An intelligent matching system for enterprise technology needs and intellectual property information

Abstract

The present invention relates to the technical field of matching enterprise technology needs with intellectual property information, and specifically to an intelligent matching system for enterprise technology needs and intellectual property information. The system includes data storage, natural language processing, machine learning and user interaction modules. The data storage module uses a distributed database to ensure high data availability and scalability; the natural language processing module analyzes and understands the text; the machine learning module implements classification, clustering and matching; and the user interaction module provides a friendly interface. The system improves matching accuracy and efficiency through natural language processing and machine learning technologies, and the distributed database ensures reliable data storage and processing. The system can help enterprises quickly find intellectual property information that meets their needs and promote technological innovation and development.

Description

Technical Field

[0001] The present invention relates to the technical field of matching enterprise technology requirements with intellectual property information, and in particular to an intelligent matching system for enterprise technology requirements with intellectual property information. Background Art

[0002] In today's rapidly developing technological era, the technological innovation capability of enterprises has become an important part of their core competitiveness. However, how to effectively obtain and utilize relevant intellectual property information is a key issue. The existing intellectual property information matching methods have the disadvantages of low efficiency and low accuracy, which is difficult to meet the actual needs of enterprises.

[0003] With the continuous development of information technology, by building an intelligent matching system, it is possible to efficiently process and analyze a large amount of enterprise technology demand information and intellectual property information. The application of machine learning algorithms can achieve accurate classification, clustering and matching of technology requirements and intellectual property rights, and improve the accuracy and efficiency of matching. In addition, it is also crucial to have a friendly user interface. Therefore, the development of an intelligent matching system for enterprise technology requirements and intellectual property information can help enterprises quickly find intellectual property information that meets their needs, and provide strong support for the technological innovation and development of enterprises. Summary of the invention

[0004] In view of the technical problems existing in the above-mentioned intellectual property information matching technology, the present invention proposes a method which is simple, easy to operate and can effectively solve the shortcomings of low efficiency and low accuracy.

[0005] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is that an intelligent matching system of enterprise technology requirements and intellectual property information includes a data storage module, a natural language processing module, a machine learning module and a user interaction module;

[0006] The data storage module is used to store enterprise technology demand information, intellectual property information and related metadata, and adopts distributed database technology to ensure high availability and scalability of data;

[0007] The natural language processing module includes a word vector generation unit, a named entity recognition unit and a semantic analysis unit, which are used to perform text analysis and semantic understanding on enterprise technical requirements and intellectual property documents;

[0008] The machine learning module includes a clustering algorithm unit, a classification algorithm unit and a recommendation algorithm unit, which are used to classify, cluster and match technical requirements and intellectual property rights;

[0009] The user interaction module provides a friendly interface for enterprise users to interact. Enterprise users can submit technical requirements through web pages or mobile applications and query matching intellectual property information for display.

[0010] Preferably, in the data storage module, the specific operation of using distributed storage is:

[0011] Step 1: Split the database horizontally into multiple shards. Each shard is stored on a different server. The data distribution location is determined by selecting the shard key. The range sharding strategy is used to achieve balanced data distribution and horizontal expansion of the system. Assuming there are M shards, define the range boundary of each shard and distribute the data to the corresponding shard according to the data value. For a given value V, the sharding formula is Where R1...R M To define the boundary value of the shard range;

[0012] Step 2: Store copies of data on multiple nodes to improve the system's fault tolerance and availability; use multi-master replication to synchronize data between multiple nodes to ensure that services can still be provided when a node fails;

[0013] Step 3: Use a consistency protocol to achieve data consistency in a distributed system to ensure that all copies of the data remain in the same state after the operation. Consistency is achieved through communication between the proposer and the recipient. When more than half agree, the proposal is considered a consensus to keep the data consistent;

[0014] Step 4: Use polling scheduling strategy to evenly distribute requests to different servers to prevent single point overload and improve system processing capacity and response speed. Polling scheduling is implemented by counter Next_Server=(Current_Index+1)modN, where N is the total number of servers.

