd."

[0072] The server responds to the target user's click supply and demand operation, and first determines the target object of the click supply and demand operation. Since the click is on supply information released by other companies, the server adopts the corresponding filtering strategy to pre-process the supply and demand information determined by the click supply and demand operation.

[0073] Filtering strategies may include removing irrelevant information, such as advertising descriptions and repetitive content. Furthermore, the supply and demand information is further extracted and filtered to highlight key features. For example, the server will focus on extracting key technical parameters and applicable processing techniques for "high-end CNC machine tools," while removing information about company culture and development history that is less relevant to the current supply and demand match. After preprocessing, a feature vector of the undetermined supply and demand content is obtained, which contains the filtered and extracted key feature information.

[0074] The server uses Elastic Search to perform vector retrieval based on tags, regions, and industries for the above-obtained feature vectors of pending supply and demand content.

[0075] Taking tags as an example, the server searches Elastic Search for all documents containing tags related to "high-end CNC machine tools." These documents may come from different suppliers, and their supply and demand tags may contain similar keywords such as "CNC machine tools" and "high-precision machine tools." The server also considers regional and industry factors. Regarding region, it may prioritize searching for supply information from areas that have business dealings with or are relatively close to the target user's location (Beijing). Regarding industry, it only searches for supply and demand information related to the manufacturing industry, excluding information from other unrelated industries.

[0076] Through this search operation, the server retrieves from Elastic Search a series of documents that are relevant to the pending supply and demand content feature vector in terms of tags, region, industry, and other aspects. Each document corresponds to a supply and demand content feature vector, and these vectors form the candidate supply and demand content feature vector set. For example, the candidate set might include feature vectors for "high-precision CNC machine tools" supply information published by "Nanjing XX Machine Tool Manufacturing Co., Ltd." and feature vectors for "Tianjin XX Machinery Parts Factory" supply information related to machine tool processing.

[0077] The server uses a preset distance formula to calculate the feature similarity between each candidate supply and demand content feature vector in the candidate supply and demand content feature vector set and the pending supply and demand content feature vector. Common preset distance formulas, such as the cosine distance formula, measure similarity by calculating the cosine of the angle between two vectors. The closer the cosine of the angle is to 1, the more similar the two vectors are.

[0078] For example, the server uses the feature vector of the supply information for "high-precision CNC machine tools" published by Nanjing XX Machine Tool Manufacturing Co., Ltd. and the feature vector of the undetermined supply and demand information to calculate the similarity between the two vectors. The server then performs this calculation for each vector in the candidate set.

[0079] After the calculation is complete, the server ranks each candidate supply and demand content feature vector by feature similarity. Vectors with higher similarity scores are ranked higher in the ranking. Ultimately, the supply and demand content feature similarity calculation results are obtained. For example, the ranking results may show that the supply information of "Nanjing XX Machine Tool Manufacturing Co., Ltd." is most similar to the supply information of "High-end CNC Machine Tools" clicked by the target user, followed by the relevant tool supply information of "Tianjin XX Machinery Parts Factory."

[0080] These similarity calculation results provide valuable insights for target users. Based on the sorting results, target users can prioritize supply and demand information that is most similar to their needs, allowing them to more quickly find suppliers or products that meet their needs, improving the efficiency and accuracy of supply and demand matching. Furthermore, the server can use these calculation results to further optimize the recommendation algorithm and display method for supply and demand information, providing better support for enterprises' supply and demand matching services.

[0081] Through the detailed scenario examples above, we can clearly understand the specific operations and functions of each step in the actual application of the method for calculating the similarity of supply and demand content features based on the enterprise dictionary system. This method can help enterprises more effectively process and analyze supply and demand data, achieve accurate supply and demand matching, and improve the operational efficiency and competitiveness of enterprises.

[0082] In an embodiment of the present invention, extracting the basic supply and demand data further includes:

[0083] Extract transaction scale and transaction time limit information from the supply and demand basic data.

[0084] In an embodiment of the present invention, for example, after obtaining the basic supply and demand data, the server will extract the transaction size and transaction time limit information therein. The following is an example of a specific supply and demand scenario.

