Multi-view science and technology data portrait system and method thereof

Abstract

The invention discloses a multi-view science and technology data portrait system and method, a science and technology index integrated evaluation module constructs a science and technology index portrait, and realizes multi-dimensional science and technology competitiveness transverse / longitudinal comparative analysis; the science and technology hot spot label labeling module constructs science and technology hot spot portraits and displays science and technology hot spots in a period of time in the future in a label word cloud form; the science and technology atlas construction and analysis module constructs science and technology atlas portraits, determines potential relationships among science and technology entities, presents full life cycle evolution of the science and technology entities, and displays time-space association among the science and technology entities; the science and technology competition situation awareness module constructs a science and technology field portrait, monitors the latest key science and technology field trend, and reveals the competition situation of key field resources; and the science and technology portrait multi-dimensional display module is used for comprehensive display of science and technology index portraits, science and technology hotspot portraits, science and technology map portraits and science and technology field portraits. According to the method, visual science and technology general picture analysis and game research and judgment are realized, and the efficiency of science and technology big data analysis is improved.

Description

Technical Field

[0001] The present invention relates to the field of science and technology big data portraits, and in particular to a multi-perspective science and technology data portrait system and method. Background Art

[0002] Science and technology data portraits are regarded as an effective means to integrate and analyze multi-source science and technology information, present a comprehensive picture of science and technology strength and provide valuable insights. Through digital analysis of widely collected multi-source science and technology data, science and technology data portraits can achieve a comprehensive evaluation and comparison of science and technology strength, revealing the advantages and shortcomings of different fields. At the same time, using labeling technology to gain insight into the characteristics of science and technology strength helps to discover important laws and trends hidden in the data. At present, the existing public science and technology indicator system provides important data support and analysis perspectives for the construction of science and technology data portraits. In the basic idea of ​​building science and technology data portraits, it is necessary to review and integrate the existing public science and technology big data indicator system, comprehensively analyze and obtain relevant data, and build a science and technology digital portrait that meets actual needs through data labeling and selective screening. However, science and technology data are of various types, from a wide range of sources and in large amounts of data. How to effectively integrate big data processing, data mining and artificial intelligence technologies to establish science and technology data portraits has become an important challenge. In particular, how to achieve cross-domain and cross-disciplinary cross-analysis and understanding of science and technology data from different perspectives by organically combining domain knowledge, data science and technology, and decision analysis methods has become an important challenge. In summary, there is currently a lack of research on a multi-perspective science and technology data portrait method and its system in the relevant field, that is, to provide an integrated solution for multi-perspective science and technology data portrait modeling, analysis and application based on the establishment of full-picture data. Summary of the invention

[0003] The purpose of the present invention is to propose a multi-perspective science and technology data portrait system and method, so as to depict the overview of scientific and technological strength from multiple perspectives and provide fine-grained scientific and technological value insight support in the field.

[0004] The technical solution to achieve the purpose of the present invention is: a multi-perspective science and technology data portrait system, including a science and technology indicator integrated evaluation module, a science and technology hotspot labeling module, a science and technology map construction and analysis module, a science and technology competition and cooperation situation awareness module, and a science and technology portrait multi-dimensional display module, wherein:

[0005] The integrated evaluation module of science and technology indicators is used to build a portrait of science and technology indicators. A three-level science and technology indicator system is established from the dimensions of science and technology input, output and benefit. The relative importance of indicators at each level is determined through the "Delphi method", and the geometric analysis method is used to quantify and rank indicators, so as to achieve multi-dimensional horizontal / vertical comparative analysis of science and technology competitiveness.

[0006] The technology hotspot labeling module is used to build a technology hotspot portrait. The TextRank algorithm is used to extract technology keywords from a large amount of technology text data in parallel, and a technology theme word library is established through the similarity mapping between technology keywords and technology theme word vector encodings. By building a technology hotspot label prediction model and building a sample set training based on the technology theme word library, the technology hotspots in the future are displayed in the form of a label word cloud.

[0007] The technology graph construction and analysis module is used to construct technology graph portraits. Through knowledge graph modeling and technology entity relationship extraction, the connection between technology entities is established. Through technology graph association analysis, technology graph time series analysis, and technology graph spatiotemporal analysis, the potential relationship between technology entities is discovered, the evolution of the entire life cycle of technology entities is presented, and the spatiotemporal relationship between technology entities is displayed.

[0008] The technology competition and cooperation situation awareness module is used to build a portrait of the technology field. It extracts core viewpoints from think tank texts, monitors the latest trends in key technology fields, and calculates import dependence and trade stability index by building a trade competition and cooperation situation network, revealing the competition and cooperation situation of resources in key fields.

