Patent classification retrieval method and system based on multivariate data fusion

Abstract

The invention relates to the technical field of patent classification retrieval, and particularly discloses a patent classification retrieval method and system based on multivariate data fusion, which comprises the steps of updating a domain knowledge base in real time, dynamically adjusting the definition and application of terms and ensuring the consistency and accuracy of the terms in different technical fields. In combination with an intelligent analysis technology, terms and contexts thereof in patent documents are deeply analyzed, meanings of the terms are accurately understood, semantic accuracy of the patent documents is evaluated, and understanding deviation caused by term ambiguity or unknown contexts is avoided; by analyzing technical problems and solutions in patent texts and combining technical context and technical background information of a domain knowledge base, the patent text analysis accuracy is further improved, and therefore the patent classification accuracy and the correlation of retrieval results are enhanced; according to the method, the classification and retrieval process of patent literatures is optimized by establishing an accurate semantic recognition mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of patent classification retrieval, and in particular to a patent classification retrieval method and system based on multivariate data fusion. Background Art

[0002] With the rapid development of global technological innovation, the accumulation of patent documents in various technical fields is becoming increasingly large. How to efficiently and accurately classify, retrieve and analyze patents has become an important issue in intellectual property management. At present, the classification and retrieval of patent documents mostly rely on the standardization of terminology and the understanding of technical context. However, in the prior art, there are often inconsistencies in the definition and application of terms, especially in cross-domain or multi-domain technical documents. The differences in the use of terms between different patents are large, resulting in inaccurate patent classification, irrelevant search results, and even affecting the efficiency and quality of patent review. Existing patent classification and retrieval methods fail to effectively integrate domain knowledge bases, lack dynamic updates and intelligent analysis of technical contexts and term meanings, resulting in frequent term ambiguity and unclear context.

[0003] The prior art has the following deficiencies:

[0004] Existing patent classification and retrieval technologies have many deficiencies in terms of accuracy and efficiency, which are mainly manifested in the inability to adjust the definition and application of terms in real time and dynamically, resulting in low classification accuracy and poor relevance of retrieval results. Especially when the technical context of patent documents changes greatly, the existing technology lacks in-depth analysis of technical content and accurate semantic recognition mechanism, and cannot effectively solve the diversity and complexity of term meanings. Therefore, in response to these problems in the prior art, the present invention proposes a dynamic update mechanism based on a domain knowledge base, combined with intelligent analysis methods, to accurately identify the meaning of terms in patent documents, and through in-depth analysis of the technical context, improve the accuracy and efficiency of patent classification and retrieval, thereby effectively solving the problems of inconsistent terminology and inaccurate classification retrieval in the prior art. Summary of the invention

[0005] The purpose of the present invention is to provide a patent classification retrieval method and system based on multivariate data fusion to solve the above-mentioned background problems.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] The patent classification retrieval method based on multivariate data fusion includes the following steps:

[0008] S1: Build a domain knowledge base in the patent-related technical fields, including terminology and technical background information, to provide accurate semantic benchmarks for patent texts and ensure semantic consistency of texts in classification and retrieval;

[0009] The terminology includes: technical vocabulary, professional terms, abbreviations; the technical background information includes: application scenarios of the terminology, development history of related technical fields, technical implementation methods, technical problems and solutions;

[0010] S2: Based on the constructed domain knowledge base, the knowledge base is updated in real time, and the definition and application of terms are dynamically adjusted according to different technical fields. All technical terms in patent documents are comprehensively analyzed with the standard terms in the knowledge base to evaluate the accuracy of patent classification and retrieval;

[0011] S3: Analyze the meaning of terms and evaluate the semantic accuracy of text information by intelligently analyzing the terms in patent documents and the technical content in the context;

[0012] S4: Based on accurate patent classification and retrieval and accurate semantics of text information, and combined with the technical context and domain knowledge of patent texts, establish a semantically accurate recognition mechanism in the classification and retrieval process.

[0013] As a further solution of the present invention: the construction of a domain knowledge base in the patent-related technical field specifically includes:

[0014] For each technical field, the terminology and technical background information of the corresponding technical field are obtained, the terminology and technical background information of the patent document are analyzed, and based on the analysis results, a comprehensive feature index is generated. Based on the comprehensive feature index, the semantic consistency of the text in classification and retrieval is evaluated.