[0015] Step 5: Use the heartbeat mechanism to regularly detect node status, detect and handle faults in a timely manner, and automatically redirect requests to other healthy nodes through the failover mechanism to ensure high availability and continuous operation of the system.

[0016] Preferably, in the natural language processing module: the word vector generation unit generates word vectors using a pre-trained word embedding model or a context-based self-training model to represent the semantic relationship between words in the text;

[0017] The named entity recognition unit is used to identify and extract specific named entities in text data, including but not limited to technical terms, patent names, inventors, applicants and other related entity information; the unit analyzes the text word by word through a predefined entity category label set combined with a neural network model, and annotates related entities;

[0018] The semantic analysis unit is used to perform deep semantic analysis on the generated word vectors and identified named entities, and understand the semantic relationship between the enterprise's technical requirements and the intellectual property documents through context analysis, semantic role labeling or dependency parsing, thereby achieving semantic matching and correlation calculation between texts.

[0019] Preferably, the machine learning module includes a clustering algorithm unit, a classification algorithm unit and a recommendation algorithm unit, and the implementation order and specific operations of the three units are as follows:

[0020] First, the cluster analysis unit: Step 1, first extract feature vectors from technical requirements and intellectual property documents;

[0021] Step 2: Use K-means algorithm to group these feature vectors Where K is the number of clusters, x is the data point, μ i is the centroid of the ith cluster, C i is the set of the i-th cluster;

[0022] Step 3: Finally, output a set of cluster labels to identify which cluster each data point belongs to;

[0023] Then comes the classification algorithm unit: Step 1: First, use the labeled data set to train the classification model. Each data point contains a feature vector and a corresponding category label;

[0024] Step 2: Then train the support vector machine classification model Where F is the function space, x i is the input feature vector, y i is the true value label, Γ is the loss function, and N is the number of training samples;

[0025] Step 3: Finally, classify the clustered data and output the category to which each data point belongs;

[0026] Finally, the recommendation algorithm unit: Step 1: First, perform similarity calculation and use the cosine similarity formula to calculate the similarity between technical requirements and intellectual property information. Where A and B are two feature vectors, and the cosine similarity ranges from -1 to 1. For text information, the similarity value is between 0 and 1;

[0027] Step 2: Then sort the potential matching results according to the similarity scores and recommend the technical requirements and intellectual property pairs with the highest scores;

[0028] Step 3: Finally, generate a recommendation list and output the matching results of technical requirements and intellectual property information.

[0029] Preferably, the technical requirements submitted in the friendly interactive interface of the user interaction module include requirement description, technical field, priority, budget range, and implementation time frame; the display of intellectual property information supports users to browse, filter, sort and view in detail the matching intellectual property information.

[0030] Compared with the prior art, the advantages and positive effects of the present invention are that the use of distributed database technology ensures high availability and scalability of data, and can cope with the storage and processing needs of massive data; the natural language processing module can perform in-depth text analysis and semantic understanding of enterprise technical requirements and intellectual property documents, and accurately extract key information; the clustering, classification and recommendation algorithms in the machine learning module can achieve accurate classification, clustering and matching of technical requirements and intellectual property rights, and provide the most relevant intellectual property information for enterprises; the user interaction module provides a friendly interface, which is convenient for enterprise users to submit technical requirements and browse, filter, sort and view the matching intellectual property information in detail, thereby improving the user experience. DETAILED DESCRIPTION

[0031] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0032] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.

[0033] Embodiment: In the current technological innovation environment, enterprises are faced with the challenge of how to efficiently match their own technological needs with relevant intellectual property information. Traditional matching methods are usually inefficient and inaccurate, and cannot fully support the innovative development needs of enterprises. Therefore, this system aims to improve the accuracy and efficiency of matching through technical means such as natural language processing and machine learning, and provide enterprises with more intelligent solutions. This embodiment describes in detail a system for intelligently matching enterprise technological needs with intellectual property information, including the following main modules: data storage module, natural language processing module, machine learning module, and user interaction module