[0009] The multi-dimensional display module of science and technology portraits provides a unified user portal for the comprehensive display of science and technology indicator portraits, science and technology hotspot portraits, science and technology map portraits, and science and technology field portraits.

[0010] Furthermore, the science and technology indicator integrated evaluation module includes a science and technology indicator evaluation system construction module, a multi-dimensional science and technology competitiveness index ranking module and a science and technology indicator portrait construction module, wherein:

[0011] The construction module of the science and technology indicator evaluation system is composed of the science and technology development environment index , Science and Technology Development Resource Index , Science and Technology Innovation Output Index , Science and Technology and Industrial Security Index , Comprehensive Science and Technology Contribution Index There are 5 first-level indicators in total; the first-level indicators can be decomposed into multiple second-level indicators, expressed as ; The second-level indicators are composed of multiple directly quantifiable third-level indicators, expressed as ; The indicator system is quantified by formula (1a); (1a); st (1b); Among them, i, j, and k represent the serial numbers of the first-level indicator, the second-level indicator, and the third-level indicator respectively. and Indicates the third level indicator The maximum and minimum values ​​of ; and Indicates the corresponding three-level indicators and secondary indicators The weight coefficient of Indicates the first level indicator Secondary indicators included The number of Represents secondary index The three-level indicators included the number of

[0012] The multi-dimensional science and technology competitiveness index ranking module realizes the science and technology competitiveness index ranking from five indicator dimensions: science and technology development environment index, science and technology development resource index, science and technology innovation output index, science and technology and industrial security index, and science and technology comprehensive contribution index;

[0013] The science and technology indicator portrait construction module uses charts and numerical analysis methods to conduct a horizontal comparison of the first-level evaluation indicator rankings of different evaluation objects in the same time period from five indicator dimensions, including the science and technology development environment index, science and technology development resource index, science and technology innovation output index, science and technology and industrial security index, and science and technology comprehensive contribution index. It also conducts horizontal / vertical comparisons of different indicator dimensions of different evaluation objects in different time periods by "drilling down" from the second-level and third-level indicators through the first-level indicators.

[0014] Furthermore, the secondary index weight coefficient and weight coefficient of the third-level indicators The solution process is: First, the Delphi method is used to determine any two indicators at the corresponding level. and The relative importance of , Respectively Compared to Equally important, slightly important, important, relatively important, very important; then construct a comparison matrix of the relative importance of indicators , the geometric analysis method of formula (2) is used to calculate the index weight; (2); in, Indicates any indicator at each level The weight coefficient of , D represents the indicator dimension of the corresponding level.

[0015] Furthermore, the science and technology indicator portrait construction module displays the ranking change trend of different indicator values ​​of the same evaluation object over time through a line graph, displays the ranking comparison of different evaluation objects in the same indicator dimension through a radar chart, and displays the changes of the evaluation object in different indicator dimensions in recent time periods through a heat map.

[0016] Furthermore, the technology hotspot label marking module includes a technology theme word library construction module, a technology hotspot label prediction module and a technology hotspot word cloud marking module, wherein:

[0017] The technology thesaurus construction module first uses papers, patents, think tanks, and reports as data sources for technology texts, and constructs a technology text keyword relationship graph G in parallel. The technology text keyword relationship graph G uses technology keywords as nodes and the co-occurrence relationship of technology keywords in the text as edges. The co-occurrence relationship is only valid if the co-occurrence frequency of two technology keywords in the same text is greater than a threshold. A relationship edge is established when the word vector is generated; then the TextRank algorithm is used to rank the technology keywords in the technology text keyword relationship graph G to generate a (technology keyword, score) set; then, Word2Vec is used to encode the technology keywords in the technology text keyword relationship graph G, calculate the similarity between the technology keywords and the technology subject words, and replace the technology keywords with the technology subject words with the greatest similarity to achieve standardization; finally, the scores of different technology keywords with the same technology subject words are accumulated and aggregated, and the subject words with the top K scores are taken as technology hotspot tags to form a technology subject word library; The technology hotspot label prediction module is used to predict the labels that may become technology hotspots of the evaluated object in the future. The technology hotspot label prediction model consists of a three-layer LSTM network cascade, and a 32-node fully connected layer with a SoftMax activation function is added at the end. The input is the embedding coding sequence of the technology theme words in the previous m stages, and the output is the probability of the technology theme words in the next stage. The technology theme word library constructed by the technology texts of papers, patents, think tanks, and reports is used to construct a training sample set {([embedded coding sequence], technology theme words)}, and the network structure of the technology hotspot label prediction model is fine-tuned in an incremental training manner. The discount coefficient Optimize sample training loss to improve model performance, where is the serial number of the time period to which the sample belongs, and Q is the number of time period divisions;

[0018] The technology hot word cloud annotation module adjusts the size and color of the label according to the technology hot label probability value output by the technology hot label prediction model, forming a cloud-like image with technology keywords as visual elements, showing the technology hot spots of the evaluation object in the future time period.