[0085] On the supply side, the server obtains supply information from an electronic component manufacturer located in Guangzhou. This company produces various basic electronic components such as resistors and capacitors and has a considerable scale in the industry. When the server extracts its basic supply and demand data:

[0086] Extracting transaction size information: The supply information clearly states, "This quarter, we can supply 5 million resistors and 3 million capacitors." Using text recognition and parsing technology, the server accurately extracts the transaction size information from this description, indicating that the supply quantity is 5 million resistors and 3 million capacitors. Furthermore, the information also states, "A single order must be no less than 100,000 units." This minimum purchase quantity requirement is also extracted by the server as part of the transaction size information.

[0087] Extracting transaction time limit information: The supply information states "Product supply cycle is within this quarter, and delivery must be completed within 15 working days after receiving the order." The server extracts two key transaction time limit information from this information: the supply time range is within this quarter, and the order delivery time must be within 15 working days after receiving the order.

[0088] On the demand side, the server obtains demand information from an electronics assembly company in Shenzhen. The company plans to produce a batch of smart bracelets soon and requires a large number of electronic components. The server extracts the basic supply and demand data:

[0089] Extracting transaction size information: The request information states, "This request requires 8 million resistors and 5 million capacitors, planned for two batches." The server extracts the requested quantity of 8 million resistors and 5 million capacitors, along with information about the transaction size, including the two-batch purchase. Furthermore, the requester also requested, "The purchase quantity per batch shall not exceed 5 million units," which the server also extracts and records.

[0090] Extraction of transaction time limit information: The demand information mentioned that "the first batch of purchases must be delivered within half a month, and the second batch must be delivered within one month." The server accurately extracted these two transaction time limit information and clarified the demander's delivery time requirements for different batches of purchases.

[0091] After extracting this transaction size and transaction time limit information, the server will integrate it with the previously extracted supply and demand tags, region, industry, publisher, and publisher basic information. This information is crucial for subsequent supply and demand data processing. When integrating feature vectors, transaction size and time limit information can be incorporated as important feature dimensions, making the characteristics of supply and demand data richer and more comprehensive. This information can also serve as important screening and matching conditions when searching and calculating feature similarity through Elastic Search. For example, if the supplier's delivery time cannot meet the demander's time limit requirements, then even if other features have high similarity, it may be lowered in priority in the final sorting, helping companies to more accurately achieve supply and demand matching.

[0092] In an embodiment of the present invention, the characteristic vectors of supply and demand are obtained from the vector database, and the characteristic vectors are integrated with the basic data of supply and demand. The following implementation methods are also provided.

[0093] The basic characteristics of the target user are integrated into the supply and demand basic data, wherein the basic characteristics of the user include the industry to which the user belongs, years of employment, and company size.

[0094] In the embodiment of the present invention, for example, when the server obtains the supply and demand feature vector from the vector database and integrates it with the supply and demand basic data, it also incorporates the basic characteristics of the target user. Detailed description is given below using specific scenarios.

[0095] Let's assume the target user is a medium-sized auto parts manufacturer located in Hangzhou, named "Hangzhou XX Auto Parts Co., Ltd." Its basic characteristics are: it belongs to the automotive manufacturing industry, has 12 years of experience in the industry, has 400 employees, and has an annual turnover of 80 million yuan.

[0096] On the supply side, the server retrieves a supply information from "Shanghai XX Metal Materials Co., Ltd." Its basic supply and demand data includes: the supply and demand tag is "high-quality steel, aluminum alloy plate," the supply and demand region is Shanghai, the supply and demand industry is raw material supply, and the publisher is "Shanghai XX Metal Materials Co., Ltd." Basic information about the publisher is that it was established in 2005, has 200 employees, and specializes in the research and development and sales of metal materials.

[0097] The server retrieves the feature vectors corresponding to "high-quality steel and aluminum alloy plates" from the vector database. These feature vectors are generated based on the enterprise dictionary system and cover multiple dimensions of information, including the material's composition, performance, and specifications.

[0098] Next, the server integrates the basic characteristics of the target user "Hangzhou XX Auto Parts Co., Ltd." into the supply and demand basic data of this supply information. Specifically:

[0099] Industry Integration: Because the target user belongs to the automotive industry, the server pays special attention to the compatibility of materials in the supply information with automotive manufacturing during integration. For example, within the feature vectors for steel and aluminum alloy sheet properties, performance indicators that meet the requirements of automotive parts manufacturing, such as high strength and corrosion resistance, are highlighted. Furthermore, the textual description of the basic supply and demand data also includes information on common automotive applications for these materials, such as in the manufacture of vehicle body frames and engine components.

