66 results about "Citation network" patented technology

In a corpus of scientific articles such as a digital library, documents are connected by citations and one document plays two different roles in the corpus: document itself and a citation of other documents. A Bernoulli Process Topic (BPT) model is provided which models the corpus at two levels: document level and citation level. In the BPT model, each document has two different representations in the latent topic space associated with its roles. Moreover, the multi-level hierarchical structure of the citation network is captured by a generative process involving a Bernoulli process. The distribution parameters of the BPT model are estimated by a variational approximation approach.

A visualization-based interactive legal research tool that generates from a multi-dimensional citation network a semantics-constrained citation sub-network that focuses on one individual issue in which a user is interested, and puts the sub-network on an interactive user interface (“UI”), which allows the researcher to browse, navigate, and jump over to start new sub-networks on different issues that are relevant to original issues.

The invention discloses a document citation network visualization and document recommendation method and system, and relates to the fields of document influence analysis and information visualization. The method comprises the following steps of calculating the importance degree of documents according to authors, time information, citation numbers and other inherent attributes, document similarities and transfer values generated by introducing behavior quantitative analysis, and sorting the documents; then performing clustering on the sorted documents and performing visualization on the clustered result to establish a dual-layer network model, and displaying important documents in a clear manner; and finally, recommending the documents in the center of clustering displayed in the visualization manner to a user. The document citation network visualization and document recommendation method and system are high in usability and can help science researchers to rapidly screen out most authoritative papers.

The present invention provides a method for patent valuation, comprising steps of providing a patent group for establishing a citation network and selecting a cluster within the citation network, wherein the citation network comprises a plurality of links, each link is formed between any two patent cases having citation relationship with each other; afterwards, designating weighting value for each link connecting each pair of patent cases; finally, calculating a plurality of centrality indices with respect to the plurality of patent cases in the cluster respectively according to the weighting value of each link. By means of the method of the present invention, it is capable of obtaining a relative value or significance of a specific patent case with respect to the other patent cases within the cluster.

Systems and methods facilitate a search and identify documents and associated metadata reflecting content of the documents. In one implementation, a method receives a query comprising a set of search terms, identifies a stored document in response to the query, and determines a score value for the retrieved document based on a similarity between one or more of the query search terms and metadata associated with the identified document. The method locates the identified document in a citation network of baseline query results, the citation network comprising a first set of documents that cite to the identified document and a second set of documents cited to by the identified document. The method further determines a new score value of the identified document as a function of the score value and a quantity and a quality of documents within the first and second set of documents.

The invention discloses a method for analyzing and sequencing academic influence of theme literature in a citation database. The method includes performing theme retrieval in the citation database to acquire metadata information of the theme literature and citation relation among the theme literature; according to the acquired metadata information of the theme literature and the acquired citation relation among the theme literature, creating a theme citation network; on the basis of the citation network, building a factor graph model containing various literature academic influence factors; using a circulating maximum sum algorithm for approximate reasoning on a factor graph to determine marginal probability values of all variables in the factor graph; according to a descending order result of the marginal probability values, acquiring and outputting an academic influence sequence of all theme literature. By the method, sequencing of the academic influence of literature with a theme given by a user can be realized.

A computer-implemented method of generating an issue library from a corpus of documents defining a citation network includes semantically linking, by a computing device, electronic documents within the corpus of documents by pairing reasons-for-citing in citing documents with cited-text-areas in cited documents and storing citation entries associated with the semantically-linked electronic documents in a citation-pairing metadata file. The method may further include searching the citation-pairing metadata file for reasons-for-citing and cited-text-areas that are semantically similar to at least one issue to thereby group similar documents and reasons-for-citing by issue, and storing information regarding groups of semantically-similar reasons-for-citing and cited-text-areas in a plurality of issue library metadata entities. Each issue library metadata entity is associated with an individual issue.