[0019] Furthermore, the technology graph construction and analysis module includes a technology graph construction module and a technology knowledge association mining module, wherein:

[0020] The technology graph construction module first integrates three types of unstructured technology text data: technology projects, technology bulletins, and technology funds. It uses the BERT pre-training model to encode the input technology text, extract the semantic information of the text, and generate a semantic feature vector sequence for each word. , P is the number of words in the text; then, the output of the BERT pre-trained model is further processed using a bidirectional LSTM network , capturing the contextual dependencies between words and obtaining an enhanced semantic feature vector sequence , and the output of the bidirectional LSTM network As the input of the conditional random field, the boundaries and types of scientific and technological entities in the text are identified, and scientific and technological entities and attribute information are screened out according to the type, where scientific and technological entities include scientific and technological organizations, scientific and technological funds, scientific and technological projects, and scientific and technological talents; then, the scientific and technological entity relationships are extracted. The scientific and technological entity relationships are determined by the co-occurrence relationship of scientific and technological entities in the text and the similarity of the semantic feature vectors between scientific and technological entities. That is, if the frequency of co-occurrence of two scientific and technological entities in the same context exceeds the threshold And the similarity of semantic feature vectors exceeds the threshold , then there is a relationship edge between the two scientific and technological entities; finally, the identified scientific and technological entities and scientific and technological entity relationships are stored in a graph database or a relational database to build a complete scientific and technological graph;

[0021] The science and technology knowledge association mining module discovers the potential relationship between science and technology entities and events from the science and technology graph through three methods: science and technology graph association analysis, science and technology graph time series analysis, and science and technology graph time-space analysis. Among them, the science and technology graph association analysis queries the multi-hop relationship of the science and technology graph to form a subgraph of the object of interest, and obtains the hidden association relationship between science and technology entities by checking other entities directly and indirectly related to the science and technology entity; the science and technology graph time series analysis dynamically presents the existence time of the science and technology entity and the start and end time of the science and technology entity relationship, and plays back the science and technology graph state at any time within the start and end time range, presenting the evolution process of the entire life cycle of the science and technology entity; the science and technology graph time-space analysis displays the time-space association relationship between science and technology entities by associating the science and technology graph with the geographical location information of the science and technology entity and displaying it on the three-dimensional earth.

[0022] Furthermore, the technology competition and cooperation situation awareness module includes a key field trend monitoring module, a key field resource competition and cooperation situation analysis module and a key field portrait construction module, wherein:

[0023] The key science and technology field trend monitoring module collects think tank reports in key fields to form a think tank text library. It uses an open source large language model to automatically extract the content framework, summary and core viewpoints of think tank express reports, and displays the regional distribution of key fields in the form of a map. The latest core viewpoints of think tank express reports in key fields are displayed when the mouse is hovered.

[0024] The module for analyzing resource competition and cooperation in key science and technology fields builds a trade competition and cooperation network based on resource reserves and production and trade exchanges. In the trade competition and cooperation network, the import dependence is calculated to show the degree of dependence on trade imports. The stability of trade is measured by calculating the trade stability index, i.e., the network structure entropy, to analyze the trade competition and cooperation situation of the industrial chain and supply chain.

[0025] The module for constructing portraits of key scientific and technological fields conducts a network analysis of the competition and cooperation in the trade of important mineral resources involved in the integrated circuit and large-capacity battery industries, forms a resource portrait based on the output and reserve data of mineral resource demand, and displays the distribution of mineral resources in order to analyze the potential competition and cooperation between key minerals.

[0026] A multi-perspective technology data portrait method is based on the multi-perspective technology data portrait system to realize multi-perspective technology data portrait, specifically:

[0027] The science and technology indicator integrated evaluation module is used to build a science and technology indicator portrait, and a three-level science and technology indicator system is established from the dimensions of science and technology input, output and benefit. The relative importance of indicators at each level is determined through the "Delphi method", and the geometric analysis method is used to quantify and rank the indicators, so as to achieve multi-dimensional horizontal / vertical comparative analysis of science and technology competitiveness;

[0028] Use the technology hotspot labeling module to build a technology hotspot portrait, extract technology keywords from technology text data in parallel through the TextRank algorithm, establish a technology theme word library through the similarity mapping between technology keywords and technology theme word vector encoding, and build a technology hotspot label prediction model and build a sample set training based on the technology theme word library to display the technology hotspots in the future in the form of a label word cloud;

[0029] Use the technology graph construction analysis module to build a technology graph portrait. Through knowledge graph modeling and technology entity relationship extraction, establish the connection between technology entities. Through technology graph association analysis, technology graph time series analysis, and technology graph spatiotemporal analysis, determine the potential relationship between technology entities, present the evolution of the entire life cycle of technology entities, and show the spatiotemporal relationship between technology entities.

