A scientific research information retrieval and tracking method and device

Abstract

The application discloses a scientific research information retrieval tracking method and device, the method comprises the following steps: S1, obtaining retrieval keyword information; S2, performing retrieval processing on the retrieval keyword information to obtain preliminary retrieval information; S3, performing tracking retrieval processing on the preliminary retrieval information to obtain a retrieval tracking information set; the S3 comprises: setting a retrieval time interval; in each retrieval time interval, using a large model tool to perform semantic expansion on the retrieval keyword information to obtain first expanded keyword information; using a knowledge graph to perform structural expansion on the retrieval keyword information to obtain second expanded keyword information; based on a scientific research literature database, performing retrieval processing on the first expanded keyword information and the second expanded keyword information to obtain a tracking literature information set; performing fusion processing on the tracking literature information set and the preliminary retrieval information to obtain the retrieval tracking information set in the retrieval time interval.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of text data processing, graph data network processing and artificial intelligence technology, and in particular to a scientific research information retrieval and tracking method and device. BACKGROUND

[0002] In scientific research activities, timely and accurate acquisition and tracking of relevant field scientific research information is the basis for carrying out research work. With the development of information technology, scientific research literature data presents an explosive growth, and the traditional scientific research information retrieval method gradually exposes many deficiencies.

[0003] Currently, scientific researchers mainly retrieve in academic databases through keyword matching, but this method highly depends on the accuracy and comprehensiveness of the initial keywords, and is prone to missing due to keyword setting limitations. At the same time, the terminology in the field of scientific research has rich semantic associations, such as synonyms, hypernyms, and application scenario words, and simple keyword matching cannot cover these semantic associations, resulting in insufficient relevance and completeness of the retrieval results. Traditional database retrieval relies on precise keywords, and is prone to miss relevant literature due to terminology differences (such as "deep learning" and "deep neural network").

[0004] In addition, scientific research information has the characteristics of dynamic updating, and new research results and literature will continue to emerge, and scientific researchers need to track the progress of the field regularly. However, the existing tracking methods mostly rely on manual repeated retrieval, which is not only time-consuming and laborious, but also difficult to ensure the timeliness and continuity of tracking. Although some tools introduce simple keyword expansion functions, they lack deep mining of semantic associations and knowledge structures, and the expansion effect is limited, and still cannot meet the needs of scientific researchers for comprehensive and accurate tracking of scientific research information. SUMMARY

[0005] The present application mainly solves the problem of how to accurately and timely retrieve and track the relevant literature information of scientific research keywords based on semantic information association, and discloses a scientific research information retrieval and tracking method and device.

[0006] In a first aspect, a scientific research information retrieval and tracking method is disclosed, comprising:

[0007] S1, acquiring retrieval keyword information;

[0008] S2, performing retrieval processing on the retrieval keyword information to obtain preliminary retrieval information;

[0009] S3, performing tracking retrieval processing on the preliminary retrieval information to obtain a set of retrieval tracking information. The retrieval processing on the retrieval keyword information to obtain preliminary retrieval information comprises:

[0010] S21, using a scientific literature database, performing retrieval processing on the retrieval key information to obtain a retrieval literature information set;

[0011] S22, performing first mapping processing on the retrieval literature information set to obtain a first matrix set;

[0012] S23, performing second mapping processing on the retrieval literature information set to obtain a second matrix set;

[0013] S24, performing fusion clustering on the first matrix set and the second matrix set to obtain a fusion matrix set;

[0014] S25, performing vector text conversion processing on the fusion matrix set to obtain preliminary retrieval information.

[0015] The tracking retrieval processing on the preliminary retrieval information obtains a retrieval tracking information set, comprising:

[0016] S31, setting a retrieval time interval;

[0017] S32, in each retrieval time interval, using a large model tool to perform semantic expansion on the retrieval keyword information to obtain first expanded keyword information;

[0018] S33, using a knowledge graph to perform structural expansion on the retrieval keyword information to obtain second expanded keyword information;

[0019] S34, based on a scientific literature database, performing retrieval processing on the first expanded keyword information and the second expanded keyword information to obtain a tracking literature information set;

[0020] S35, performing fusion processing on the tracking literature information set and the preliminary retrieval information to obtain a retrieval tracking information set in the retrieval time interval;

[0021] S36, in each retrieval time interval, performing S32 to S35 to obtain a retrieval tracking information set for each retrieval time interval.