The invention provides an academic recommendation method and system based on a citation network and user proficiency analysis. The academic recommendation method includes the following steps that step 1, a model is adopted to perform analysis so as to obtain a first candidate set; step 2, the academic citation network is adopted to perform analysis so as to obtain a second candidate set; step 3, a third candidate set is obtained through the user proficiency analysis; step 4, the first candidate set, the second candidate set and the third candidate set are filtered and sorted to obtain a recommended candidate set; step 5, the recommended candidate set is displayed. Compared with the prior art, the academic recommendation method has the advantages that the academic recommendation personalization is enhanced and a recommendation system more suitable for researchers is created by combining with the analysis on the citation network and field proficiency on the basis of a traditional recommendation system.

A visualization-based interactive legal research tool that generates from a multi-dimensional citation network a semantics-constrained citation sub-network that focuses on one individual issue in which a user is interested, and puts the sub-network on an interactive user interface (“UI”), which allows the researcher to browse, navigate, and jump over to start new sub-networks on different issues that are relevant to original issues.

The invention provides a data recommendation method based on citation network community discovery, and the method comprises the following steps: constructing a citation network based on the cooperation relation between authors and the co-citation and coupling relation between papers; for the citation network, discovering a community network with similar or related research contents by utilizing amodular Louvain algorithm; based on the similarity between the paper and the data set, establishing association between the data set and the community network; and superposing and de-duplicating the paper nodes in the community network associated with the data set, and then performing data recommendation.

The invention discloses a method for assessing and sorting citation network academic influences based on credibility, and belongs to the technical field of academic influence assessment. By combining the characteristics of a citation network, a series of rules are defined according to background information of an article, a seed set mechanism is selected by improving a TrustRank algorithm and an Anti-TrustRank algorithm, and after a reputation value and a non-reputation value of a network node is obtained through cyclic iterative calculation, a score is obtained based on results of the reputation value and the non-reputation value; according to a descending sorting result of a final comprehensive reputation value, the result of academic influence sorting of scientific literature in the citation network is obtained and output. The invention aims to provide the reasonable and fair assessment method which can assess the influences of the literature accurately and then select a high-quality paper material in a subject, and researchers can also search for literature materials quickly, grasp a current hot research direction, and allocate more time in the study of scientific theory; the method for assessing and sorting the citation network academic influences based on the credibility has important theoretical significance to understanding the structure and propagation characteristics of the citation network and assessing the influences of the literature.

The invention provides a citation recommendation method based on deep learning and link prediction. The citation recommendation method comprises the following steps: step 1, obtaining feature information of all paper nodes in a known paper library and a citation network of the known paper library; step 2, respectively propagating the paper nodes in a citation network to obtain embedded representation of each paper node; step 3, inputting a paper node of a to-be-recommended citation, and calculating embedded representation of the paper node of the to-be-recommended citation; step 4, according to the embedded representation of the paper nodes of the citation to be recommended and the embedded representation of each paper node in the known paper library, recommending the citation to be recommended; and calculating cosine similarities between the paper nodes of the citation to be recommended and each paper node in the known paper library, and selecting the paper nodes corresponding to thefirst t cosine similarities as a citation list of the paper nodes of the citation to be recommended.

The invention provides a screening and valuing method of a technical similarity patent, comprising an establishing procedure of docking of stereotactic orientation analysis technology and citation networks TSU-CN microstructure, a procedure of establishing a related subnet, and a procedure of collecting monitoring data of internal and external technical flow based on target patent-similar technology-main technology environment, and a procedure of extracting technological spectrum parameters basic data and a procedure of establishing technological spectrum parameters based on a dynamic TSU-CN microstructure. By macroscopic and microcosmic docking of target patent technical flow in similar technology and main technology environment and stereotactic and quantitative control, relative objectivity and accuracy of valuing and screening results are guaranteed. Thus, enterprises are capable of positioning the 'technology value of target patent under the realistic background' reasonably and accurately in the introduction of technology and the development of foreign invalid patents so as to avoid the competitive failure caused by blind technology introduction and patent development. Meantime, the method helps enterprises make competitiveness evaluation on the technical patent in the international similar technology, as well as the option of stock investment.