[0030] Use the technology competition and cooperation situation awareness module to build a portrait of the technology field, extract the core content by analyzing think tank texts, monitor the latest trends in key technology fields, and calculate the import dependence and trade stability index by building a trade competition and cooperation situation network to reveal the competition and cooperation situation of resources in key fields;

[0031] The multi-dimensional display module of science and technology portraits is used to unify the user portal for the comprehensive display of science and technology indicator portraits, science and technology hotspot portraits, science and technology map portraits, and science and technology field portraits.

[0032] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the multi-perspective scientific and technological data portrait method is implemented to realize multi-perspective scientific and technological data portrait.

[0033] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the multi-perspective scientific and technological data portrait method to achieve multi-perspective scientific and technological data portrait.

[0034] Compared with the prior art, the present invention has the following significant advantages: 1) It proposes a multi-perspective technology data portrait method, which expands the breadth and depth of technology development trend analysis. 2) It provides a technology data portrait value insight system, which realizes perspective technology overview analysis and game judgment from the application perspective, and improves the efficiency of technology big data analysis. 3) Based on the above effects, the present invention can realize the rapid integration and portrait presentation of multi-modal technology big data in engineering applications, and has strong operability and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a structural diagram of a multi-perspective scientific and technological data portrait system of the present invention.

[0036] Figure 2 It is the structural diagram of the science and technology indicator evaluation system.

[0037] Figure 3 It is a quantitative flow chart of scientific and technological indicators.

[0038] Figure 4 It is a flow chart for constructing a science and technology thesaurus.

[0039] Figure 5 This is the architecture diagram of the technology hotspot prediction model.

[0040] Figure 6 It is a structural diagram of the analysis module for constructing the science and technology map.

[0041] Figure 7 It is the flow chart of the technology competition and cooperation situation awareness module. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0043] This invention realizes the construction and quantitative analysis of science and technology data portraits and provides valuable insights by integrating science and technology data source integration technology, science and technology hotspot annotation technology, science and technology graph analysis technology, and science and technology competition and cooperation situation awareness technology. The overall architecture is as follows: Figure 1 shown.

[0044] A multi-perspective science and technology data portrait system, including: science and technology index integrated evaluation module, science and technology hotspot labeling module, science and technology map construction and analysis module, science and technology competition and cooperation situation awareness module, science and technology portrait multi-dimensional display module. Figure 2-7 The composition and function of each module are described in detail.

[0045] The technology indicator integrated evaluation module is used to construct a technology indicator portrait. A three-level technology indicator system is established from the dimensions of technology input, output and benefit. The relative importance of indicators is determined through the "Delphi method", and the geometric analysis method is used to quantify the ranking of indicators at each level, so as to achieve horizontal / vertical comparative analysis of technology competitiveness in different indicator dimensions. Figure 2-3 , the specific implementation steps are as follows:

[0046] (1) Construction of science and technology indicator evaluation system

[0047] Reference Figure 2 The three-level science and technology indicator system consists of the science and technology development environment index , Science and Technology Development Resource Index , Science and Technology Innovation Output Index , Science and Technology and Industrial Security Index , Comprehensive Science and Technology Contribution Index There are 5 first-level indicators in total; the first-level indicators can be decomposed into multiple second-level indicators, expressed as ; The second-level indicators are composed of multiple directly quantifiable third-level indicators, expressed as , as shown in Table 1: Table 1 Three-level science and technology indicator system

[0048] The science and technology indicator system adopts the "Delphi method" to determine the relative importance of indicators, constructs an indicator relative importance comparison matrix, determines the weight coefficients of indicators at each level through geometric analysis, and calculates the indicator values ​​at each level through reverse weighting of third-level indicators → second-level indicators → first-level indicators, as shown in formula (1a).

[0049] (1a)

[0050] st

[0051] (1b)

[0052] Among them, i, j, and k represent the serial numbers of the first-level indicator, the second-level indicator, and the third-level indicator respectively. and Indicates the third level indicator The maximum and minimum values ​​of ; and Indicates the corresponding three-level indicators and secondary indicators The weight coefficient of Indicates the first level indicator Secondary indicators included The number of Represents secondary index The three-level indicators included The number of

[0053] The secondary index weight coefficient and weight coefficient of the third-level indicators The solution process is to first use the "Delphi method" to determine any two indicators at each level. and The relative importance of , Respectively indicate indicators Compared to Equally important, slightly important, important, relatively important, very important; then construct a comparison matrix of the relative importance of indicators , the geometric analysis method is used to calculate formula (2).