[0034] First, in the data storage module, the system stores the company's technical demand information and intellectual property data in a distributed database. The database adopts a sharding strategy, dividing the data into multiple shards by range and storing them on different servers to ensure balanced data distribution and system scalability. For example, suppose the database is split into 4 shards, and each shard stores a specific range of data. For a technical requirement keyword "AI-driven", the system will distribute the keyword data to the corresponding shard by calculating the sharding formula. To improve the reliability of the system, the data of each shard will be stored in copies on multiple nodes. If a node fails, other nodes can quickly take over to ensure continuity of service. In order to maintain data consistency, the system uses a consistency protocol to reach data consensus among all nodes. Consistency is achieved through communication between proposers and recipients. When more than half of the votes agree, the proposal is considered a consensus to keep the data consistent. The system also uses a polling scheduling strategy, which uses a counter to implement Next_Server = (Current_Index + 1) mod N, where N is the total number of servers; requests are evenly distributed to different servers. When a server has a problem, the system detects the failure through the heartbeat mechanism and automatically switches the request to a healthy node to ensure that users are not affected.

[0035] Next, the system uses the natural language processing module to perform semantic analysis on the company's technical requirements text and intellectual property documents. The system uses a pre-trained model to convert the input text into a word vector to represent the semantic relationship between words in the text. For example, after the keyword "AI-driven" is processed, a high-dimensional vector will be generated to represent the meaning of the term in different contexts. The system uses named entity recognition technology to extract key entity information from the text, such as patent names, inventors, etc. For example, when the input contains a technical requirement for "image recognition", the system can identify and extract related entities such as "computer vision" and "deep learning". The generated word vectors and extracted named entities will be further semantically analyzed to understand the deep relationship between corporate needs and intellectual property documents. For example, through dependency parsing, the system can identify the semantic similarity between "AI-driven image recognition algorithm" and "image processing patent based on deep learning".

[0036] After semantic analysis, the system enters the machine learning module and refines the demand and patent information through clustering and classification algorithms. The system uses the K-means algorithm to group these feature vectors Where K is the number of clusters, x is the data point, μ i is the centroid of the ith cluster, C iis the set of the ith cluster; output a set of cluster labels to identify which cluster each data point belongs to. In the classification algorithm unit, the labeled data set is first used to train the classification model. Each data point contains a feature vector and a corresponding category label; then the support vector machine classification model is trained. Where F is the function space, x i is the input feature vector, y i is the true value label, Γ is the loss function, and N is the number of training samples; finally, the clustered data is classified and the category to which each data point belongs is output. Finally, the recommendation algorithm unit first performs similarity calculation and uses the cosine similarity formula to calculate the similarity between technical requirements and intellectual property information. Among them, A and B are two feature vectors, and the cosine similarity ranges from -1 to 1. For text information, the similarity value is between 0 and 1. Then the potential matching results are sorted according to the similarity score, and the technical requirements and intellectual property pairs with the highest scores are recommended. Finally, a recommendation list is generated and the matching results of technical requirements and intellectual property information are output. After the operation of the machine learning module, the system will recommend the most matching patent information. For example, when a company submits a demand for "AI-driven image recognition", the system may recommend the most relevant "deep learning image processing" patent.

[0037] Finally, the user interaction module provides enterprises with a simple and easy-to-use interface. Enterprise users can submit detailed technical requirements through this interface, including description, technical field, priority, etc. At the same time, the system also supports users to sort and filter matching results. Enterprise users can enter a demand description, such as "We need an AI-based real-time image recognition system with high priority, a budget of 1 million yuan, and we expect it to be realized within 6 months." The system will generate a recommendation list to display the patent information that best matches the user's needs. Users can filter and sort the results according to conditions such as matching degree and release time to quickly find the most suitable patent technology.