The invention discloses a scientific and technological paper clustering analysis method based on variational diagram auto-encoder and K-Means, which comprises the following steps of: constructing a citation network G=(V, E, F) by utilizing existing scientific and technological paper data, and constructing a variational diagram auto-encoder consisting of an encoder and a decoder according to an adjacent matrix A of a citation relationship between papers and a characteristic matrix F of paper keyword attributes, taking minimization of distance measurement between a reconstructed adjacency matrixand an original adjacency matrix A and divergence of node representation vector distribution and normal distribution as targets, training in an unsupervised mode to obtain multi-dimensional Gaussiandistribution, and sampling from the distribution to obtain a low-dimensional embedded vector z of a node; and then clustering the low-dimensional embedded vector z by using the K-Means algorithm to obtain a division result of the science and technology paper, and performing two-dimensional visual display after dimension reduction by using a tSNE algorithm. According to the method, the accuracy ofscientific and technological paper clustering analysis is improved, and the calculation cost of analysis is reduced.

The invention provides a paper text similarity detection method based on a citation network. According to the method, candidate document set filtering based on combination of booklist coupling and semantic fingerprints is provided on the basis of a citation network. On the candidate document set, a sentence-level inverted index is established based on the words. The similar sentence detection andviewpoint fragment detection is performed and a similar text is generated. The calculation is performed to obtain a copy ratio of the to-be-detected document so as to judge the similarity of the papertext. According to the sentence comparison and viewpoint detection based on the word vectors, the word vectors and synonyms are introduced into text similarity calculation, the sentence similarity calculation effect is improved. The method has the advantage of being high in calculation speed, and text fragments possibly related to viewpoint plagiarism in the thesis text can be detected. The method is high in speed, and has a good effect on sentence similarity detection in the forms of word replacement, sentence recombination and the like.

The invention discloses a ranking method of gold citation authors based on paper citation relation, which comprises the following steps: firstly, selecting a paper group range in a paper database, obtaining the citation relation and the author information; Secondly, the initial paper citation network model is constructed according to the mutual citation relationship of papers in the paper group and the author information. Thirdly, calibrate the citation attributes of initial papers, establish self-citation revised citation coefficients, and generate self-citation excluded paper citation network; Then, author citation network is generated by linear mapping from other citation network. Finally, the author narrows down the scope of the citation network by recursive deletion until the result converges, outputs the gold citation authors, calculates the author importance parameters and sorts them. The paper author ranking method provided by the invention is easy to implement, and can effectively eliminate self-citation and low-quality and low-efficiency other-citation interference of papers, so as to help researchers to quickly and accurately rank authors' importance.

A visualization-based interactive legal research tool that generates from a multi-dimensional citation network a semantics-constrained citation sub-network that focuses on one individual issue in which a user is interested, and puts the sub-network on an interactive user interface (“UT”), which allows the researcher to browse, navigate, and jump over to start new sub-networks on different issues that are relevant to original issues.

A computer-implemented method of generating an issue library from a corpus of documents defining a citation network includes semantically linking, by a computing device, electronic documents within the corpus of documents by pairing reasons-for-citing in citing documents with cited-text-areas in cited documents and storing citation entries associated with the semantically-linked electronic documents in a citation-pairing metadata file. The method may further include searching the citation-pairing metadata file for reasons-for-citing and cited-text-areas that are semantically similar to at least one issue to thereby group similar documents and reasons-for-citing by issue, and storing information regarding groups of semantically-similar reasons-for-citing and cited-text-areas in a plurality of issue library metadata entities. Each issue library metadata entity is associated with an individual issue.