[0015] As a further solution of the present invention: the process of obtaining the comprehensive feature index is:

[0016] Convert the terms and their background information in patent documents into vector representations;

[0017] Convert each term and technical background into a vector representation of fixed dimension;

[0018] For each term T in the patent document i , word vector representation is obtained through word embedding method

[0019] Obtain technical background information B j , using document-based embedding to transform technical background information into background vectors

[0020] For each term T in a patent document i and corresponding technical background information B j , the semantic matching degree between the term and the background is calculated by cosine similarity;

[0021] The calculation expression of the cosine similarity is:

[0022]

[0023] In the formula, Indicates the similarity between the terms in the patent document and the technical background information, represents the dot product of the term vector and the background vector, represents the Euclidean norm of the term vector, represents the Euclidean norm of the background vector, i represents the i-th term in the patent document, and j represents the j-th term in the technical background information;

[0024] For each term in the patent document, calculate the similarity with all technical background information and obtain the similarity vector, and calculate the comprehensive feature index by weighted average;

[0025] Wherein, the calculation expression of the comprehensive characteristic index is:

[0026]

[0027] In the formula, C i represents the comprehensive feature index, n represents the total amount of technical background information, and w j Indicates technical background information B j The weight coefficient of .

[0028] As a further solution of the present invention: the comprehensive analysis of technical terms in all patent documents and standard terms in the knowledge base to evaluate the accuracy of patent classification and retrieval specifically includes:

[0029] Obtain technical terms in patent documents, calculate embedded representations on text and images in patent documents, and generate technical term accuracy coefficients based on the calculation results;

[0030] Obtain standard terms in the knowledge base of patent documents, analyze the connectivity and semantic consistency of term nodes, and calculate the accuracy coefficient of standard terms based on the analysis results;

[0031] The accuracy coefficients of technical terms and standard terms are comprehensively calculated and analyzed, and the accuracy coefficients are calculated based on the analysis results to evaluate the accuracy of patent classification and retrieval.

[0032] As a further solution of the present invention: the process of obtaining the accuracy coefficient of the technical term is:

[0033] Represent the text part of the patent document as an embedding vector and extract the image features associated with the technical terms in the patent document;

[0034] Fusion of text embedding and image embedding to form a multimodal embedding vector for terms

[0035] Based on the multimodal embedding vectors, the text and image consistency scores of technical terms are calculated by cosine similarity;

[0036] According to the frequency of occurrence of technical terms in the literature and the importance weight in the technical field knowledge base, the semantic weight of the technical terms is calculated. The calculation expression is:

[0037] W = log(1+f)*w1;

[0038] In the formula, W represents the semantic weight, w1 represents the importance weight, and f represents the frequency of occurrence of technical terms in patent documents;

[0039] The consistency score and semantic weight are processed comprehensively to calculate the accuracy coefficient of technical terms. The calculation expression is:

[0040]

[0041] Where TTAC represents the technical term accuracy coefficient, λ represents the adjustment parameter, F represents the consistency score, and max(W) represents the maximum semantic weight in the term set.

[0042] As a further solution of the present invention: the process of obtaining the accuracy coefficient of the standard term is as follows:

[0043] Structuring term definitions, domain background, synonyms, and hyponyms in the knowledge base into graphs;

[0044] Graph nodes represent terms, and graph edges represent semantic relationships between terms;

[0045] The frequency of standard terms in the literature in the corresponding technical field is normalized and calculated to obtain the field weight parameter;

[0046] Sum all the domain weight parameters in the patent document to get the connectivity G of the language node;

[0047] Calculate the semantic similarity between the standard term definition and its upper and lower nodes. The calculation expression is:

[0048]

[0049] Where, T a represents the standard term, a and b represent the number of standard terms in the patent document, sin represents the similarity between standard term vectors, N represents the total number of standard terms, S(T a ) indicates the semantic similarity between the standard term definition and its upper and lower nodes;

[0050] The connectivity G of the language nodes is related to the semantic similarity S(T a ) to calculate the ratio and obtain the standard term accuracy coefficient, denoted as TTAB.