[0022] The large model tool is used to perform semantic expansion on the retrieval keyword information to obtain first expanded keyword information, comprising:

[0023] S321, using a context understanding module of a large model tool to generate word information related to the semantic of the retrieval keyword information;

[0024] S322, encoding the retrieval keyword information and the related word information into vector information respectively through a word encoding method;

[0025] S323, cluster all the vector information to obtain category information containing the largest number of vector information;

[0026] S324, convert all the vector information of the category information into vector text to obtain first extended keyword information.

[0027] The structure expansion of the search keyword information by using the knowledge graph to obtain second extended keyword information, comprising:

[0028] S331, using a path traversal algorithm, starting from the entity corresponding to the search keyword information in the knowledge graph, traversing to obtain an entity set within N1-hop relationships;

[0029] S332, filtering the traversed entity set to obtain second extended keyword information.

[0030] The filtering of the traversed entity set to obtain second extended keyword information, comprising:

[0031] S3321, using a knowledge graph embedding model, converting the entity corresponding to the search keyword information in the knowledge graph into a standard entity vector, and converting the relationship information set of the entity corresponding to the search keyword information in the knowledge graph into a standard relationship matrix; Each row vector of the standard relationship matrix is obtained by converting each relationship information of the relationship information set of the entity;

[0032] S3322, using a knowledge graph embedding model, converting the traversed entity set to obtain a traversal entity matrix, and converting the relationship information set of each entity of the traversed entity set to obtain a corresponding traversal relationship matrix;

[0033] S3323, filtering the standard entity vector, the standard relationship matrix, the traversal entity matrix and the traversal relationship matrix to obtain a filtered entity set;

[0034] S3324, using the filtered entity set and the search keyword information to construct second extended keyword information.

[0035] The filtering of the standard entity vector, the standard relationship matrix, the traversal entity matrix and the traversal relationship matrix to obtain a filtered entity set, comprising:

[0036] Subtracting each row vector of the standard entity vector and the traversal entity matrix to obtain a corresponding difference vector;

[0037] Using each difference vector as a row vector to construct a difference matrix C;

[0038] performing multivariate variational modal decomposition transformation on the row vectors of the difference matrix to obtain corresponding transformation vectors;

[0039] constructing a transformation matrix F by using each transformation vector as a row vector;

[0040] performing feature calculation on the transformation matrix and the difference matrix to obtain a feature matrix T;

[0041] The expression of the feature calculation is:

[0042] T = C 1 / 2 FC -1 / 2 ,

[0043] wherein T is the feature matrix;

[0044] calculating rank value α, norm value β, core norm γ and singular value from the feature matrix T;

[0045] judging whether each singular value satisfies a first discrimination inequality to obtain a first discrimination result;

[0046] The expression of the first discrimination inequality is:

[0047]

[0048] wherein N is the row dimension of the feature matrix T;

[0049] constructing a first screening entity set by using entities of row vectors corresponding to singular values whose first discrimination results are yes;

[0050] performing deep screening processing on the traversal relation matrix and the standard relation matrix corresponding to the first screening entity set to obtain a screening entity set.

[0051] In a second aspect, the present application discloses a scientific research information retrieval and tracking device, which comprises:

[0052] a memory storing executable program codes;

[0053] a processor coupled with the memory;

[0054] the processor calls the executable program codes stored in the memory to execute the scientific research information retrieval and tracking method.

[0055] In a third aspect, the present application discloses a computer storage medium storing computer instructions, which are called by a computer to execute the scientific research information retrieval and tracking method.