[0038] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent matching system between enterprise technology requirements and intellectual property information, characterized in that: It includes data storage module, natural language processing module, machine learning module and user interaction module; The data storage module is used to store enterprise technology demand information, intellectual property information and related metadata, and adopts distributed database technology to ensure high availability and scalability of data; The natural language processing module includes a word vector generation unit, a named entity recognition unit and a semantic analysis unit, which are used to perform text analysis and semantic understanding on enterprise technical requirements and intellectual property documents; The machine learning module includes a clustering algorithm unit, a classification algorithm unit and a recommendation algorithm unit, which are used to classify, cluster and match technical requirements and intellectual property rights; The user interaction module provides a friendly interface for enterprise users to interact. Enterprise users can submit technical requirements through web pages or mobile applications and query matching intellectual property information for display; In the data storage module, the specific operations using distributed storage are: Step 1: Split the database horizontally into multiple shards. Each shard is stored on a different server. The data distribution location is determined by selecting the shard key. The range sharding strategy is used to achieve balanced data distribution and horizontal expansion of the system. Assuming there are M shards, define the range boundary of each shard and distribute the data to the corresponding shard according to the data value. For a given value V, the sharding formula is Where R1...R M To define the boundary value of the shard range; Step 2: Store copies of data on multiple nodes to improve the system's fault tolerance and availability; use multi-master replication to synchronize data between multiple nodes to ensure that services can still be provided when a node fails; Step 3: Use a consistency protocol to achieve data consistency in a distributed system to ensure that all copies of the data remain in the same state after the operation. Consistency is achieved through communication between the proposer and the recipient. When more than half agree, the proposal is considered a consensus to keep the data consistent; Step 4: Use polling scheduling strategy to evenly distribute requests to different servers to prevent single point overload and improve system processing capacity and response speed. Polling scheduling is implemented by counter Next_Server=(Current_Index+1)modN, where N is the total number of servers. Step 5: Use the heartbeat mechanism to regularly detect node status, detect and handle faults in a timely manner, and automatically redirect requests to other healthy nodes through the failover mechanism to ensure high availability and continuous operation of the system; The machine learning module includes a clustering algorithm unit, a classification algorithm unit and a recommendation algorithm unit. The implementation order and specific operations of the three units are as follows: First, the cluster analysis unit: Step 1, first extract feature vectors from technical requirements and intellectual property documents; Step 2: Use K-means algorithm to group these feature vectors Where K is the number of clusters, x is the data point, μ i is the centroid of the ith cluster, C i is the set of the i-th cluster; Step 3: Finally, output a set of cluster labels to identify which cluster each data point belongs to; Then comes the classification algorithm unit: Step 1: First, use the labeled data set to train the classification model. Each data point contains a feature vector and a corresponding category label; Step 2: Then train the support vector machine classification model Where F is the function space, x i is the input feature vector, y i is the true value label, Γ is the loss function, and N is the number of training samples; Step 3: Finally, classify the clustered data and output the category to which each data point belongs; Finally, the recommendation algorithm unit: Step 1: First, perform similarity calculation and use the cosine similarity formula to calculate the similarity between technical requirements and intellectual property information. Where A and B are two feature vectors, and the cosine similarity ranges from -1 to 1. For text information, the similarity value is between 0 and 1; Step 2: Then sort the potential matching results according to the similarity scores and recommend the technical requirements and intellectual property pairs with the highest scores; Step 3: Finally, generate a recommendation list and output the matching results of technical requirements and intellectual property information.

2. According to claim 1, the intelligent matching system between enterprise technology requirements and intellectual property information is characterized in that: In the natural language processing module: the word vector generation unit generates word vectors using a pre-trained word embedding model or a context-based self-training model to represent the semantic relationship between words in the text; The named entity recognition unit is used to identify and extract specific named entities in text data, including but not limited to technical terms, patent names, inventors, applicants and other related entity information; the unit analyzes the text word by word through a predefined entity category label set combined with a neural network model, and labels the related entities; The semantic analysis unit is used to perform deep semantic analysis on the generated word vectors and identified named entities, and understand the semantic relationship between the enterprise's technical requirements and the intellectual property documents through context analysis, semantic role labeling or dependency parsing, thereby achieving semantic matching and correlation calculation between texts.

3. According to claim 1, the intelligent matching system between enterprise technology requirements and intellectual property information is characterized in that: The technical requirements submitted in the user interaction module's friendly interactive interface include requirement description, technical field, priority, budget range, and implementation time frame; the display of intellectual property information supports users to browse, filter, sort, and view in detail the matching intellectual property information.

Images