[0051] As a further solution of the present invention: the analysis of the meaning of the term and the evaluation of the semantic accuracy of the text information specifically include:

[0052] Wherein, the said terms include technical terms and standard terms;

[0053] Modeling the contextual content of terms in patent documents as a graph;

[0054] The graph convolutional network is used to aggregate the neighborhood features of nodes layer by layer and update the semantic representation of the nodes. The calculation expression is:

[0055]

[0056] In the formula, represents the feature vector of node q in the l+1 layer, q represents the number of nodes, z represents the number of neighbor nodes, represents the set of neighbor nodes of node q, d q Indicates the number of connections between node q and node d z represents the number of connections of the neighbor node z, W (l) represents the weight matrix of the lth layer, l is the number of graph convolutional layers, b (l) represents the bias vector of the lth layer, σ represents the activation function;

[0057] The updated feature vectors of term nodes and context nodes are used to calculate the semantic similarity between the term and the context through cosine similarity;

[0058] The semantic accuracy coefficient of a term is calculated by aggregating the similarity scores of the term node and all its neighbor nodes. The calculation expression is:

[0059]

[0060] Where Y q represents the semantic accuracy coefficient, represents the set of neighbor nodes, x(q) represents the semantic similarity score between the term and the neighbor nodes, and q represents the number of nodes.

[0061] As a further solution of the present invention: the establishment of a semantic accurate recognition mechanism in the classification and retrieval process specifically includes:

[0062] The semantic accuracy coefficient of patent documents and the accuracy coefficient of patent document classification and retrieval are normalized to calculate the comprehensive recognition score. According to the comprehensive recognition score, the semantic accuracy of patent classification and retrieval is divided, and the semantics of document classification and retrieval are divided into qualified semantics and unqualified semantics.

[0063] Patent classification retrieval system based on multivariate data fusion, including:

[0064] A domain knowledge base construction module, which constructs a domain knowledge base in the patent-related technical field, including terminology and technical background information, to provide an accurate semantic benchmark for the patent text and ensure the semantic consistency of the text in classification and retrieval;

[0065] A domain knowledge base update and terminology standardization module, which updates the domain knowledge base in real time, dynamically adjusts the definition and application of terms according to different technical fields, analyzes the matching between terms in patent documents and standard terms in the knowledge base, and evaluates the accuracy of patent classification and retrieval;

[0066] A term semantic analysis and evaluation module, which evaluates the semantic accuracy of terms by intelligently analyzing the terms and their contexts in patent documents;

[0067] The semantic accurate recognition and classification retrieval module is based on accurate patent classification and retrieval, combined with the technical context and domain knowledge of the patent text, to establish an effective semantic accurate recognition mechanism to improve the accuracy of patent text classification and retrieval.

[0068] Beneficial effects of the present invention:

[0069] (1) The present invention achieves consistency between patent document terms and standard terms in the knowledge base by constructing and continuously updating a domain knowledge base and combining it with advanced semantic processing technology, thus successfully solving the accuracy problem caused by inconsistent terms in patent classification and retrieval. The knowledge base not only contains various professional terms, but also covers the application scenarios, technical background information and evolution of related technical fields of the terms, thereby providing in-depth technical context support. By dynamically updating and adjusting the terms in real time, the present invention can adapt to the ever-changing technological progress and ensure that patent documents in each technical field can be classified under the latest and most accurate semantic framework. This method improves the accuracy of patent document classification, avoids the errors and deviations that may exist in the traditional classification system, and ensures that the retrieval results are highly correlated with user needs, making patent retrieval more accurate and efficient. Users can quickly locate the patent documents that best meet their technical needs, effectively improving the utilization efficiency and retrieval effect of patent information.

[0070] (2) The present invention uses intelligent analysis methods to deeply explore the terms and their contextual information in patent documents, thereby achieving accurate interpretation of the meaning of terms and efficient evaluation of semantic accuracy. By comprehensively analyzing the use of terms in specific technical contexts and combining them with the rich technical background information and development context in the domain knowledge base, the present invention can effectively eliminate potential understanding biases caused by term ambiguity or unclear context, ensuring a comprehensive and accurate grasp of the various technical contents in the patent text. In particular, in terms of identifying technical problems and solutions in patent documents, the present invention accurately locates and analyzes the innovative points and technical elements in patent documents through the deep association between terms and technical background, thereby improving the analysis accuracy and semantic accuracy of patent texts. This innovative method provides more reliable support for processes such as patent examination, retrieval and technical evaluation, ensures the accuracy of patent classification and retrieval, and further promotes technological innovation and precise information management in the field of intellectual property. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] The present invention will be further described below in conjunction with the accompanying drawings.

