Method and device for extracting keywords based on document set
By clustering document sets and using large language models to determine the technical field, combined with keyword extraction algorithms, the accuracy and efficiency of keyword extraction in large-scale and multi-themed documents in the existing technology are solved, and more accurate and efficient keyword extraction.
Patent Information
- Application Number
- CN202411341024.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-25
- Publication Date
- 2024-12-24
AI Technical Summary
When existing keyword extraction technology deals with large-scale and multi-themed document sets, it is difficult to accurately distinguish the subtle theme differences in the document, resulting in the extracted keywords that cannot accurately reflect the true content of the text, and there is accuracy, explanatory and efficiency. The challenge.
By clustering document sets, clustering clusters with similar contents are formed, and the technical fields of clustering clusters are determined using large language models, so as to obtain relevant technical terms, and keyword extraction algorithms are used to extract clustering clusters. Increase the weight of technical terms to extract keywords that fit the content of the document more.
Improves the accuracy and relevance of keyword extraction, ensures that the extracted keywords can better reflect the real content of the document, and improves computing efficiency, especially when dealing with large-scale document sets.
Smart Images

Figure CN119179759A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of natural language processing, and in particular to a method and device for extracting keywords based on a document set. Background Art
[0002] As an effective text analysis tool, keyword extraction can quickly provide an overview of the text and help users understand and process large amounts of data. This technology is widely used in many fields such as information retrieval, content recommendation, text summary generation, retrieval enhancement generation RAG, etc.
[0003] Although existing keyword extraction technologies, such as BM25 based on statistical methods and keyBERT based on language models, have performed well in multiple application scenarios, they face multiple challenges in accuracy, interpretability, and efficiency when processing large-scale and multi-topic document sets. In particular, in terms of the relevance and precision of keywords, traditional methods are often unable to effectively distinguish subtle topic differences within a document, resulting in the extracted keywords not accurately reflecting the true content of the text. Therefore, a method is needed to extract keywords that are more in line with the content of the document. Summary of the invention
[0004] The present invention describes a method and device for extracting keywords based on a document set to solve the above problems.
[0005] In a first aspect, a method for extracting keywords based on a document set is provided, comprising:
[0006] Get a document set containing multiple documents;
[0007] Performing a clustering operation on the documents in the document set to obtain a plurality of document clusters;
[0008] For any target cluster in each document cluster, determine the first k documents that are close to the cluster center of the target cluster, input the k documents into the large language model, and let the large language model determine the target field to which the target cluster belongs based on the k documents;
[0009] Acquire target domain terms related to the target domain;
[0010] Based on the target domain terms, keywords are extracted from the target clusters to obtain a target keyword list.
[0011] In a possible implementation, it further includes:
[0012] Get the first keyword entered by the user;
[0013] Determine a number of document clusters in the keyword list that include the first keyword;
[0014] Based on the documents in the plurality of document clusters, a query result related to the first keyword is determined.
[0015] In a possible implementation, it further includes:
[0016] Obtain a prompt word to be input into the large language model, and extract a second keyword from it;
[0017] Determine a number of document clusters in the keyword list that contain the second keyword;
[0018] The prompt word and the documents in the plurality of document clusters are input into a large language model to obtain a response text of the large language model for the prompt word.
[0019] In a possible implementation, a clustering operation is performed on the documents in the document set to obtain a plurality of document clusters, including:
[0020] Performing embedding processing on each document in the document set to obtain a document embedding representation of each document;
[0021] Clustering operations are performed based on the embedded representations of each document to obtain multiple document clusters.
[0022] In a possible implementation, the clustering operation includes at least one of the following: K-means clustering, K-medoids clustering, and Gaussian mixture model clustering.
[0023] In a possible implementation, the k documents are input into a large language model, and the large language model is enabled to determine the target domain to which the target cluster belongs according to the k documents, including:
[0024] Acquire a first prompt word template; the first prompt word template is used to indicate the domain to which the input multiple documents belong as determined by the large language model;
[0025] Filling the k documents into a first prompt word template to obtain a first prompt word;
[0026] The first prompt word is input into the large language model to obtain the target domain.
[0027] In a possible implementation, based on the target domain terms, keyword extraction is performed on the target cluster to obtain a target keyword list, including:
[0028] Based on the keyword extraction algorithm, determining a plurality of candidate keywords in the target cluster and their corresponding scores;
[0029] Increase the score corresponding to each candidate keyword belonging to the target domain term by a first ratio;
[0030] Based on the updated candidate keyword scores, m candidate keywords with the highest scores are selected and determined as the target keyword list.