[0100] Incorporating years of experience: Considering the target user has 12 years of experience, indicating a certain level of experience and standards in automotive parts manufacturing, the server will prioritize information related to quality assurance, technical support, and other aspects of the supply information during integration. For example, if the supplier provides relevant quality certifications and technological R&D results, the server will analyze and match these with the target user's industry experience, highlighting the value of this information to experienced companies.

[0101] Enterprise size integration: The target user's enterprise size is medium-sized, with an annual turnover of 80 million yuan. Based on this size information, the server adjusts and correlates the transaction size, pricing strategy, and other aspects of the supply information. For example, if a supplier offers preferential pricing policies for enterprises of different sizes, the server will highlight pricing information applicable to medium-sized enterprises. Furthermore, based on the target user's projected purchase volume, the transaction size information will be further refined and annotated, such as the relationship between the projected purchase volume and the supplier's volume discount.

[0102] By integrating the target user's basic characteristics into the supply and demand base data and then integrating them with the feature vector, the server ensures that the final supply and demand data is more closely aligned with the target user's actual needs and characteristics. This provides a more accurate and targeted data foundation for subsequent document indexing, retrieval, and similarity calculation, helping to improve the quality and efficiency of supply and demand matching, allowing target users to more easily find supply information that meets their requirements.

[0103] In the embodiment of the present invention, the target object based on the click supply and demand operation adopts a corresponding filtering strategy, which can be implemented through the following examples.

[0104] If the target object of the click supply and demand operation is the supply and demand information published by the target user himself, a first filtering strategy is adopted to remove duplicate or low-value self-related information;

[0105] If the target object of the click supply and demand operation is the supply and demand information published by other users, a second filtering strategy is adopted to filter out information that has a high degree of matching with the target user's needs.

[0106] In the embodiment of the present invention, illustratively, when responding to a click supply and demand operation of a target user, the server will adopt corresponding filtering strategies according to the different click objects, which will be described in detail below through specific scenarios.

[0107] Let's assume the target user is "Chengdu XX Electronic Technology Co., Ltd.". This company has posted both demand information for a certain chip and supply information for independently developed electronic components in the system. One day, while browsing the system, a staff member of this company clicked on the demand information for a certain chip posted by their own company.

[0108] After detecting the click-supply-and-demand operation, the server determines that the target object of the click-supply-and-demand operation is the supply-and-demand information released by the target user himself, and then adopts the first filtering strategy.

[0109] To remove duplicate information, the server compares these historical demand information, as the company may have published demand information for the same chip multiple times over different time periods, and some of the content may be repeated. For example, if a previously published demand information repeatedly mentions "need XX model chip, monthly demand of 5,000 pieces," the server will identify the duplicate statement and retain only the most recent and complete demand description, removing the remaining duplicate content.

[0110] For example, when a company posts demand information, it includes promotional content about internal cultural activities. This information has no direct relevance to the chip demand itself and is considered low-value information. The server removes this information from the demand information and retains only the key content closely related to the chip demand, such as the chip's technical parameter requirements, quality standards, and delivery time, resulting in filtered and refined information.

[0111] The target user is Chengdu XX Electronic Technology Co., Ltd., which needs to purchase new electronic components. While browsing the system, the staff clicked on the electronic component supply information posted by Shenzhen XX Component Manufacturing Co., Ltd.

[0112] The server determines that the target object of the clicked supply and demand operation is the supply and demand information published by other users, and thus adopts the second filtering strategy.

[0113] First, the server determines the target user, "Chengdu XX Electronic Technology Co., Ltd.", based on their historical purchase history, previously published demand information, and basic user characteristics to determine their demand preferences. For example, the electronic components purchased by this company in the past were mostly high-precision, high-stability products, primarily used in communications equipment manufacturing.

[0114] The server then matches and filters the supply information published by "Shenzhen XX Component Manufacturing Co., Ltd." For product parameters mentioned in the supply information, if the accuracy, stability, and other indicators of a particular electronic component meet the target user's requirements, the server will retain the relevant content and mark it as a high-match. However, the server will filter out information related to functions or parameters that the target user clearly does not need, such as the supplier's information on low-precision components suitable for consumer electronics.