[0072] Figure 1 It is a flowchart of the specific steps of the patent classification retrieval method based on multivariate data fusion of the present invention;

[0073] Figure 2 It is a flowchart of the patent classification retrieval system based on multivariate data fusion in the present invention. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] See also Figure 1 As shown, the present invention is a patent classification retrieval method based on multivariate data fusion, comprising the following steps:

[0076] S1: Build a domain knowledge base in the patent-related technical fields, including terminology and technical background information, to provide accurate semantic benchmarks for patent texts and ensure semantic consistency of texts in classification and retrieval;

[0077] The terminology includes: technical vocabulary, professional terms, abbreviations; the technical background information includes: application scenarios of the terminology, development history of related technical fields, technical implementation methods, technical problems and solutions;

[0078] S2: Based on the constructed domain knowledge base, the knowledge base is updated in real time, and the definition and application of terms are dynamically adjusted according to different technical fields. All technical terms in patent documents are comprehensively analyzed with the standard terms in the knowledge base to evaluate the accuracy of patent classification and retrieval;

[0079] S3: Analyze the meaning of terms and evaluate the semantic accuracy of text information by intelligently analyzing the terms in patent documents and the technical content in the context;

[0080] S4: Based on accurate patent classification and retrieval and accurate semantics of text information, and combined with the technical context and domain knowledge of patent texts, establish a semantically accurate recognition mechanism in the classification and retrieval process to improve retrieval accuracy and efficiency.

[0081] In S1, a domain knowledge base of patent-related technical fields is constructed, including terminology and technical background information, which is used to provide accurate semantic benchmarks for patent texts and ensure the semantic consistency of texts in classification and retrieval, including:

[0082] For each technical field, obtain the terminology and technical background information of the corresponding technical field, analyze the terminology and technical background information of the patent document, generate a comprehensive feature index based on the analysis results, and evaluate the semantic consistency of the text in classification and retrieval based on the comprehensive feature index;

[0083] The process of obtaining the comprehensive feature index is as follows:

[0084] When constructing a domain knowledge base in the field of patent-related technology, the cosine similarity is used to calculate the comprehensive feature index. First, the terms and technical background information in the patent document need to be vectorized, and then the similarity between the text and the standard terms in the domain knowledge base is calculated, and then a comprehensive feature index is generated to evaluate the semantic consistency of the text, including:

[0085] Convert the terms and their background information in patent documents into vector representations;

[0086] Convert each term or technical background into a vector representation of fixed dimension;

[0087] For each term T in the patent document i , word vector representation is obtained through word embedding method

[0088] Obtain technical background information B j , using document-based embedding to transform technical background information into background vectors

[0089] For each term T in a patent document i and corresponding technical background information Bj , the semantic matching degree between the term and the background is calculated by cosine similarity;

[0090] The calculation expression of the cosine similarity is:

[0091]

[0092] In the formula, Indicates the similarity between the terms in the patent document and the technical background information, represents the dot product of the term vector and the background vector, represents the Euclidean norm of the term vector, represents the Euclidean norm of the background vector, i represents the i-th term in the patent document, and j represents the j-th term in the technical background information;

[0093] For each term in the patent document, calculate the similarity with all technical background information and obtain the similarity vector, and calculate the comprehensive feature index by weighted average;

[0094] Wherein, the calculation expression of the comprehensive characteristic index is:

[0095]

[0096] In the formula, C i represents the comprehensive feature index, n represents the total amount of technical background information, and w j Indicates technical background information B j The weight coefficient of

[0097] comparing the comprehensive characteristic index with a preset threshold;

[0098] If the comprehensive feature index is greater than or equal to the preset threshold, it means that the semantics of the corresponding patent in classification and retrieval are consistent;

[0099] If the comprehensive feature index is less than the preset threshold, it means that the semantics of the corresponding patent in classification and retrieval are inconsistent;

[0100] It should be noted that the comprehensive feature index reflects whether the semantics of the patent are consistent during the classification and retrieval process, and the larger the value of the comprehensive feature index, the higher the degree of semantic consistency in the corresponding patent classification and retrieval process.