[0031] In a possible implementation, the keyword extraction algorithm includes at least one of the following: TF-IDF algorithm, BM25 algorithm, KeyBERT algorithm.
[0032] In a possible implementation, it further includes:
[0033] Obtaining a pre-set mapping relationship between a domain and a keyword extraction algorithm;
[0034] According to the mapping relationship and the target domain, a target keyword extraction algorithm corresponding to the target cluster is determined.
[0035] In a second aspect, a device for extracting keywords based on a document set is provided, comprising:
[0036] An acquisition unit is configured to acquire a document set including a plurality of documents;
[0037] A clustering unit configured to perform a clustering operation on the documents in the document set to obtain a plurality of document clusters;
[0038] A domain determination unit is configured to, for any target cluster in each document cluster, determine the first k documents that are close to the cluster center of the target cluster, input the k documents into the large language model, and let the large language model determine the target domain to which the target cluster belongs based on the k documents;
[0039] A term determination unit configured to obtain target domain terms related to the target domain;
[0040] The keyword extraction unit is configured to extract keywords from the target cluster based on the target domain terms to obtain a target keyword list.
[0041] According to a third aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed in a computer, the computer is caused to execute the method of the first aspect.
[0042] According to a fourth aspect, a computing device is provided, comprising a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the method of the first aspect is implemented.
[0043] The present invention proposes a method and device for extracting keywords based on a document set. The method first clusters documents based on their contents, and clusters documents with similar contents into the same cluster. Then, a large language model is used to assist in determining the technical field to which the document contents in the cluster belong as the technical field of the cluster. Next, technical terms related to the technical field are obtained from external prior knowledge. Finally, taking the cluster as a unit, a keyword extraction algorithm is used to extract keywords from the documents in the cluster, and the weight of the technical terms in the candidate keywords is increased to extract keywords in the document cluster in a more targeted manner. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] In order to more clearly illustrate the technical solutions of the multiple embodiments disclosed in this specification, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only the multiple embodiments disclosed in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0045] Figure 1 A schematic diagram of an implementation scenario of a method for extracting keywords based on a document set disclosed in an embodiment of the present invention;
[0046] Figure 2 A flowchart of a method for extracting keywords based on a document set disclosed in an embodiment of the present invention;
[0047] Figure 3 A schematic block diagram of a device for extracting keywords based on a document set disclosed in an embodiment of the present invention. DETAILED DESCRIPTION
[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.
[0049] Figure 1 Schematic diagram of an implementation scenario of a method for extracting keywords based on a document set disclosed in an embodiment of the present invention. Figure 1As shown, the document set contains multiple documents, represented by small circles. First, the documents in the document set are clustered to obtain multiple document clusters, represented by dotted ellipses. The documents represented by small circles in the same dotted ellipse belong to the same cluster. Then, the same operation process is performed on each cluster, and the cluster in the upper right corner is used as an example to illustrate. For this cluster, several documents (represented by gray small circles) that are close to the cluster center of the cluster are selected as representative documents of the cluster, and are input into the pre-trained large language model. The large language model uses these representative documents to determine the field to which the cluster belongs. Next, domain terms related to the field are obtained from external knowledge. Finally, with the assistance of domain terms, keyword extraction is performed on the cluster to obtain a corresponding keyword list.
[0050] It should be noted that Figure 1 It does not constitute a limitation to the technical solution of the present invention.
[0051] The specific implementation steps of the above-mentioned method for extracting keywords based on a document set are described below in conjunction with a specific embodiment.
[0052] Figure 2 Flow chart of a method for extracting keywords based on a document set disclosed in an embodiment of the present invention. Figure 2 As shown, the method at least includes: step 201, obtaining a document set including multiple documents; step 202, clustering the documents in the document set to obtain multiple document clusters; step 203, for any target cluster in each document cluster, determining the first k documents that are close to the cluster center of the target cluster, inputting the k documents into a large language model, and allowing the large language model to determine the target field to which the target cluster belongs based on the k documents; step 204, obtaining target field terms related to the target field; step 205, based on the target field terms, performing keyword extraction on the target cluster to obtain a target keyword list.
[0053] The specific execution process of each of the above steps is described below.
[0054] First, in step 201, a document set including multiple documents is obtained.
[0055] The document contents in the document set may belong to multiple technical fields. If keyword extraction is performed directly, it is difficult for the existing keyword extraction algorithm to accurately and reasonably use these documents for some documents, resulting in unsatisfactory results of the keyword list finally proposed. Therefore, the present invention first clusters the document set, and then extracts keywords for each cluster respectively, so that the keyword extraction algorithm can be performed in a more accurate document environment.