[0115] The server also considers factors such as supply and demand regions, transaction size, and time constraints. If the supplier's delivery location is acceptable to the target user, and the transaction size and time constraints also meet the target user's needs, the relevant information will be retained and highlighted. Otherwise, information that does not meet the requirements will be filtered out, ultimately selecting information that highly matches the target user's needs, providing more accurate data for subsequent searches and similarity calculations.

[0116] In this embodiment of the present invention, for example, assume that the target user is "Chengdu XX Electronic Technology Co., Ltd.". This company has posted both demand information for a certain chip and supply information for independently developed electronic components in the system. One day, while browsing the system, a staff member of this company clicks on the demand information for a certain chip posted by their own company.

[0117] After detecting the click-supply-and-demand operation, the server determines that the target object of the click-supply-and-demand operation is the supply-and-demand information released by the target user himself, and then adopts the first filtering strategy.

[0118] To remove duplicate information, the server compares these historical demand information, as the company may have published demand information for the same chip multiple times over different time periods, and some of the content may be repeated. For example, if a previously published demand information repeatedly mentions "need XX model chip, monthly demand of 5,000 pieces," the server will identify the duplicate statement and retain only the most recent and complete demand description, removing the remaining duplicate content.

[0119] For example, when a company posts demand information, it includes promotional content about internal cultural activities. This information has no direct relevance to the chip demand itself and is considered low-value information. The server removes this information from the demand information and retains only the key content closely related to the chip demand, such as the chip's technical parameter requirements, quality standards, and delivery time, resulting in filtered and refined information.

[0120] The target user is Chengdu XX Electronic Technology Co., Ltd., which needs to purchase new electronic components. While browsing the system, the staff clicked on the electronic component supply information posted by Shenzhen XX Component Manufacturing Co., Ltd.

[0121] The server determines that the target object of the clicked supply and demand operation is the supply and demand information published by other users, and thus adopts the second filtering strategy.

[0122] First, the server determines the target user, "Chengdu XX Electronic Technology Co., Ltd.", based on their historical purchase history, previously published demand information, and basic user characteristics to determine their demand preferences. For example, the electronic components purchased by this company in the past were mostly high-precision, high-stability products, primarily used in communications equipment manufacturing.

[0123] The server then matches and filters the supply information published by "Shenzhen XX Component Manufacturing Co., Ltd." For product parameters mentioned in the supply information, if the accuracy, stability, and other indicators of a particular electronic component meet the target user's requirements, the server will retain the relevant content and mark it as a high-match. However, the server will filter out information related to functions or parameters that the target user clearly does not need, such as the supplier's information on low-precision components suitable for consumer electronics.

[0124] The server also considers factors such as supply and demand regions, transaction size, and time constraints. If the supplier's delivery location is acceptable to the target user, and the transaction size and time constraints also meet the target user's needs, the relevant information will be retained and highlighted. Otherwise, information that does not meet the requirements will be filtered out, ultimately selecting information that highly matches the target user's needs, providing more accurate data for subsequent searches and similarity calculations.

[0125] In the embodiment of the present invention, before performing vector retrieval based on tags, regions, and industries on the feature vectors of the pending supply and demand content through Elastic Search, the following implementation manner is also provided.

[0126] The undetermined supply and demand content feature vector is normalized to eliminate the dimensional effects of different feature dimensions.

[0127] In this embodiment of the present invention, for example, assume that the target user is "Wuhan XX Machinery Manufacturing Co., Ltd." and a purchasing representative from the company clicks on a supply listing for "high-precision gears" posted by "Changsha XX Parts Supplier." The server responds to the click and, after filtering the listing, obtains a feature vector for the pending supply and demand content.

[0128] This undetermined supply and demand feature vector contains information from multiple dimensions, such as the gear's accuracy grade (expressed as a numerical value, possibly ranging from 1 to 10), material hardness (expressed in Rockwell hardness HR), applicable machine types (which can be represented by codes, such as 01 for machine tools and 02 for automotive manufacturing), and price (in RMB). These different dimensional features have different dimensions. The accuracy grade is a relatively small numerical value, the material hardness has a different numerical range and unit than the accuracy grade, the price may have a larger numerical range, and the applicable machine type is a discrete coded value.

[0129] Before performing vector retrieval based on tags, regions, and industries through Elastic Search, the server needs to normalize the feature vector of the pending supply and demand content to eliminate the dimensional effects of different feature dimensions.