[0101] In S2, based on the constructed domain knowledge base, the knowledge base is updated in real time, and the definition and application of terms are dynamically adjusted according to different technical fields. All technical terms in patent documents are comprehensively analyzed with the standard terms in the knowledge base to evaluate the accuracy of patent classification and retrieval, including:

[0102] Obtain technical terms in patent documents, calculate embedded representations on text and images in patent documents, and generate technical term accuracy coefficients based on the calculation results;

[0103] Obtain standard terms in the knowledge base of patent documents, analyze the connectivity and semantic consistency of term nodes, and calculate the accuracy coefficient of standard terms based on the analysis results;

[0104] The accuracy coefficients of technical terms and standard terms are comprehensively calculated and analyzed, and the accuracy coefficients are calculated based on the analysis results to evaluate the accuracy of patent classification and retrieval, including:

[0105] The process of obtaining the accuracy coefficient of the technical term is as follows:

[0106] Represent the text part of the patent document as an embedding vector and extract the image features associated with the technical terms in the patent document;

[0107] Fusion of text embedding and image embedding to form a multimodal embedding vector for terms

[0108] Based on the multimodal embedding vectors, the text and image consistency scores of technical terms are calculated by cosine similarity;

[0109] According to the frequency of occurrence of technical terms in the literature and the importance weight in the technical field knowledge base, the semantic weight of the technical terms is calculated. The calculation expression is:

[0110] W = log(1+f)*w1;

[0111] In the formula, W represents the semantic weight, w1 represents the importance weight, and f represents the frequency of occurrence of technical terms in patent documents;

[0112] The consistency score and semantic weight are processed comprehensively to calculate the accuracy coefficient of technical terms. The calculation expression is:

[0113]

[0114] Where TTAC represents the technical term accuracy coefficient, λ represents the adjustment parameter, F represents the consistency score, and max(W) represents the maximum semantic weight in the term set;

[0115] The process of obtaining the accuracy coefficient of the standard term is as follows:

[0116] Structuring term definitions, domain background, synonyms, and hyponyms in the knowledge base into graphs;

[0117] Graph nodes represent terms, and graph edges represent semantic relationships between terms;

[0118] The frequency of standard terms in the literature in the corresponding technical field is normalized and calculated to obtain the field weight parameter;

[0119] Sum all the domain weight parameters in the patent document to get the connectivity G of the language node;

[0120] Calculate the semantic similarity between the standard term definition and its upper and lower nodes. The calculation expression is:

[0121]

[0122] Where, T a represents the standard term, a and b represent the number of standard terms in the patent document, sin represents the similarity between standard term vectors, N represents the total number of standard terms, S(T a ) indicates the semantic similarity between the standard term definition and its upper and lower nodes;

[0123] The connectivity G of the language nodes is related to the semantic similarity S(T a ) to calculate the ratio and obtain the standard term accuracy coefficient, which is recorded as TTAB;

[0124] Among them, the calculation expression of the accuracy coefficient is:

[0125]

[0126] In the formula, Z d represents the accuracy coefficient, o1 and o2 represent the preset proportional coefficients, and both o1 and o2 are greater than 0.

[0127] In S3, the meaning of terms is analyzed through intelligent analysis of the terms in the patent documents and the technical content in the context, and the semantic accuracy of the text information is evaluated, including:

[0128] Wherein, the said terms include technical terms and standard terms;

[0129] Modeling the contextual content of terms in patent documents as a graph;

[0130] The graph convolutional network is used to aggregate the neighborhood features of nodes layer by layer and update the semantic representation of the nodes. The calculation expression is:

[0131]

[0132] In the formula, represents the feature vector of node q in the l+1 layer, q represents the number of nodes, z represents the number of neighbor nodes, represents the set of neighbor nodes of node q, d q Indicates the number of connections between node q and node d zrepresents the number of connections of the neighbor node z, W (l) represents the weight matrix of the lth layer, l is the number of graph convolutional layers, b (l) represents the bias vector of the lth layer, σ represents the activation function;