[0056] The fourth aspect of the embodiment of the present application discloses an information data processing terminal, which is used for realizing the scientific research information retrieval tracking method.

[0057] The present application has the following advantages:

[0058] The present application combines scientific research literature databases, large model tools and knowledge graphs to build an integrated scientific research information retrieval tracking system, effectively solving the problems of keyword limitations and insufficient semantic association mining in traditional retrieval methods. By expanding the semantics and structure of the retrieval keywords, the present application can cover a wider range of related terms and knowledge associations, significantly improving the comprehensiveness of the retrieved information and reducing the missed literature caused by insufficient keyword settings.

[0059] The present application converts the retrieved literature information into a matrix form through first mapping processing and second mapping processing, and performs fusion clustering, realizing the deep integration and structured processing of the literature information, making the preliminary retrieval information more systematic and relevant, and facilitating researchers to quickly grasp the core content of the field.

[0060] The present application sets a retrieval time interval and automatically performs tracking retrieval, combines the semantic understanding ability of the large model and the structured knowledge of the knowledge graph, dynamically generates expanded keywords, realizes the automation and periodic tracking of scientific research information, ensures the timeliness and continuity of the tracking information, and reduces the manual tracking burden of researchers.

[0061] The present application introduces vector encoding and clustering processing in semantic expansion, and uses knowledge graph embedding models and multi-dimensional matrix screening in structural expansion, ensuring the high relevance of the expanded keywords and the original keywords, avoiding irrelevant information interference, and improving the accuracy of tracking retrieval.

[0062] The present application realizes the accumulation and update of scientific research information through the fusion processing of the tracking literature information and the preliminary retrieval information, forms a complete retrieval tracking information set, provides full-process information support for researchers from initial retrieval to dynamic tracking, and helps to improve the research efficiency and innovation ability. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 The present application is a method for implementing the flowchart. DETAILED DESCRIPTION

[0064] In order to better understand the content of the present application, an embodiment is given.

[0065] Figure 1 The present application is a method for implementing the flowchart.

[0066] The first aspect of the embodiment of the present application discloses a scientific research information retrieval tracking method, comprising:

[0067] S1, obtaining retrieval keyword information;

[0068] S2, performing retrieval processing on the retrieval keyword information to obtain preliminary retrieval information;

[0069] S3, performing tracking retrieval processing on the preliminary retrieval information to obtain a set of retrieval tracking information;

[0070] The retrieval processing on the retrieval keyword information to obtain preliminary retrieval information comprises:

[0071] S21, performing retrieval processing on the retrieval keyword information by using a scientific research literature database to obtain a set of retrieval literature information;

[0072] S22, performing first mapping processing on the set of retrieval literature information to obtain a first matrix set;

[0073] S23, performing second mapping processing on the set of retrieval literature information to obtain a second matrix set;

[0074] S24, performing fusion clustering on the first matrix set and the second matrix set to obtain a fusion matrix set;

[0075] S25, performing vector text conversion processing on the fusion matrix set to obtain preliminary retrieval information;

[0076] The first mapping processing or the second mapping processing is realized by a method of text-to-graph data conversion, comprising:

[0077] S2201, performing preprocessing on the set of retrieval literature information to obtain a preprocessed text data set;

[0078] S2202, performing entity extraction processing on each data subset of a literature in the preprocessed text data set to obtain a corresponding set of entity nodes;

[0079] S2203, performing relationship extraction processing on each data subset of a literature in the preprocessed text data set to obtain a corresponding set of edge nodes;

[0080] S2204, constructing a graph object data corresponding to a literature based on the set of entity nodes and the set of edge nodes;

[0081] S2205, performing Euclidean space network embedding representation on the graph object data to obtain a first matrix corresponding to the literature; and performing hyperbolic space network embedding representation on the graph object data to obtain a second matrix corresponding to the literature;

[0082] S2206, a first matrix set is constructed by using the first matrix of all literatures, and a second matrix set is constructed by using the second matrix of all literatures.