[0054] (2)

[0055] in, Indicates any indicator at each level The weight coefficient of , D represents the indicator dimension of the corresponding level.

[0056] (2) Multi-dimensional science and technology competitiveness index ranking

[0057] The science and technology competitiveness index refers to the comprehensive strength of the evaluation object in the field of science and technology, reflecting its level of scientific and technological innovation ability, scientific and technological achievement transformation ability, scientific and technological resource allocation efficiency and scientific and technological talent team quality. This system realizes the ranking of science and technology competitiveness index from five indicator dimensions: science and technology development environment index, science and technology development resource index, science and technology innovation output index, science and technology and industrial security index, and science and technology comprehensive contribution index.

[0058] Reference Figure 3 The calculation and ranking of the science and technology competitiveness index includes the following steps:

[0059] ①Set the set of objects to be evaluated , set the starting year for the evaluation and end year ; Initialize variables , , , is 0;

[0060] ② , take an evaluation object from C and start calculating the indicator value;

[0061] ③ Determine whether all evaluation objects in C have been traversed. If yes, go to step ⑩, otherwise go to step ④;

[0062] ④ Determine whether the current evaluation object indicator has been calculated. If yes, go to step ③; otherwise, go to step ⑤;

[0063] ⑤Judgment , Is it 0? If yes, go to step ⑥, otherwise go to step ⑦;

[0064] ⑥ Obtain the relative importance of the expert scoring evaluation indicators and use the geometric analysis method of formula (2) to determine the weights of the secondary and tertiary indicators , ;

[0065] ⑦Judgment , Is it 0? If yes, go to step ⑧, otherwise go to step ⑨;

[0066] ⑧Traverse the set C and calculate , ;

[0067] ⑨Use formula (1a) to calculate the value of the first-level indicator currently being evaluated;

[0068] ⑩ For all evaluated objects, according to the first-level indicators Ranking is performed to obtain the ranking of the evaluated object in the five dimensions of science and technology development environment, science and technology development resources, science and technology innovation output, science and technology and industrial security, and comprehensive science and technology contribution in time period t;

[0069] ⑪Judgment , if yes, go to step ⑬, otherwise go to step ⑫;

[0070] ⑫ , go to step ③;

[0071] ⑬End the evaluation.

[0072] (3) Construction of science and technology indicator portrait

[0073] The construction of the science and technology indicator portrait is based on the three-level science and technology indicator system, from the five indicator dimensions of science and technology development environment index, science and technology development resource index, science and technology innovation output index, science and technology and industrial security index, and science and technology comprehensive contribution index, to achieve horizontal / vertical comparative analysis of science and technology development status and trends in the form of charts combined with numerical analysis.

[0074] Specifically, for the horizontal comparison of the status and trend of scientific and technological development, a horizontal comparative analysis is conducted on the rankings of the scientific and technological development environment index, scientific and technological development resource index, scientific and technological innovation output index, scientific and technological and industrial security index, and comprehensive scientific and technological contribution index of different evaluation objects (i.e. the subjects of scientific and technological evaluation) in the same time period;

[0075] For the longitudinal comparison of the status and trends of scientific and technological development, a longitudinal comparative analysis is conducted on the tertiary, secondary, and primary indicators and rankings of specific assessment objects in different time periods. The ranking growth rate of different indicators in different time periods is calculated to reveal the scientific and technological development status of the assessment objects in different periods.

[0076] Furthermore, a line chart is used to show the ranking trend of different indicator values ​​of the same evaluation object over time, a radar chart is used to show the ranking comparison of different evaluation objects in the same indicator dimension, and a heat map is used to show the changes in indicator dimensions such as scientific and technological development environment, scientific and technological development resources, and scientific and technological innovation output of the evaluation object in recent time periods.

[0077] Furthermore, by “drilling down” the first-level indicators to the second-level and third-level indicators step by step, horizontal / vertical comparisons of different indicator dimensions of different evaluation objects in different time periods are carried out to reveal the advantages and disadvantages of different indicator dimensions of different evaluation objects in scientific and technological development.

[0078] The technology hotspot labeling module is used to construct a technology hotspot portrait by extracting technology keywords from massive technology text data to establish a technology theme word library. Furthermore, by constructing a technology hotspot label prediction model, accurate prediction and trend analysis of the direction of technology development can be achieved in the form of a label word cloud. Figure 4-5 , the specific implementation steps are as follows:

[0079] (1) Construction of science and technology thesaurus

[0080] Reference Figure 4 The construction of the science and technology thesaurus is based on the massive science and technology text data of papers, patents, think tanks, and reports. By analyzing the science and technology keywords from a large amount of science and technology text data and scoring the science and technology keywords, and mapping them to the science and technology thesaurus for standardization, the top K science and technology thesaurus with cumulative scores are taken as the hot science and technology tags within the evaluation period. The specific implementation steps are as follows:

[0081] 1) Collect scientific and technological text data within the evaluation period of the evaluated object as the data source for scientific and technological hot spot analysis to form a text collection , where N is the number of texts. Perform word segmentation and stop word removal in parallel to form ;

[0082] 2) Randomly divided into M batches, forming The size of the 1st to M-1st batch is , the size of the Mth batch is ;

[0083] 3) For any , using technology keywords as nodes , construct a subgraph based on their co-occurrence relationship in the text as edge R The co-occurrence relationship is only established when the co-occurrence frequency of two technology keywords in the same text is greater than the threshold. Merge different subgraphs through the same node Get the technology text keyword relationship graph G, use the TextRank algorithm to rank the technology keywords in the technology text keyword relationship graph G, and form a (technology keyword, score) set ,in Is TextRank for technology keywords The score value of , E is the number of technology keywords;

[0084] 4) For any technology keyword , use Word2Vec for word embedding, and calculate the cosine similarity between the scientific keyword and all the scientific keyword vector encodings in the scientific keyword table, take the scientific keyword with the largest similarity to replace the scientific keyword to achieve standardization, accumulate the scores of different scientific keywords for the same scientific keyword as the score of the scientific keyword, and form a (scientific keyword, score) set , where H is the number of scientific and technological keywords. In particular, the scientific and technological keyword table is a complete set of scientific and technological keyword sets constructed by domain experts and is available as an open source.

[0085] 5) Sort the science and technology keywords from large to small according to the score value, take the top K science and technology keywords as the science and technology hotspot labels of the evaluated object in the time period, and form a science and technology keyword library by calculating the science and technology hotspot labels in all historical time periods.

[0086] (2) Prediction of technology hotspot tags

[0087] Reference Figure 5 The technology hotspot label prediction model consists of three layers of LSTM networks cascaded together, and a 32-node fully connected layer with a SoftMax activation function is added at the end to predict the probability of technology hotspot labels (technology keywords) in the future time period. The technology hotspot label prediction model uses a technology keyword library constructed from papers, patents, think tanks and report text data to generate training samples. , fine-tune the neural network structure parameters by incremental training, and optimize the sample training loss through the discount coefficient to improve the model performance. The specific implementation steps are as follows:

[0088] 1) Organize the embedded codes of historical science and technology keywords into long sequences according to the chronological order of time periods;

[0089] 2) Use a sliding window of length m to divide the long sequence, and each time the sliding window moves back one time period. The embedded coding of the scientific and technological subject words in the m time periods within the sliding window constitutes an embedded coding sequence, and the scientific and technological subject words in the next time period immediately after the sliding window are used as the labels of the subsequence, thus generating a number of A training sample set, where Q represents the number of time periods divided;

[0090] 3) The technology hotspot label prediction model is first trained using a historical sample set. As time goes by, new samples are continuously generated, and the model continuously fine-tunes the neural network structure parameters through incremental training. At the same time, a discount coefficient is used during the training process. The sample training loss is scaled to improve the role of new samples in model parameter adjustment, where is the serial number of the time period to which the sample belongs, and Q is the number of time period divisions;

[0091] 4) Use the technology keyword embedding coding sequence of the last time period to input the trained technology hotspot label prediction model to predict the probability of technology hotspot labels (technology keyword) in the future time period, and select the K labels with the highest probability values ​​as the technology hotspots of the evaluation object in the future time period.

[0092] (3) Technology hot word cloud annotation

[0093] The technology word cloud uses the size and color of the technology hotspot label probability value to show the hot spots in the technology field of the evaluation object in the future time period. Specifically, according to the size of the technology hotspot label probability value output by the technology hotspot label prediction model and the adjustment of the label color, a cloud-shaped image with technology keywords as visual elements is formed.

[0094] The technology graph construction and analysis module is used to construct a technology graph portrait. Through knowledge graph modeling and technology entity relationship extraction, potential connections between technology entities are established. Through technology graph association analysis, technology graph time series analysis, and technology graph spatiotemporal analysis, hidden associations between technology entities are discovered, the evolution of the entire life cycle of technology entities is presented, and the spatiotemporal associations between technology entities are displayed. Figure 6 , the specific implementation steps are as follows:

[0095] (1) Construction of science and technology map

[0096] 1) Technology data integration. The sources of technology data include three types of unstructured technology text data: technology projects, technology bulletins, and technology funds. The technology data integration first cleans up errors and false information from the Internet; then removes punctuation marks, special characters, and meaningless stop words; and finally performs word segmentation, stem extraction, and part-of-speech tagging.