[0130] For the dimension of precision level, assume that the current precision level of "high-precision gear" is level 8. The server uses the minimum-maximum normalization method. It is known that the minimum value of the precision level of this type of gear is 1 and the maximum value is 10. Through the formula Calculate it, which is approximately equal to 0.78, and normalize it

[0131] to the range of 0-1.

[0132] For the material hardness dimension, assuming that the current gear material hardness is 50HR, it is known that the minimum hardness of this type of gear material is 30HR and the maximum hardness is 60HR. Similarly, the minimum-maximum normalization method is used to calculate x' to be approximately equal to 0.67, completing the normalization of this dimension.

[0133] For the price dimension, assuming that the unit price of the high-precision gear is 200 yuan, and the minimum price of similar gears on the market is known to be 100 yuan and the maximum price is 500 yuan, x' is calculated to be 0.25 according to the above normalization formula.

[0134] For the discrete encoding dimension of applicable machine type, the server can use one-hot encoding to convert it into a suitable numerical form for normalization. For example, if "01" represents a machine tool, when the gear is suitable for machine tools, the corresponding encoding vector is [1, 0, ..., 0] (assuming there are multiple machine type codes). Appropriate scaling and other operations are then performed to ensure that it meets normalization requirements.

[0135] After normalizing each dimension of the pending supply and demand content feature vector, the dimensionality effects of different feature dimensions are eliminated, ensuring that each dimension has a relatively balanced weight in subsequent vector retrieval and similarity calculations. The server then uses this normalized pending supply and demand content feature vector in Elastic Search retrieval operations, enabling more accurate identification of matching candidate supply and demand content feature vectors from a wide range of supply and demand information. This improves retrieval accuracy and reliability, laying the foundation for subsequent precise supply and demand matching.

[0136] In the embodiment of the present invention, after calculating the feature similarity using the preset distance formula, the following implementation is also provided.

[0137] The feature similarity calculation result is weighted according to the target user's historical click behavior data, wherein the historical click behavior data includes information such as click frequency, dwell time after clicking, whether a transaction is generated, etc.

[0138] In this embodiment of the present invention, for example, assume that the target user is "Chongqing XX Auto Parts Factory." The server detects that the target user clicked on a supply information post about "special steel" by "Guangzhou XX Metal Products Company." Following the process, the server first calculates the feature similarity between this supply information and the feature vectors of other candidate supply and demand content using a preset distance formula. The server then weights the feature similarity calculation results based on the historical click behavior data of the target user "Chongqing XX Auto Parts Factory."

[0139] The server found that in the past three months, "Chongqing XX Auto Parts Factory" had a high frequency of clicks on special steel supply information, with an average of 5 clicks on related information per week. This shows that the company has a sustained and high level of attention to special steel. Therefore, when adjusting the weight, for those supply information that is similar in characteristics to the "special steel" supply information clicked this time and is also related to special steel, the server will appropriately increase its weight in the feature similarity calculation results. For example, there was originally a special steel supply information from "Wuhan XX Steel Enterprise". When the click frequency was not taken into account, its feature similarity calculation score was 0.6. Due to the high click frequency of the target user on special steel related information, the server increased its weight by 10%, and the adjusted score became 0.6×(1+0.1)=0.66.

[0140] Server records show that after clicking on the special steel supply information, the average stay time of "Chongqing XX Auto Parts Factory" is 10 minutes. For some supply information that eventually generates transactions, the stay time is often longer, reaching more than 15 minutes. For the supply information of "Guangzhou XX Metal Products Company" clicked this time, the target user stayed for 12 minutes. When adjusting the weight, the server will increase the weight of those supply information that have similar characteristics to the supply information and have had a longer stay time of the target user in the past. For example, there is a special steel supply information of "Shanghai XX Steel Supplier". The feature similarity was originally 0.5. The average stay time of the target user who clicked on similar information of this supplier in the past was 13 minutes. The server increased its weight by 8% based on the stay time factor, and the adjusted score became 0.5×(1+0.08)=0.54.

[0141] Server statistics show that 20% of the specialty steel supply information previously clicked by "Chongqing XX Auto Parts Factory" ultimately resulted in transactions. The server assigns higher weight to other supply information with similar characteristics to the supply information that resulted in transactions. For example, suppose the supply information for "Chengdu XX Steel Distributor" has characteristics similar to a supply information that the target user previously transacted with. The feature similarity calculation result is originally 0.4. Due to this similarity with the transaction information, the server increases its weight by 15%, resulting in an adjusted score of 0.4 × (1 + 0.15) = 0.46.