[0083] The fusion clustering of the first matrix set and the second matrix set obtains a fusion matrix set, and the fusion clustering comprises:

[0084] A clustering method model set is constructed, and the clustering method model set comprises a K-means clustering method, a spectral clustering method and a hierarchical clustering method.

[0085] The first matrix set is clustered by using the clustering method model set to obtain a first clustering information set, and the first clustering information comprises label information of each first matrix corresponding to a literature under the operation of the three clustering methods.

[0086] The first clustering information set is fused based on a Co-association Matrix method to obtain first fusion label information of the literature corresponding to each first matrix.

[0087] The second matrix set is clustered by using the clustering method model set to obtain a second clustering information set, and the second clustering information comprises label information of each second matrix corresponding to a literature under the operation of the three clustering methods.

[0088] The second clustering information set is fused based on a Co-association Matrix method to obtain second fusion label information of the literature corresponding to each second matrix.

[0089] The first fusion label information and the second fusion label information of all literatures are clustered and fused by using a meta-feature clustering method to obtain fusion label information of all literatures.

[0090] The fusion matrix set is constructed by using the first matrix or the second matrix of the literature belonging to the same fusion label information.

[0091] The process of the fusion clustering of the first matrix set and the second matrix set to obtain the fusion matrix set is improved in robustness and anti-interference by multi-clustering method fusion.

[0092] The process first builds a clustering method model set containing K-means, spectral clustering, and hierarchical clustering, and performs clustering on the first matrix set (Euclidean space embedding) and the second matrix set (hyperbolic space embedding) respectively. K-means is good at processing linearly separable data with spherical distribution, and is suitable for capturing the local density of literature features; spectral clustering is more sensitive to data with non-linear and complex topological structure, and can identify implicit relationships between literature; hierarchical clustering can preserve the hierarchical relationship between literature (such as the progression of "basic theory-derivative method-application scenario"), and is suitable for reflecting the evolution logic of research topics.

[0093] Through the combination of multiple clustering methods, the inherent limitations of single clustering algorithm (such as K-means sensitivity to initial center and spectral clustering sensitivity to noise) are avoided. Subsequent fusion of clustering results based on co-association matrix (Co-association Matrix) (the frequency of statistical samples being classified into the same category in different clusters) can filter random errors and strengthen consensus information, and the final first / second fusion label is closer to the real clustering structure of literature, improving the robustness of clustering within a single space.

[0094] Integrating the complementary features of Euclidean and hyperbolic spaces enhances the comprehensiveness of clustering. The first matrix set is generated based on Euclidean space network embedding, which is good at capturing the local linear correlation of literature features (such as the association of similar research methods and the same experimental data); the second matrix set is generated based on hyperbolic space network embedding, which is more suitable for describing global hierarchical structure (such as the literature concept hierarchy from "machine learning" to "deep learning" to "Transformer"). By clustering and fusing the two types of matrix sets respectively, we can first fully explore the literature rules under different space characteristics: Euclidean space clustering focuses on "local similarity", and hyperbolic space clustering focuses on "global hierarchy". The two types of fusion labels carry different dimensional features of the literature, laying the foundation for subsequent cross-space fusion.

[0095] The meta-feature clustering realizes the complementarity of cross-space information and improves the accuracy of the final clustering. After obtaining the first fusion label (Euclidean space feature) and the second fusion label (hyperbolic space feature), the two are fused again through meta-feature clustering. Meta-feature clustering integrates the distribution characteristics of the labels themselves (such as the coincidence, difference, and hierarchical correlation of the labels), rather than directly clustering the original vectors, which can more efficiently capture the complementarity of the two types of labels: for documents with local similarity but belonging to different concept levels (such as “CNN image classification” and “Transformer image classification”), meta-feature clustering can retain their local association of “image classification” and distinguish their hierarchical differences in “deep learning model evolution”; for documents belonging to different local features but belonging to the same core topic (such as “reinforcement learning algorithm design” and “application of reinforcement learning in robot control”), meta-feature clustering can integrate their global association of “reinforcement learning” and avoid being misclassified due to local feature differences.