[0133] The updated feature vectors of term nodes and context nodes are used to calculate the semantic similarity between the term and the context through cosine similarity;

[0134] The semantic accuracy coefficient of a term is calculated by aggregating the similarity scores of the term node and all its neighbor nodes. The calculation expression is:

[0135]

[0136] Where Y q represents the semantic accuracy coefficient, represents the set of neighbor nodes, x(q) represents the semantic similarity score between the term and the neighbor nodes, and q represents the number of nodes;

[0137] comparing the semantic accuracy coefficient with a preset threshold;

[0138] If the semantic accuracy coefficient is greater than or equal to the preset threshold, it means that the semantics of the corresponding patent document term is accurate;

[0139] If the semantic accuracy coefficient is less than a preset threshold, it means that the semantics of the corresponding patent document term is inaccurate.

[0140] In S4, based on accurate patent classification and retrieval and accurate text information semantics, combined with the technical context and domain knowledge of patent texts, a semantic accurate recognition mechanism is established in the classification and retrieval process to improve retrieval accuracy and efficiency, including:

[0141] Normalize the semantic accuracy coefficient of patent documents and the accuracy coefficient of patent document classification and retrieval, calculate the comprehensive recognition score, divide the semantic accuracy of patent classification and retrieval according to the comprehensive recognition score, and divide the semantics of document classification and retrieval into qualified semantics and unqualified semantics;

[0142] The calculation expression of the comprehensive recognition score is:

[0143]

[0144] In the formula, E z represents the comprehensive recognition score;

[0145] Identify the accuracy of advanced semantics for each patent document group during classification search, including:

[0146] comparing the composite recognition score to a preset threshold;

[0147] If the comprehensive recognition score is greater than or equal to the preset threshold, it means that the corresponding patent document group has high accuracy in advanced semantics during classification retrieval and is recorded as qualified semantics;

[0148] If the comprehensive recognition score is less than the preset threshold, it means that the accuracy of the advanced semantics of the corresponding patent document group during classification retrieval is low and is recorded as unqualified semantics.

[0149] See also Figure 2 As shown in the figure, the patent classification retrieval system based on multivariate data fusion includes:

[0150] A domain knowledge base construction module, which constructs a domain knowledge base in the patent-related technical field, including terminology and technical background information, to provide an accurate semantic benchmark for the patent text and ensure the semantic consistency of the text in classification and retrieval;

[0151] A domain knowledge base update and terminology standardization module, which updates the domain knowledge base in real time, dynamically adjusts the definition and application of terms according to different technical fields, analyzes the matching between terms in patent documents and standard terms in the knowledge base, and evaluates the accuracy of patent classification and retrieval;

[0152] A term semantic analysis and evaluation module, which evaluates the semantic accuracy of terms by intelligently analyzing the terms and their contexts in patent documents;

[0153] The semantic accurate recognition and classification retrieval module is based on accurate patent classification and retrieval, combined with the technical context and domain knowledge of the patent text, to establish an effective semantic accurate recognition mechanism to improve the accuracy of patent text classification and retrieval.

[0154] The working principle of the present invention is as follows: by constructing a domain knowledge base in the field of patent-related technical fields, including technical terms, background information, application scenarios, and technology development history, an accurate semantic benchmark is provided for patent texts to ensure the semantic consistency of texts in classification and retrieval; on this basis, the semantic matching degree of terms and technical background information is calculated by cosine similarity to generate a comprehensive feature index, thereby evaluating the semantic consistency of patent documents in classification and retrieval; secondly, the method intelligently analyzes the terms and contexts in patent documents, uses graph neural networks to aggregate the features of term nodes and neighbor nodes, calculates the semantic accuracy coefficient of terms, and further analyzes the semantic accuracy of patent documents. In addition, the accuracy coefficients of technical terms and standard terms in patent documents are calculated by multimodal embedding vectors, and the connectivity and semantic similarity of term nodes are combined with graph analysis to evaluate the accuracy of patent classification and retrieval; finally, based on the accuracy coefficient and semantic consistency, a semantic accurate recognition mechanism is established in the classification and retrieval process, and the patent documents are semantically accurately divided by comprehensive recognition scores to optimize retrieval accuracy and efficiency. This method can dynamically adjust the term definitions, improve the classification and retrieval effect of patent documents, and is suitable for patent information management and retrieval systems in various technical fields.