[0097] 2) Technology entity identification. The technology entity is the representation of a thing in the technology map, including the name, alias, description information, type, duration (start and end time), geographic location and other attribute information of technology organizations, technology funds, technology projects, and technology talents. The implementation steps of technology entity identification are as follows:

[0098] a) Use the BERT pre-trained model to encode the input scientific text, extract the semantic information of the text, and obtain the semantic feature vector sequence of each word ,in is the semantic feature vector of word j, and P is the number of words.

[0099] b) Optionally, further process the output of BERT using a bidirectional LSTM network (BiLSTM) , capturing the contextual dependencies between words and outputting an enhanced sequence of semantic feature vectors .

[0100] c) The output of the BiLSTM network As the input of the conditional random field, the boundaries and types of scientific and technological entities in the text are identified. Further screening out scientific and technological entities and attribute information of scientific and technological organizations, scientific and technological funds, scientific and technological projects, and scientific and technological talents.

[0101] 3) Scientific and technological relationship extraction. The scientific and technological relationship is the representation of the connection between scientific and technological entities in the scientific and technological graph, including the start and end points of the relationship between scientific and technological entities and the start and end time of the relationship. The start and end time of the relationship between scientific and technological entities is represented by the intersection of the start and end scientific and technological entity existence times. The scientific and technological entity relationship extraction is based on the extraction of scientific and technological entity information, and is determined by the co-occurrence relationship of scientific and technological entities in scientific and technological texts and the similarity (cosine similarity) of the semantic feature vectors between scientific and technological entities. That is, if the frequency of co-occurrence of two scientific and technological entities in the same context exceeds a threshold, And the similarity of semantic feature vectors exceeds the threshold , then there is a relationship edge between the two technology entities.

[0102] 4) Technology graph generation. Use graph databases (Neo4j, ArangoDB) or relational databases (MySQL, PostgreSQL) to store identified technology entities and relationships. Use knowledge graph reasoning technology to reason between technology entities and relationships to further enrich the knowledge graph. Use visualization tools (Gephi, Graphviz) to display the constructed knowledge graph.

[0103] (2) Mining the association of scientific and technological knowledge

[0104] The scientific and technological knowledge association mining refers to the process of discovering the association between new scientific and technological entities and events from the scientific and technological map, thereby revealing the knowledge structure, development trends and innovation opportunities in the field of science and technology. Figure 6 The scientific and technological knowledge association mining includes three methods: scientific and technological graph association analysis, scientific and technological graph time series analysis, and scientific and technological graph time-space analysis:

[0105] 1) Association analysis of technology graphs. By querying the multi-hop relationships of the technology graph, a subgraph of the object of interest is formed. By checking other entities directly or indirectly associated with the technology entity, the hidden associations between technology entities can be discovered;

[0106] 2) Technology graph time series analysis. Dynamically present technology entities and their relationships through the duration of technology entities in the technology graph and the start and end time of technology entity relationships. Support playback of technology graph status at any time within the start and end time range, and present the evolution of the entire life cycle of technology entities.

[0107] 3) Spatiotemporal analysis of technology maps. By displaying the technology maps on the three-dimensional earth according to the geographical location information of technology entities, the spatiotemporal relationship between technology entities is displayed, thus improving the breadth and depth of technology map analysis.

[0108] The technology competition and cooperation situation awareness module is used to build a portrait of the technology field. By analyzing think tank texts to extract core viewpoints, the latest trends in key technology fields are monitored. By building a trade competition and cooperation situation network including core indicators of trade objects and network core indicator definitions and calculations, the competition and cooperation situation of resources in key fields is revealed. Figure 7 , the specific implementation steps are as follows:

[0109] (1) Monitoring trends in key science and technology fields

[0110] By collecting think tank reports in key areas, a think tank text library is formed. The open source large language model is used to automatically extract the content framework, summary and core viewpoints of think tank bulletins. The regional distribution of key areas is displayed in the form of a map, and the latest core viewpoints of think tank bulletins in key areas are displayed when the mouse is hovered.

[0111] (2) Analysis of competition and cooperation in key scientific and technological fields

[0112] The competitive and cooperative situation in key scientific and technological fields refers to the current status of trade competition and cooperation in the industrial chain and supply chain in specific fields.

[0113] By building a trade competition and cooperation network based on resource reserves and production and trade relationship data, the following key core indicators are displayed in the trade competition and cooperation network:

[0114] Core indicators of trade partners: Import dependence = import volume / (internal production + import volume),

[0115] Network Core Indicators: Trade Stability Index

[0116] (3)

[0117] in is the proportion of imports from trading partner j to the total imports of all trading partners, and L is the number of trading partners. Used to measure the complexity and stability of the entire trade network.