[0096] The final generated fusion label can simultaneously consider the local similarity and global hierarchy of the documents, making the clustering results corresponding to the fusion matrix set more consistent with the internal logic of scientific research information (such as topic association, method evolution, and application scenarios).

[0097] The preliminary retrieval information obtained by converting the vector text of the fusion matrix set can significantly reduce the interference of irrelevant documents while retaining the association relationships of key documents (such as similar document clusters and topic evolution context), making it easier for researchers to quickly locate core document clusters and understand the research context through the clustering structure, thereby improving the efficiency and depth of information acquisition.

[0098] In summary, the fusion clustering process effectively overcomes the limitations of single clustering method and single space embedding through the three-layer strategy of “multiple methods to enhance the robustness of a single space, cross-space to capture complementary features, and meta-feature fusion to improve accuracy”, ultimately achieving the unity of robustness, comprehensiveness, and accuracy in scientific literature clustering, and laying a core foundation for high-quality scientific information retrieval and tracking.

[0099] The vector text conversion processing of the fusion matrix set to obtain preliminary retrieval information is to convert the first matrix or the second matrix of the documents with the same fusion label information into text data to obtain the preliminary fusion retrieval information of the documents, and to construct the preliminary retrieval information using the preliminary fusion retrieval information of all documents.

[0100] The tracking retrieval processing of the preliminary retrieval information to obtain a set of retrieval tracking information includes:

[0101] S31, set a retrieval time interval;

[0102] S32, in each retrieval time interval, using a large model tool, the retrieval keyword information is semantically expanded to obtain first expanded keyword information;

[0103] S33, using a knowledge graph, the retrieval keyword information is structurally expanded to obtain second expanded keyword information;

[0104] S34, based on the scientific literature database, the first expanded keyword information and the second expanded keyword information are retrieved and processed to obtain a set of tracking literature information;

[0105] S35, the set of tracking literature information and the preliminary retrieval information are fused to obtain a set of retrieval tracking information in the retrieval time interval;

[0106] S36, for each retrieval time interval, S32 to S35 are executed to obtain a set of retrieval tracking information for each retrieval time interval.

[0107] The large model tool is used to expand the semantic of the retrieval keyword information to obtain the first expanded keyword information, comprising:

[0108] S321, using the context understanding module of the large model tool, generating semantic related word information (including synonyms, hypernyms, and application scenario words) of the retrieval keyword information.

[0109] S322, the retrieval keyword information and the related word information are respectively encoded into vector information by a word encoding method;

[0110] S323, all vector information is clustered to obtain class information containing the maximum number of vector information;

[0111] S324, the vector information of the class information is converted into a vector text to obtain the first expanded keyword information;

[0112] The large model tool is used to expand the semantic of the retrieval keyword information to obtain the first expanded keyword information, and the specific mode comprises:

[0113] Call open source LLM (such as Llama 3, Qwen), input prompt word to generate expanded word candidate set.

[0114] Calculate the semantic similarity (threshold ≥0.7) between the candidate word and the original keyword using Sentence-BERT, and filter irrelevant words.

[0115] The knowledge graph is used to expand the structure of the retrieval keyword information to obtain the second expanded keyword information, comprising:

[0116] S331, using a path traversal algorithm, traversing to obtain an entity set within N1-hop relationships from an entity corresponding to the search keyword information in the knowledge graph (such as "machine learning"→"deep learning"→"convolutional neural network");

[0117] S332, performing screening processing on the traversed entity set to obtain second expanded keyword information.

[0118] N1 is a positive integer greater than or equal to 2;

[0119] The path traversal algorithm can be Floyd algorithm or Dijkstra algorithm.

[0120] The screening processing on the traversed entity set to obtain the second expanded keyword information comprises:

[0121] Using a knowledge graph embedding model, converting the entity corresponding to the search keyword information in the knowledge graph into a standard entity vector, and converting the relationship information set of the entity corresponding to the search keyword information in the knowledge graph into a standard relationship matrix; each row vector of the standard relationship matrix is obtained by converting each relationship information in the relationship information set of the entity.