[0056] Then, in step 202, a clustering operation is performed on the documents in the document set to obtain a plurality of document clusters.
[0057] In one embodiment, step 202 specifically includes step 2021 and step 2022 .
[0058] In step 2021, each document in the document set is embedded to obtain a document embedding representation of each document.
[0059] The embedding process may use existing embedding algorithms, such as word2vec, BERT (Bidirectional Encoder Representations from Transformers), etc., which are not limited here.
[0060] Then, in step 2022, a clustering operation is performed based on the embedding representation of each document to obtain a plurality of document clusters.
[0061] In a more specific embodiment, the clustering operation in step 2022 includes at least one of the following: K-means clustering, K-medoids clustering, and Gaussian mixture model clustering.
[0062] In other embodiments, clustering may also be performed directly based on the similarity of text content, which is not limited here.
[0063] Next, in step 203, for any target cluster in each document cluster, the first k documents that are close to the cluster center of the target cluster are determined, and the k documents are input into the large language model, so that the large language model determines the target field to which the target cluster belongs based on the k documents.
[0064] The first k documents that are close to the cluster center can be regarded as the documents that best represent the cluster. Using these documents as representative documents of the target cluster and inputting them into the large language model can enable the large language model to better determine the target field to which the target cluster belongs.
[0065] In one embodiment, step 203 of inputting the k documents into a large language model, and allowing the large language model to determine the target domain to which the target cluster belongs according to the k documents, specifically includes steps 2031 to 2033.
[0066] In step 2031, a first prompt word template is obtained; the first prompt word template is used to indicate the domain to which the large language model determines that the input multiple documents belong.
[0067] Exemplarily, the first prompt word template may be as follows:
[0068] “You are a professional text classification assistant. Your task is to analyze multiple given documents and determine the fields or topics to which they belong.
[0069] Please read the following documents carefully:
[0070] [Insert document content here]
[0071] Based on the content of these documents, please perform the following steps:
[0072] Identify keywords, terms, and themes in each document.
[0073] Look for commonalities and similarities between these documents.
[0074] Consider the different areas or disciplines these documents might cover.
[0075] Based on your analysis, identify a domain or discipline that best captures the common themes across all documents.”
[0076] In step 2032, the k documents are filled into a first prompt word template to obtain a first prompt word.
[0077] The first prompt word is obtained by filling k documents into the corresponding spaces in the first prompt word template. In the above example, the space is "[insert document content here]".
[0078] In step 2033, the first prompt word is input into a large language model to obtain a target domain.
[0079] By applying the process of step 203 to each document cluster respectively, the field to which each document cluster belongs is obtained.
[0080] Then, in step 204, target domain terms related to the target domain are obtained.
[0081] The target domain terminology can be obtained from an external prior knowledge base, such as a technical dictionary and the like.
[0082] Finally, in step 205, based on the target domain terms, keyword extraction is performed on the target cluster to obtain a target keyword list.
[0083] When extracting keywords from the target cluster, by increasing the weight of each field term, keywords related to the technical field can be better captured.
[0084] For example, in a professional paper, a technical term may only be mentioned 1-2 times, and traditional keyword extraction algorithms may find it difficult to regard it as a keyword. However, by increasing the weight of the technical term, the probability of capturing the technical term can be increased.
[0085] In one embodiment, step 205 specifically includes steps 2051 to 2053 .
[0086] In step 2051, based on a keyword extraction algorithm, a plurality of candidate keywords in the target cluster and their corresponding scores are determined.
[0087] The keyword extraction algorithm will use the corresponding method to calculate the score (or weight, confidence, etc.) of each word (candidate keyword) in the document.
[0088] In a more specific embodiment, the keyword extraction algorithm includes at least one of the following: TF-IDF algorithm (term frequency-inverse document frequency), BM25 algorithm (BestMatching 25), KeyBERT algorithm.
[0089] In step 2052, the scores corresponding to the candidate keywords belonging to the target domain terms are increased by a first ratio.
[0090] If a candidate keyword is a target domain term, its score is increased.
[0091] In step 2053, based on the updated candidate keyword scores, m candidate keywords with higher scores are selected and determined as the target keyword list.
[0092] By using the process from step 204 to step 205 for each document cluster respectively, a keyword list corresponding to each document cluster can be obtained, and the keyword extraction process is completed.
[0093] In some possible implementations, before step 205, the method further includes steps 212 to 213:
[0094] In step 212, a preset mapping relationship between a domain and a keyword extraction algorithm is obtained.