The invention discloses a literature selection and learning sequence determination method. The method comprises the steps of constructing a related citation network in the literature network based onthe search words and the mutual reference relationship; calculating the importance degree value of each literature on the basis of the frequency of being led; and eliminating the influence generated by publication years by using a standardized formula, deleting unimportant literature nodes in a related citation network, constructing a literature knowledge map, summing the literature node importance of literature learning paths in the literature knowledge map, and performing descending sort to obtain a literature learning path recommendation list. According to the method, the literature nodes in the literature learning path and the sequential learning relationship of the literature nodes, the problems of information overload, learning carelessness and learning backtracking in current literature learning are solved, the time consumption of literature selection is saved, the literature learning efficiency is improved, the literature learning effect is improved, and the method has an extremely good application prospect.

The invention provides a classification method based on a hyper-graph transformation network, which is used for solving the problem that in the prior art, high-order semantic information in a heterogeneous network cannot be deeply explored, so that classification cannot be accurately carried out. According to the method, an end-to-end hyper-graph transform network (HGTN) is provided, the communication ability between hyper-edge amplification nodes is used for learning a high-order relation, and semantic information between different types of nodes is mined. Specifically, an attention mechanism is utilized to distribute weights for different types of hyper-graphs, high-order semantic information implied in an original heterogeneous hyper-graph is subjected to cascade learning, useful meta-paths are generated, node embedding features are learned in an end-to-end mode, and a node classification task is completed. The method has good accuracy and universality, and is suitable for node classification tasks of heterogeneous networks such as citation networks, media networks, social networks and the like.

The embodiment of the invention provides a scientific research hotspot theme analysis method and device and electronic equipment, and the method comprises the steps of building a citation network based on scientific research literatures, and calculating the literature sorting weight of each document through a random walk algorithm, so as to filter the literatures, and obtain citation data and a corpus; and evaluating and screening the obtained citation data and corpus based on a probability topic hierarchical combination model to obtain a hotspot scientific research theme. According to the embodiment of the invention, the theme evolution diagram is constructed according to the literature structure information and the theme affiliation information, and theme description optimization, themesorting and theme evolution are combined through the theme evolution diagram, so that a more accurate result is obtained on the premise of improving the time efficiency.

The invention relates to a baseline based journal evaluation method and belongs to the technical field of computer natural language processing. The method comprises: firstly, establishing a network of baseline citations of a journal, so that a baseline evaluation value Vbaseline of the journal can be conveniently calculated; secondly, establishing a journal citation network, so that evaluation values Vcount of all citations of the journal can be conveniently calculated; and finally, combining the Vbaseline and the Vcount to obtain a baseline value based journal evaluation index V. Compared with the prior art, the method can comprehensively consider the attention degree and novelty of a thesis: a citation rate of the thesis represents the attention degree of the thesis, and the baseline value of the thesis represents the importance and novelty of the thesis. Through the baseline based journal evaluation index, the problem of two evaluation indexes for balancing the quantity and quality of the thesis is effectively solved, the importance of originality, novelty and pioneering performance of the journal is emphasized, and a fairer method for measuring the importance of the journal is provided.

The invention discloses an article recommendation method based on multi-attribute features, and belongs to the field of information processing. According to the method, more article features are extracted by using a multi-attribute article feature recommendation method, and the recommendation performance is improved: a struc2vec embedding vector based on an article quotation network, a metapath2vec embedding vector based on a heterogeneous network with article author and organization information, and an embedding vector of an article title and abstract content doc2vec are utilized, and on the basis of an original quotation network, through a graph reconstruction method, the embedding results of the isomorphic quotation network, the heterogeneous article network and the text information can be combined according to the weight. For a multi-attribute feature reconstruction network, graph embedding is carried out by using a method capable of combining structure information and homogeneous information, recommendation performance is improved, an embedding vector, containing the structure information and the homogeneous information, of an article node is obtained through a node2vec method, and finally recommendation is carried out through vector similarity.