[0122] Using a knowledge graph embedding model, converting the traversed entity set to obtain a traversal entity matrix, and converting the relationship information set of each entity in the traversed entity set to obtain a corresponding traversal relationship matrix.

[0123] Screening processing the standard entity vector, the standard relationship matrix, the traversal entity matrix and the traversal relationship matrix to obtain a screened entity set.

[0124] Using the screened entity set and the search keyword information to construct the second expanded keyword information.

[0125] The conversion processing of the traversed entity set using the knowledge graph embedding model to obtain the traversal entity matrix is a conversion processing of each entity in the traversed entity set to obtain a corresponding row vector, and the traversal entity matrix is constructed using all row vectors.

[0126] The conversion processing of the relationship information set of each entity in the traversed entity set to obtain the corresponding traversal relationship matrix comprises:

[0127] The conversion processing of each relationship information in the relationship information set of each entity in the traversed entity set to obtain a corresponding row vector, and the corresponding traversal relationship matrix is constructed using all row vectors.

[0128] The screening processing on the standard entity vector, the standard relation matrix, the traversal entity matrix and the traversal relation matrix obtains a screened entity set, comprising:

[0129] The standard entity vector is subtracted from each row vector of the traversal entity matrix to obtain a corresponding difference vector;

[0130] A difference matrix C is constructed by using each difference vector as a row vector;

[0131] A multivariate variational modal decomposition transformation is performed on the row vectors of the difference matrix to obtain a corresponding transformation vector;

[0132] A transformation matrix F is constructed by using each transformation vector as a row vector;

[0133] Feature calculation is performed on the transformation matrix and the difference matrix to obtain a feature matrix T;

[0134] The expression of the feature calculation is:

[0135] T = C 1 / 2 FC -1 / 2 ,

[0136] Wherein, T is a feature matrix;

[0137] Rank value α, norm value β, core norm γ and singular value are calculated from the feature matrix T;

[0138] It is judged whether each singular value satisfies a first discriminant inequality to obtain a first discrimination result;

[0139] The expression of the first discriminant inequality is:

[0140]

[0141] Wherein, N is the row dimension of the feature matrix T;

[0142] A first screened entity set is constructed by using the entities corresponding to the row vectors of all singular values with the first discrimination result being yes;

[0143] The traversal relation matrix and the standard relation matrix corresponding to the first screened entity set are subjected to a deep screening processing to obtain a screened entity set.

[0144] The singular value is obtained by singular value decomposition on the feature matrix T, and one singular value corresponds to one row vector of the feature matrix T;

[0145] The multivariate variational modal decomposition transformation is an MVMD transformation;

[0146] The application realizes accurate screening of singular values in the feature matrix T through a first discriminant inequality. The formula combines the rank value, norm value, kernel norm and row dimension of the feature matrix, and constructs a quantitative discriminant standard through a trigonometric function and a proportional relationship, which can effectively distinguish the correlation strength of the entities corresponding to the original keywords in the semantic and structural aspects. The singular value reflects the importance of the entity features. The threshold value comprehensively considers the dimension characteristics and structural characteristics of the matrix, avoiding the over-screening or screening problem caused by a single threshold value. The entities corresponding to the singular values screened through the inequality can ensure semantic association with the original keywords while retaining entities with significant features, providing a high-quality candidate set for subsequent structure expansion, reducing the interference of irrelevant entities on retrieval tracking, and improving the accuracy of structure expansion.

[0147] The first screening entity set is subjected to a depth screening process to obtain a screened entity set, including:

[0148] The kernel norm difference of the traversal relationship matrix corresponding to each entity in the first screening entity set and the standard relationship matrix is calculated to obtain a first difference value C1 corresponding to each entity.

[0149] The fusion difference of the traversal relationship matrix corresponding to each entity in the first screening entity set and the standard relationship matrix is calculated to obtain a second difference value C2 corresponding to each entity.