[0142] By comprehensively considering historical click behavior data such as click frequency, post-click dwell time, and whether a transaction occurred, the server adjusts the weights of the feature similarity calculation results. This adjustment more accurately reflects the target user's true interest in different supply and demand information and their potential for transaction. This provides the target user with a more tailored supply and demand information ranking, improving the accuracy and effectiveness of supply and demand matching.

[0143] In an embodiment of the present invention, the preset distance formula is a cosine distance formula.

[0144] In this embodiment of the present invention, for example, let's assume the target user is "Hangzhou XX Automation Equipment Factory," and a purchasing representative from the factory clicks on a supply listing for "high-precision servo motors" posted by "Shenzhen XX Technology Company" on the platform. After completing a series of preliminary processing steps, the server obtains the undetermined supply and demand feature vector corresponding to this clicked information. It then retrieves multiple candidate supply and demand feature vectors from the database and prepares to calculate their feature similarity using the cosine distance formula.

[0145] The undetermined supply and demand characteristics of "high-precision servo motors" clicked by "Hangzhou XX Automation Equipment Factory" are quantified across multiple dimensions. For example, the accuracy dimension reflects the motor's high positioning accuracy; the power dimension indicates that it has the appropriate power to operate the automation equipment; and the interface compatibility dimension indicates that it has specific interface types suitable for automated production lines. The server combines these characteristics into a feature vector.

[0146] Let's look at another candidate supply and demand item: "High-precision servo motors" supplied by "Nanjing XX Motor Manufacturing Enterprise." This information also forms a feature vector based on accuracy, power, and adapter interface. Its accuracy is slightly lower than the target click information, while its power levels are similar and its adapter interface types differ.

[0147] The server uses the cosine distance formula to calculate the feature similarity between the two. Simply put, the cosine distance formula compares the degree of similarity between two vectors in direction. If you imagine these two feature vectors as two directed line segments in space, their starting points are both at the origin. When the "high-precision servo motor" features represented by the two vectors are very similar in all dimensions, the directions of these two directed line segments in space will be very close, and the similarity value calculated by the cosine distance formula will be very high, close to 1. This means that the "high-precision servo motor" supply information of "Nanjing XX Motor Manufacturing Enterprise" and the information clicked by "Hangzhou XX Automation Equipment Factory" are very similar in features. For example, when the accuracy is very high, the power is adapted, and the interface can meet the requirements, the similarity will approach 1.

[0148] On the contrary, if the "high-precision servo motor" of "Nanjing XX Motor Manufacturing Enterprise" is significantly different from the information clicked by "Hangzhou XX Automation Equipment Factory" in some key feature dimensions, such as the accuracy is far from enough, the power is not matched, and the adapter interface is completely different, then the directions of these two directed line segments in space will be significantly different, and the calculated similarity value will be low, possibly approaching 0.

[0149] In this way, the server calculates the similarity between the feature vectors of all candidate supply and demand information and the feature vectors of the pending supply and demand information. Finally, the server ranks the candidate supply and demand information based on these similarity results, prioritizing the most similar information to Hangzhou XX Automation Equipment Factory, helping the factory more quickly identify high-precision servo motor suppliers that meet its needs.

[0150] In order to more clearly describe the solution provided by the embodiment of the present invention, a relatively complete implementation method is provided below. Figure 2 , Figure 2 A schematic diagram of a framework for calculating the similarity of supply and demand content features based on an enterprise dictionary system provided by an embodiment of the present invention.

[0151] (1) Extracting data from the supply and demand data published by users, mainly extracting supply and demand tags, supply and demand regions, supply and demand industries, publishers, and basic information of publishers;

[0152] (2) Obtain the characteristic vectors of supply and demand through the database and fuse them into the basic data of supply and demand;

[0153] (3) Build the doc index to Elastic Search 8.0+;

[0154] (4) Obtaining the basic information and feature vector information of the item through the user click strategy;

[0155] (5) Pass Calculate vector similarity where the vector dimension is 100; Item A = X A Where A is a set of a, a∈[0,100);

[0156] Item B=X B Where B is the b set, b∈[0,100);

[0157]

[0158] (6) Use ElasticSearch to search vectors based on tags, regions, and industries, and sort based on vector similarity calculations