[0155] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0156] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0157] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0158] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0159] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. Patent classification retrieval method based on multivariate data fusion, characterized in that: The following steps are involved: S1: Build a domain knowledge base in the patent-related technical fields, including terminology and technical background information, to provide accurate semantic benchmarks for patent texts and ensure semantic consistency of texts in classification and retrieval; The terminology includes: technical vocabulary, professional terms, abbreviations; the technical background information includes: application scenarios of the terminology, development history of related technical fields, technical implementation methods, technical problems and solutions; S2: Based on the constructed domain knowledge base, the knowledge base is updated in real time, and the definition and application of terms are dynamically adjusted according to different technical fields. All technical terms in patent documents are comprehensively analyzed with the standard terms in the knowledge base to evaluate the accuracy of patent classification and retrieval; S3: Analyze the meaning of terms and evaluate the semantic accuracy of text information by intelligently analyzing the terms in patent documents and the technical content in the context; S4: Based on accurate patent classification and retrieval and accurate semantics of text information, and combined with the technical context and domain knowledge of patent texts, establish a semantically accurate recognition mechanism in the classification and retrieval process.

2. The patent classification retrieval method based on multivariate data fusion according to claim 1 is characterized in that: The construction of a domain knowledge base in the field of patent-related technology specifically includes: For each technical field, the terminology and technical background information of the corresponding technical field are obtained, the terminology and technical background information of the patent document are analyzed, and based on the analysis results, a comprehensive feature index is generated. Based on the comprehensive feature index, the semantic consistency of the text in classification and retrieval is evaluated.

3. The patent classification retrieval method based on multivariate data fusion according to claim 2 is characterized in that: The process of obtaining the comprehensive characteristic index is as follows: Convert the terms and their background information in patent documents into vector representations; Convert each term and technical background into a vector representation of fixed dimension; For each term T in the patent document i , word vector representation is obtained through word embedding method Obtain technical background information B j , using document-based embedding to transform technical background information into background vectors For each term T in a patent document i and corresponding technical background information B j , the semantic matching degree between the term and the background is calculated by cosine similarity; The calculation expression of the cosine similarity is: In the formula, Indicates the similarity between the terms in the patent document and the technical background information, represents the dot product of the term vector and the background vector, represents the Euclidean norm of the term vector, represents the Euclidean norm of the background vector, i represents the i-th term in the patent document, and j represents the j-th term in the technical background information; For each term in the patent document, calculate the similarity with all technical background information and obtain the similarity vector, and calculate the comprehensive feature index by weighted average; Wherein, the calculation expression of the comprehensive characteristic index is: In the formula, C i represents the comprehensive feature index, n represents the total amount of technical background information, and w j Indicates technical background information B j The weight coefficient of .

4. The patent classification retrieval method based on multivariate data fusion according to claim 1 is characterized in that: The above mentioned comprehensive analysis of technical terms in all patent documents and standard terms in the knowledge base to evaluate the accuracy of patent classification and retrieval includes: Obtain technical terms in patent documents, calculate embedded representations on text and images in patent documents, and generate technical term accuracy coefficients based on the calculation results; Obtain standard terms in the knowledge base of patent documents, analyze the connectivity and semantic consistency of term nodes, and calculate the accuracy coefficient of standard terms based on the analysis results; The accuracy coefficients of technical terms and standard terms are comprehensively calculated and analyzed, and the accuracy coefficients are calculated based on the analysis results to evaluate the accuracy of patent classification and retrieval.

5. The patent classification retrieval method based on multivariate data fusion according to claim 4 is characterized in that: The process of obtaining the accuracy coefficient of the technical term is as follows: Represent the text part of the patent document as an embedding vector and extract the image features associated with the technical terms in the patent document; Fusion of text embedding and image embedding to form a multimodal embedding vector for terms Based on the multimodal embedding vectors, the text and image consistency scores of technical terms are calculated by cosine similarity; According to the frequency of occurrence of technical terms in the literature and the importance weight in the technical field knowledge base, the semantic weight of the technical terms is calculated. The calculation expression is: W = log(1+f)*w1; In the formula, W represents the semantic weight, w1 represents the importance weight, and f represents the frequency of occurrence of technical terms in patent documents; The consistency score and semantic weight are processed comprehensively to calculate the accuracy coefficient of technical terms. The calculation expression is: Where TTAC represents the technical term accuracy coefficient, λ represents the adjustment parameter, F represents the consistency score, and max(W) represents the maximum semantic weight in the term set.