[0150] The expression of the fusion difference calculation is:

[0151]

[0152] Wherein, N2 and M2 are the row dimension and column dimension of the standard relationship matrix, and are the i-th elements of the first difference vector and the second difference vector, respectively, A ij and B ij are the i-th row and j-th column elements of the standard relationship matrix and the traversal relationship matrix corresponding to the entity, respectively.

[0153] It is judged whether the product of the first difference value and the second difference value of each entity is less than a preset difference discriminant threshold value to obtain a second discriminant result.

[0154] The screened entity set is constructed using all entities with a second discriminant result of yes.

[0155] The difference discriminant threshold value can be 3.5.

[0156] The kernel norm difference calculation is a kernel norm calculation of the difference matrix of the two matrices.

[0157] The present application realizes the fine quantification of the difference between the standard relation matrix and the traversal relation matrix by fusing the difference calculation expression. The first difference vector and the second difference vector respectively extract the extreme eigenvalues in each row of the standard relation matrix and the traversal relation matrix, which can highlight the significant difference between the matrices. The normalized difference ratio is used to avoid the difference misjudgment caused by the different magnitudes of the matrix elements, and the log2 operation is combined to nonlinearly amplify the difference ratio, which can strengthen the discrimination degree of the significant difference and suppress the interference of the slight fluctuation. Through the cumulative summation of the difference of each row, the C2 value can comprehensively reflect the difference degree of the two matrices in the overall structure. The formula is combined with the first difference value C1 (kernel norm difference) to form a multi-dimensional difference evaluation system, which can not only capture the overall structural difference of the matrix, but also identify the significant changes of the local features, so as to ensure that the screened entities are highly matched with the original keywords in the relationship structure, further improving the relevance and reliability of the second expanded keyword information, and providing more accurate keyword support for tracking retrieval.

[0158] The row vector of the traversal entity matrix is obtained by converting each entity in the set of entities obtained by traversal;

[0159] Each row vector of the traversal relation matrix is obtained by converting each relation information of the set of relation information of the entity in the set of entities obtained by traversal;

[0160] The knowledge graph embedding model (KGE) includes TransE, RotatE, etc.

[0161] The scientific research literature database includes CNKI, Wanfang, VIP, IEEE, etc.

[0162] The fusion processing of the tracking literature information set and the preliminary retrieval information to obtain the retrieval tracking information set within the retrieval time interval includes:

[0163] The semantic alignment processing of the tracking literature information set and the preliminary retrieval information to obtain the first fusion text information;

[0164] The content fusion processing of the first fusion text information to obtain the retrieval tracking information set within the retrieval time interval.

[0165] The content fusion processing is to fuse the method information, conclusion information, background information, paper abstract information, etc. in the tracking literature information set and the preliminary retrieval information to obtain the retrieval tracking information set. The fusion method can use attention mechanism weighted fusion or GNN fusion method.

[0166] The calculation expression in the embodiment of the present application is dimensionless to the variable involved before calculation.

[0167] The second aspect of the embodiment of the present application discloses a scientific research information retrieval and tracking device, which comprises:

[0168] A memory storing executable program codes;

[0169] A processor coupled with the memory;

[0170] The processor calls the executable program codes stored in the memory to execute the scientific research information retrieval and tracking method.

[0171] The third aspect of the embodiment of the present application discloses a computer storage medium storing computer instructions, which are called by a computer to execute the scientific research information retrieval and tracking method.

[0172] The fourth aspect of the embodiment of the present application discloses an information data processing terminal for realizing the scientific research information retrieval and tracking method.