[0159] Please refer to Figure 3 , Figure 3 An embodiment of the present invention provides a device 110 for calculating similarity between supply and demand content features based on an enterprise dictionary system, comprising:

[0160] Acquisition module 1101 is used to acquire basic supply and demand data and extract the basic supply and demand data, which includes supply and demand tags, supply and demand regions, supply and demand industries, publishers, and basic publisher information; acquire a supply and demand feature vector from a vector database, and integrate the feature vector with the basic supply and demand data to obtain integrated supply and demand data; extract the feature vector from the vector database based on an enterprise dictionary system; construct a document index from the integrated supply and demand data and store it in Elastic Search; in response to a target user's click-on supply and demand operation, pre-process the supply and demand information determined by the click-on supply and demand operation using a corresponding filtering strategy based on the target object of the click-on supply and demand operation to obtain a pending supply and demand content feature vector; perform a vector search based on tags, regions, and industries on the pending supply and demand content feature vector through Elastic Search to obtain a set of candidate supply and demand content feature vectors;

[0161] The calculation module 1102 is used to calculate the feature similarity between each candidate supply and demand content feature vector included in the candidate supply and demand content feature vector set and the undetermined supply and demand content feature vector using a preset distance formula, and sort each of the candidate supply and demand content feature vectors according to the feature similarity to obtain a supply and demand content feature similarity calculation result.

[0162] It should be noted that the implementation principles of the aforementioned device 110 for calculating supply and demand content feature similarity based on an enterprise dictionary system can be referenced to the implementation principles of the aforementioned method for calculating supply and demand content feature similarity based on an enterprise dictionary system, and will not be further elaborated here. It should be understood that the division of the various modules of the aforementioned device is merely a logical functional division. In actual implementation, they may be fully or partially integrated into a single physical entity or physically separated. Furthermore, these modules may be implemented entirely as software invoked by a processing element, entirely as hardware, or partially as software invoked by a processing element, while others may be implemented in hardware. For example, the device 110 for calculating supply and demand content feature similarity based on an enterprise dictionary system may be a separate processing element, or integrated into a chip within the aforementioned device. Furthermore, the device 110 may be stored in the memory of the aforementioned device in the form of program code, invoked by a processing element within the aforementioned device to execute the functions of the aforementioned device 110 for calculating supply and demand content feature similarity based on an enterprise dictionary system. The implementation of the other modules is similar. Furthermore, these modules may be fully or partially integrated or implemented independently. The processing element described herein may be an integrated circuit with signal processing capabilities. During implementation, each step of the above method or each module above may be completed by an integrated logic circuit of hardware in a processor element or by instructions in the form of software.

[0163] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASICs), one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented by scheduling program code on a processing element, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0164] The embodiment of the present invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned device 110 for calculating the similarity of supply and demand content features based on the enterprise dictionary system. Figure 4 As shown, Figure 4 This is a block diagram of a computer device 100 according to an embodiment of the present invention. The computer device 100 includes an apparatus 110 for calculating similarity between supply and demand content features based on an enterprise dictionary system, a memory 111 , a processor 112 , and a communication unit 113 .

[0165] In order to realize the transmission or interaction of data, the memory 111, the processor 112 and the communication unit 113 are electrically connected to each other directly or indirectly. For example, the electrical connection between these components can be realized through one or more communication buses or signal lines. The device 110 for calculating the similarity of supply and demand content features based on the enterprise dictionary system includes at least one software function module that can be stored in the memory 111 in the form of software or firmware or solidified in the operating system (OS) of the computer device 100. The processor 112 is used to execute the device 110 for calculating the similarity of supply and demand content features based on the enterprise dictionary system stored in the memory 111, such as the software function modules and computer programs included in the device 110 for calculating the similarity of supply and demand content features based on the enterprise dictionary system.

[0166] An embodiment of the present invention provides a readable storage medium, which includes a computer program. When the computer program is executed, the computer device where the readable storage medium is located is controlled to execute the aforementioned device 110 for calculating the similarity of supply and demand content features based on the enterprise dictionary system.

[0167] For illustrative purposes, the foregoing description has been made with reference to specific embodiments. However, the above illustrative discussion is not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed. Numerous modifications and variations are possible in light of the above teachings. These embodiments have been selected and described in order to best illustrate the principles of the present disclosure and its practical application, thereby enabling those skilled in the art to best utilize the present disclosure and to utilize various embodiments with various modifications as appropriate for the specific application contemplated.