[0095] The mapping relationship can be preset by engineers based on previous practical experience. A large language model can also be used to select a more appropriate keyword extraction algorithm based on the characteristics of different fields.
[0096] In step 213, a target keyword extraction algorithm corresponding to the target cluster is determined according to the mapping relationship and the target domain.
[0097] By using different keyword extraction algorithms for different clusters according to the characteristics of the documents in the clusters, the accuracy and relevance of the extraction can be further improved.
[0098] Steps 201 to 205 obtain a keyword list of each cluster, and keyword recall can be performed based on the keyword list, or retrieval-augmented generation (RAG) of a large language model can be performed.
[0099] In some possible implementations, keyword recall may be performed. After step 205 , the method further includes steps 206 to 208 .
[0100] In step 206, the first keyword input by the user is obtained.
[0101] In step 207, a number of document clusters containing the first keyword in the keyword list are determined.
[0102] In step 208, based on the documents in the plurality of document clusters, a query result about the first keyword is determined.
[0103] Step 208 may use an existing keyword recall method, which will not be described in detail here.
[0104] By predetermining a number of document clusters containing the first keyword, the speed and accuracy of keyword recall can be improved.
[0105] In some possible implementations, retrieval enhancement generation of a large language model may be performed. After step 205 , the method further includes steps 209 to 211 .
[0106] In step 209, the prompt words to be input into the large language model are obtained, and the second keywords are extracted therefrom.
[0107] In step 210, a number of document clusters containing the second keyword in the keyword list are determined.
[0108] In step 211, the prompt word and the documents in the plurality of document clusters are input into a large language model to obtain a response text of the large language model for the prompt word.
[0109] By using prompt words to retrieve related document clusters, the scope of "retrieval enhancement" in retrieval enhancement generation can be narrowed.
[0110] The present invention proposes a method for extracting keywords based on a document set, clustering the entire document set, and grouping the documents according to content similarity through a clustering algorithm to form multiple clusters with similar themes or contents. Each cluster represents a relatively homogeneous document cluster, focusing on one or more common themes. This pre-grouping not only narrows the target range of keyword extraction, allowing it to be performed in a more accurate document environment, but also due to the theme consistency of the documents within the cluster, the extracted keywords will have higher theme relevance and document representativeness. Afterwards, the keyword extraction operation is implemented within each cluster, and a specific algorithm adapted to the characteristics of each cluster can be used.
[0111] The present invention improves the relevance of keywords through clustering and optimizes the accuracy of keyword extraction. When the entire document set contains a variety of different topics, the traditional global keyword extraction method may be biased towards common keywords with higher frequencies and ignore those keywords that are only important in certain specific documents. Through the clustering method, it can be ensured that even topics that are only important in a small range can be covered by their keywords, and the keywords between different documents can be separated to the maximum extent.
[0112] In addition, the present invention improves computational efficiency by performing keyword extraction through clustering. Although clustering itself is an additional computational step, by dividing a large number of documents into smaller, thematically similar subsets, the computational load of keyword extraction on each subset can be effectively reduced. When processing large-scale document sets, this method may be more efficient than directly applying a keyword extraction algorithm to the entire data set.
[0113] According to another embodiment, a device for extracting keywords based on a document set is also provided. Figure 3 This is a schematic block diagram of a device for extracting keywords based on a document set disclosed in an embodiment of the present invention. The device can be deployed in any device, platform or device cluster with computing and processing capabilities. Figure 3 As shown, the device 300 includes:
[0114] A first acquisition unit 301 is configured to acquire a document set including a plurality of documents;
[0115] The clustering unit 302 is configured to perform a clustering operation on the documents in the document set to obtain a plurality of document clusters;
[0116] The domain determination unit 303 is configured to, for any target cluster in each document cluster, determine the first k documents that are close to the cluster center of the target cluster, input the k documents into the large language model, and let the large language model determine the target domain to which the target cluster belongs based on the k documents;
[0117] The term determination unit 304 is configured to obtain target domain terms related to the target domain;
[0118] The keyword extraction unit 305 is configured to extract keywords from the target cluster based on the target domain terms to obtain a target keyword list.
[0119] In a possible implementation manner, the device 300 further includes:
[0120] The second acquisition unit 306 is configured to acquire a first keyword input by a user;
[0121] A first determining unit 307 is configured to determine a number of document clusters in the keyword list that contain the first keyword;
[0122] The second determining unit 308 is configured to determine a query result about the first keyword based on the documents in the plurality of document clusters.