The invention discloses a multi-view comparative learning-based citation network graph representation learning system and method. The citation network graph representation learning system comprises: a sample construction module which takes original graph node representation as a positive sample and constructs a negative corresponding sample based on an original graph; a graph enhancement module which is used for enhancing the positive sample node features based on a personalized page ranking algorithm and a Laplace smoothing algorithm to obtain a positive sample graph and a negative sample graph; a fusion module which is used for extracting the positive sample graph representation and the negative sample graph representation based on an encoder, integrating the positive sample graph representation and the negative sample graph representation, and obtaining the consensus representation of the positive sample graph and the negative sample graph through a cross view concentrated fusion layer; a mutual information estimation module which is used for comparing the learning representations of the positive sample pair and the negative sample pair through a discriminator; and a difficult sample mining module which represents the consistency between the negative sample pairs according to a pre-calculated affinity vector, and selects and retains nodes which are difficult to express global or neighbor information.

The invention discloses a field expert selection method based on a citation network and a scientific research cooperation network. The method comprises the following steps: firstly, based on metadatainformation of a database, constructing a scholar-cooperation directed network model, and generating a scholar-cooperation network; secondly, constructing a literature citation model based on the literature citation information, deleting self-citation interference, and performing linear mapping to generate a scholar citation network; then, fusing the scholar-cooperation network and the scholar citation network to generate a scholar relationship network; and finally, calculating important nodes in the scholar relationship network, dividing a community structure, and taking a result as a selection expert list. According to the method, two standards of academic ability evaluation and cooperative network quality evaluation are comprehensively considered according to academic activities and academic achievements of scholars, so that corresponding experts can be quickly and accurately recommended, the skilled fields of the experts can be identified, and the problem of inaccurate expert professional matching in an existing selection method is solved; and the importance of the deep schools in scientific research work is visually reflected, and the selected experts are more reasonable.

The invention discloses a golden citation algorithm-based paper recommendation method. The method comprises the following steps of: firstly, selecting a paper group range in a paper database; secondly, constructing an initial citation network model according to mutual citation relationships between papers in the selected paper group; thirdly, demarcating citation attributes of papers in the initial citation network model, and revising a citation frequency through a self-citation coefficient so as to generate a self-citation revising paper group; and finally, iteratively narrowing the range ofthe self-citation revising paper group until a result is converged, so as to obtain a golden citation paper. Through demarcating citation attributes of papers, interferences of self-citation in the paper citation relationships are eliminated; and paper groups are recursively narrowed according to self-citation revising citation frequencies, so that the interferences of low-quality and low-efficiency cited papers are eliminated, and users can rapidly and correctly retrieve high-quality papers.

The invention discloses a citation recommendation algorithm based on a heterogeneous network. The citation recommendation algorithm specifically comprises the following steps: S1, constructing a binary heterogeneous citation network; s2, initializing author node vector representation vA and text content vector representation; s3, generating a structural information-based pacer vector representation; S4, generating a text information-based pacer vector representation; S5, performing joint interaction and mutual enhancement; S6, repeating the steps S4-5 until the model converges, and completingtraining; s7, obtaining final vectors of all the pasters and authentic in the training set, and storing the trained model; s8, calling the trained model parameters to obtain the vector representationof each follower in the test set; and S9, calculating cosine similarity of each follower in the test set and all the folders in the training set, sorting the folders from large to small according to the similarity, and taking the first K folders as a final recommendation result. According to the invention, the algorithm performance can be improved by combining the structure information, and the unknown document can be predicted.

A science and technology paper classification method based on a stacked automatic encoder and a citation network comprises the following steps: building a citation network adjacency matrix and a labelvector matrix according to existing science and technology paper data, and calculating a probability transition matrix and a co-occurrence probability matrix; constructing a science and technology paper classification model based on the stacked automatic encoder and a long-term and short-term memory network; training the stacked automatic encoder and the long-term short-term memory network by using the co-occurrence probability matrix and the label vector matrix; and using the trained model to predict the classification result of the new paper. According to the method, the reference relationship among the papers is considered, the internal relationship among the nodes in the constructed reference network is effectively extracted and mapped into the low-dimensional embedded vector space, and the category feature information is acquired by using the long-term and short-term memory network, so that the category of the new papers is accurately predicted.