Claims

1. A method for calculating the similarity of supply and demand content features based on an enterprise dictionary system, characterized in that: include: Obtaining and extracting basic supply and demand data, including supply and demand tags, supply and demand regions, supply and demand industries, publishers, and basic information of the publishers; Acquire a characteristic vector of supply and demand from a vector database, and integrate the characteristic vector with the basic supply and demand data to obtain integrated supply and demand data; the vector database extracts the characteristic vector based on an enterprise dictionary system; Constructing a document index for the integrated supply and demand data and storing it in Elastic Search; In response to a target user's click-to-supply / demand operation, pre-processing the supply and demand information determined by the click-to-supply / demand operation using a corresponding filtering strategy based on the target object of the click-to-supply / demand operation to obtain a feature vector of the pending supply and demand content; Performing a vector search based on tags, regions, and industries on the pending supply and demand content feature vectors through Elastic Search to obtain a set of candidate supply and demand content feature vectors; A preset distance formula is used to calculate the feature similarity between each candidate supply and demand content feature vector included in the candidate supply and demand content feature vector set and the undetermined supply and demand content feature vector, and each candidate supply and demand content feature vector is sorted according to the feature similarity to obtain a supply and demand content feature similarity calculation result.

2. The method according to claim 1, characterized in that The extracting of the basic supply and demand data further includes: Extract transaction scale and transaction time limit information from the supply and demand basic data.

3. The method according to claim 1, characterized in that The step of obtaining the characteristic vector of supply and demand from the vector database and integrating the characteristic vector with the basic supply and demand data further includes: The basic characteristics of the target user are integrated into the supply and demand basic data, wherein the basic characteristics of the user include the industry to which the user belongs, years of employment, and company size.

4. The method according to claim 1, wherein The target object based on the click supply and demand operation adopts a corresponding filtering strategy, including: If the target object of the click supply and demand operation is the supply and demand information published by the target user himself, a first filtering strategy is adopted to remove duplicate or low-value self-related information; If the target object of the click supply and demand operation is the supply and demand information published by other users, a second filtering strategy is adopted to filter out information that has a high degree of matching with the target user's needs.

5. The method according to claim 1, wherein Before performing a vector search based on tags, regions, and industries on the feature vectors of the pending supply and demand content through Elastic Search, the method further includes: The undetermined supply and demand content feature vector is normalized to eliminate the dimensional effects of different feature dimensions.

6. The method according to claim 1, characterized in that After calculating the feature similarity using the preset distance formula, the method further includes: The feature similarity calculation result is weighted according to the target user's historical click behavior data, wherein the historical click behavior data includes information such as click frequency, dwell time after clicking, whether a transaction is generated, etc.

7. The method according to claim 1, characterized in that The preset distance formula is the cosine distance formula.

8. A device for calculating the similarity of supply and demand content features based on an enterprise dictionary system, characterized in that: include: An acquisition module is used to acquire and extract basic supply and demand data, wherein the basic supply and demand data includes supply and demand tags, supply and demand regions, supply and demand industries, publishers, and basic information of the publishers; Acquire a characteristic vector of supply and demand from a vector database, and integrate the characteristic vector with the basic supply and demand data to obtain integrated supply and demand data; extract the characteristic vector from the vector database based on the enterprise dictionary system; construct a document index for the integrated supply and demand data and store it in Elastic Search; In response to a target user's click-through supply and demand operation, pre-process the supply and demand information determined by the click-through supply and demand operation using a corresponding filtering strategy based on the target object of the click-through supply and demand operation to obtain a pending supply and demand content feature vector; perform a vector search based on tags, regions, and industries on the pending supply and demand content feature vector using Elastic Search to obtain a set of candidate supply and demand content feature vectors; a calculation module, configured to calculate, using a preset distance formula, a feature similarity between each candidate supply and demand content feature vector included in the candidate supply and demand content feature vector set and the undetermined supply and demand content feature vector, and to sort each of the candidate supply and demand content feature vectors according to the feature similarity to obtain a supply and demand content feature similarity calculation result.

9. A computer device, characterized in that: The computer device includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device executes the method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium includes a computer program, and when the computer program is executed, the computer device where the readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

Images