[0173] The above only describes the embodiments of the present application and is not used to limit the present application. The present application can be variously changed and modified by those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A method for retrieving and tracking scientific research information, characterized in that, include: S1, Obtain search keyword information; S2, perform retrieval processing on the search keyword information to obtain preliminary search information; S3, perform follow-up retrieval processing on the preliminary retrieval information to obtain a set of retrieval follow-up information, including: S31, Set the search time interval S32, within each retrieval time interval, the retrieval keyword information is semantically expanded using a large model tool to obtain first expanded keyword information, including: S321, using the contextual understanding module of the large model tool, generates word information that is semantically related to the retrieved keyword information; S322, the search keyword information and related word information are encoded into vector information using word encoding methods respectively; S323, perform clustering on all vector information to obtain the category information with the largest number of vector information; S324, Perform vector text conversion on all vector information of the category information to obtain the first extended keyword information; S33, using a knowledge graph, the structure of the retrieved keyword information is expanded to obtain second expanded keyword information, including: S331, using a path traversal algorithm, starting from the entity corresponding to the search keyword information in the knowledge graph, traverse to obtain the set of entities within N1 jump relationships; S332, the entity set obtained by the traversal is filtered to obtain the second extended keyword information, including: S3321, using a knowledge graph embedding model, the entities corresponding to the search keyword information in the knowledge graph are converted into standard entity vectors, and the set of relation information of the entities corresponding to the search keyword information in the knowledge graph is converted into a standard relation matrix; each row vector of the standard relation matrix is ​​obtained by converting each relation information of the entity's relation information set. S3322, Using a knowledge graph embedding model, the entity set obtained by traversal is transformed to obtain a traversal entity matrix, and the relation information set of each entity in the entity set obtained by traversal is transformed to obtain a corresponding traversal relation matrix. S3323, The standard entity vector, standard relation matrix, traversed entity matrix and traversed relation matrix are filtered to obtain a filtered entity set; S3324, using the filtered entity set and retrieved keyword information, construct the second extended keyword information; S34, Based on the scientific literature database, the first extended keyword information and the second extended keyword information are retrieved and processed to obtain a set of tracked literature information; S35, the tracking document information set and the preliminary search information are fused to obtain the search tracking information set within the search time interval; S36. Within each retrieval time interval, execute S32 to S35 to obtain the retrieval tracking information set for each retrieval time interval.

2. The method for retrieving and tracking scientific research information as described in claim 1, characterized in that, The process of retrieving the keyword information to obtain preliminary retrieval information includes: S21, Using a scientific research literature database, the search keyword information is processed to obtain a set of search literature information; S22, Perform a first mapping process on the retrieved document information set to obtain a first matrix set; S23, perform a second mapping process on the retrieved document information set to obtain a second matrix set; S24, perform fusion clustering on the first matrix set and the second matrix set to obtain a fusion matrix set; S25, perform vector text conversion processing on the fusion matrix set to obtain preliminary retrieval information.

3. The method for retrieving and tracking scientific research information as described in claim 1, characterized in that, The filtering process of the standard entity vector, standard relation matrix, traversed entity matrix, and traversed relation matrix yields a filtered entity set, including: For the standard entity vector, subtract it from each row vector of the traversed entity matrix to obtain the corresponding difference vector; By using each difference vector as a row vector, a difference matrix is ​​constructed. C ; Perform multivariate variational mode decomposition transformation on the row vectors of the difference matrix to obtain the corresponding transformation vectors; By using each transformation vector as a row vector, a transformation matrix can be constructed. F ; The transformation matrix and the difference matrix are subjected to feature calculation to obtain the feature matrix T; The expression for calculating the feature is: ， Where T is the characteristic matrix; The rank value is calculated from the characteristic matrix T. Norm value nuclear norm and singular values; Determine whether each singular value satisfies the first discrimination inequality to obtain the first discrimination result; The expression for the first discriminant inequality is: , Where N is the row dimension of the feature matrix T; The first set of filtered entities is constructed by using the entities of the row vectors corresponding to all singular values ​​whose first discrimination result is yes; The traversal relation matrix and standard relation matrix corresponding to the first set of filtered entities are subjected to deep filtering to obtain the set of filtered entities.

4. A scientific research information retrieval and tracking device, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the scientific research information retrieval and tracking method as described in any one of claims 1 to 3.

5. A computer-storable medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked by the computer, are used to execute the scientific research information retrieval and tracking method as described in any one of claims 1 to 3.

Images