[0123] In a possible implementation manner, the device 300 further includes:
[0124] The third acquisition unit 309 is configured to acquire the prompt word to be input into the large language model, and extract the second keyword therefrom;
[0125] The third determining unit 310 is configured to determine a number of document clusters in the keyword list that contain the second keyword;
[0126] The model calling unit 311 is configured to input the prompt word and the documents in the plurality of document clusters into the large language model to obtain the answer text of the large language model for the prompt word.
[0127] In a possible implementation manner, the device 300 further includes:
[0128] The fourth acquisition unit 312 is configured to acquire a preset mapping relationship between a field and a keyword extraction algorithm;
[0129] The fourth determining unit 313 is configured to determine a target keyword extraction algorithm corresponding to the target cluster according to the mapping relationship and the target domain.
[0130] According to another embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed in a computer, the computer is caused to execute the method described in any one of the above embodiments.
[0131] According to yet another embodiment, a computing device is provided, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the method described in any one of the above embodiments is implemented.
[0132] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0133] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0134] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0135] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.
[0136] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for extracting keywords based on a document set, comprising: Get a document set containing multiple documents; Performing a clustering operation on the documents in the document set to obtain a plurality of document clusters; For any target cluster in each document cluster, determine the first k documents that are close to the cluster center of the target cluster, input the k documents into the large language model, and let the large language model determine the target field to which the target cluster belongs based on the k documents; Acquire target domain terms related to the target domain; Based on the target domain terms, keywords are extracted from the target clusters to obtain a target keyword list.
2. The method according to claim 1, further comprising: Get the first keyword entered by the user; Determine a number of document clusters in the keyword list that include the first keyword; Based on the documents in the plurality of document clusters, a query result related to the first keyword is determined.
3. The method according to claim 1, further comprising: Obtain a prompt word to be input into the large language model, and extract a second keyword from it; Determine a number of document clusters in the keyword list that contain the second keyword; The prompt word and the documents in the plurality of document clusters are input into a large language model to obtain a response text of the large language model for the prompt word.
4. The method according to claim 1, performing a clustering operation on the documents in the document set to obtain a plurality of document clusters, comprising: Performing embedding processing on each document in the document set to obtain a document embedding representation of each document; Clustering operations are performed based on the embedded representations of each document to obtain multiple document clusters.
5. The method according to claim 4, characterized in that The clustering operation includes at least one of the following: K-means clustering, K-medoids clustering, and Gaussian mixture model clustering.
6. The method according to claim 1, inputting the k documents into a large language model, and allowing the large language model to determine the target domain to which the target cluster belongs based on the k documents, comprising: Obtaining a first prompt word template; The first prompt word template is used to indicate the domain to which the input multiple documents belong to, determined by the large language model; Filling the k documents into a first prompt word template to obtain a first prompt word; The first prompt word is input into the large language model to obtain the target domain.
7. The method according to claim 1, based on the target domain terms, extracting keywords from the target cluster to obtain a target keyword list, comprises: Based on the keyword extraction algorithm, determining a plurality of candidate keywords in the target cluster and their corresponding scores; Increase the score corresponding to each candidate keyword belonging to the target domain term by a first ratio; Based on the updated candidate keyword scores, m candidate keywords with the highest scores are selected and determined as the target keyword list.
8. The method according to claim 7, characterized in that The keyword extraction algorithm includes at least one of the following: TF-IDF algorithm, BM25 algorithm, KeyBERT algorithm.
9. The method according to claim 8, further comprising: Obtaining a pre-set mapping relationship between a domain and a keyword extraction algorithm; According to the mapping relationship and the target domain, a target keyword extraction algorithm corresponding to the target cluster is determined.
10. A device for extracting keywords based on a document set, comprising: An acquisition unit is configured to acquire a document set including a plurality of documents; A clustering unit configured to perform a clustering operation on the documents in the document set to obtain a plurality of document clusters; A domain determination unit is configured to, for any target cluster in each document cluster, determine the first k documents that are close to the cluster center of the target cluster, input the k documents into the large language model, and let the large language model determine the target domain to which the target cluster belongs based on the k documents; A term determination unit configured to obtain target domain terms related to the target domain; The keyword extraction unit is configured to extract keywords from the target cluster based on the target domain terms to obtain a target keyword list.
Citation Information
Patent Citations
System and method for automatically extracting Chinese technical terms
CN101354712A
Keyword calculation method based on document clustering
CN105159998A
Semantic retrieval method and system for automatically extracting keywords
CN118468881A
Cited By
Document clustering method and system, computer program product and electronic device
WO2026208152A1