6. The patent classification retrieval method based on multivariate data fusion according to claim 4 is characterized in that: The process of obtaining the standard term accuracy coefficient is as follows: Structuring term definitions, domain background, synonyms, and hyponyms in the knowledge base into graphs; Graph nodes represent terms, and graph edges represent semantic relationships between terms; The frequency of standard terms in the literature in the corresponding technical field is normalized and calculated to obtain the field weight parameter; Sum all domain weight parameters in the patent document to obtain the connectivity G of the semantic node; Calculate the semantic similarity between the standard term definition and its upper and lower nodes. The calculation expression is: Where, T a represents the standard term, a and b represent the number of standard terms in the patent document, sin represents the similarity between standard term vectors, N represents the total number of standard terms, S(T a ) indicates the semantic similarity between the standard term definition and its upper and lower nodes; The connectivity G of the language nodes is related to the semantic similarity S(T a ) to calculate the ratio and obtain the standard term accuracy coefficient, denoted as TTAB.

7. The patent classification retrieval method based on multivariate data fusion according to claim 1 is characterized in that: The meaning of the terms is analyzed to evaluate the semantic accuracy of the text information. include: Wherein, the said terms include technical terms and standard terms; Modeling the contextual content of terms in patent documents as a graph; The graph convolutional network is used to aggregate the neighborhood features of nodes layer by layer and update the semantic representation of the nodes. The calculation expression is: In the formula, represents the feature vector of node q in the l+1 layer, q represents the number of nodes, z represents the number of neighbor nodes, represents the set of neighbor nodes of node q, d q Indicates the number of connections between node q and node d z represents the number of connections of the neighbor node z, W (l) represents the weight matrix of the lth layer, l is the number of graph convolutional layers, b (l) represents the bias vector of the lth layer, σ represents the activation function; The updated feature vectors of term nodes and context nodes are used to calculate the semantic similarity between the term and the context through cosine similarity; The semantic accuracy coefficient of a term is calculated by aggregating the similarity scores of the term node and all its neighbor nodes. The calculation expression is: Where Y q represents the semantic accuracy coefficient, represents the set of neighbor nodes, x(q) represents the semantic similarity score between the term and the neighbor nodes, and q represents the number of nodes.

8. The patent classification retrieval method based on multivariate data fusion according to claim 1 is characterized in that: The establishment of a semantic accurate recognition mechanism in the classification and retrieval process specifically includes: The semantic accuracy coefficient of patent documents and the accuracy coefficient of patent document classification and retrieval are normalized to calculate the comprehensive recognition score. According to the comprehensive recognition score, the semantic accuracy of patent classification and retrieval is divided, and the semantics of document classification and retrieval are divided into qualified semantics and unqualified semantics.

9. Patent classification retrieval system based on multivariate data fusion, characterized by: The patent classification search method based on multivariate data fusion as claimed in any one of claims 1 to 8 comprises: A domain knowledge base construction module, which constructs a domain knowledge base in the patent-related technical field, including terminology and technical background information, to provide an accurate semantic benchmark for the patent text and ensure the semantic consistency of the text in classification and retrieval; A domain knowledge base update and terminology standardization module, which updates the domain knowledge base in real time, dynamically adjusts the definition and application of terms according to different technical fields, analyzes the matching between terms in patent documents and standard terms in the knowledge base, and evaluates the accuracy of patent classification and retrieval; A term semantic analysis and evaluation module, which evaluates the semantic accuracy of terms by intelligently analyzing the terms and their contexts in patent documents; The semantic accurate recognition and classification retrieval module is based on accurate patent classification and retrieval, combined with the technical context and domain knowledge of the patent text, to establish an effective semantic accurate recognition mechanism to improve the accuracy of patent text classification and retrieval.

Images