[0118] (3) Construction of key area portraits

[0119] The domain portrait is constructed by analyzing the important mineral resources involved in the integrated circuit and large-capacity battery industries in key emerging industries, and a trade competition and cooperation situation network is constructed. A resource portrait is formed based on the output, reserve data, core indicators of trade objects, and network indicators of the demand for these mineral resources. The distribution and trade situation of mineral resources are displayed to analyze the potential competition and cooperation relationship of key minerals.

[0120] The multi-dimensional display module of scientific and technological portraits provides a unified user portal for the comprehensive display of four portraits: scientific and technological indicator portrait, scientific and technological hot spot portrait, scientific and technological map portrait, and scientific and technological field portrait.

[0121] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0122] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A multi-perspective scientific and technological data profiling system, It is characterized in that It includes the technology indicator integrated evaluation module, technology hotspot labeling module, technology map construction and analysis module, technology competition and cooperation situation awareness module, and technology portrait multi-dimensional display module, among which: The technology indicator integrated evaluation module is used to construct a technology indicator portrait, establish a three-level technology indicator system from the dimensions of technology input, output and benefit, determine the relative importance of indicators at each level through the "Delphi method", and use the geometric analysis method to quantify and rank indicators, so as to achieve multi-dimensional horizontal / vertical comparative analysis of technology competitiveness; The technology hotspot labeling module is used to construct a technology hotspot portrait, extract technology keywords from technology text data in parallel through the TextRank algorithm, establish a technology theme word library through the similarity mapping of technology keywords and technology theme word vector encoding, and construct a technology hotspot label prediction model and build a sample set training based on the technology theme word library to display the technology hotspots in the future in the form of a label word cloud; The technology graph construction and analysis module is used to construct a technology graph portrait, establish connections between technology entities through knowledge graph modeling and technology entity relationship extraction, determine the potential relationship between technology entities through technology graph association analysis, technology graph time series analysis, and technology graph spatiotemporal analysis, present the evolution of the entire life cycle of technology entities, and display the spatiotemporal relationship between technology entities; The technology competition and cooperation situation awareness module is used to build a portrait of the technology field, extract core viewpoints by analyzing think tank texts, monitor the latest trends in key technology fields, and calculate import dependence and trade stability index by building a trade competition and cooperation situation network to reveal the competition and cooperation situation of resources in key fields; The multi-dimensional display module of science and technology portraits provides a unified user portal for the comprehensive display of science and technology indicator portraits, science and technology hotspot portraits, science and technology map portraits, and science and technology field portraits.

2. According to the multi-perspective scientific and technological data portrait system of claim 1, It is characterized in that The science and technology indicator integrated evaluation module includes a science and technology indicator evaluation system construction module, a multi-dimensional science and technology competitiveness index ranking module and a science and technology indicator portrait construction module, among which: The construction module of the science and technology indicator evaluation system is composed of the science and technology development environment index , Science and Technology Development Resource Index , Science and Technology Innovation Output Index , Science and Technology and Industrial Security Index , Comprehensive Science and Technology Contribution Index There are 5 first-level indicators in total; the first-level indicators can be decomposed into multiple second-level indicators, expressed as ; The second-level indicators are composed of multiple directly quantifiable third-level indicators, expressed as ; The indicator system is quantified by formula (1a); （1a）； st （1b）； Among them, i, j, and k represent the serial numbers of the first-level indicator, the second-level indicator, and the third-level indicator respectively; and Indicates the third level indicator The maximum and minimum values ​​of ; and Indicates the corresponding three-level indicators and secondary indicators The weight coefficient of Indicates the first level indicator Secondary indicators included The number of Represents secondary index The three-level indicators included the number of The multi-dimensional science and technology competitiveness index ranking module realizes the science and technology competitiveness index ranking from five indicator dimensions: science and technology development environment index, science and technology development resource index, science and technology innovation output index, science and technology and industrial security index, and science and technology comprehensive contribution index; The science and technology indicator portrait construction module uses charts and numerical analysis methods to conduct a horizontal comparison of the rankings of first-level evaluation indicators of different evaluation objects in the same time period from five indicator dimensions, namely, science and technology development environment index, science and technology development resource index, science and technology innovation output index, science and technology and industrial security index, and science and technology comprehensive contribution index. It also conducts horizontal / vertical comparison of different indicator dimensions of different evaluation objects in different time periods by "drilling down" from the first-level indicators to the second-level and third-level indicators.

3. According to the multi-perspective scientific and technological data portrait system of claim 2, It is characterized in that Secondary index weight coefficient and weight coefficient of the third-level indicators The solution process is: First, the "Delphi method" is used to determine any two indicators at the corresponding level. and The relative importance of , Respectively Compared to Equally important, slightly important, important, relatively important, very important; then construct a comparison matrix of the relative importance of indicators , the geometric analysis method of formula (2) is used to calculate the index weight; （2）； in, Indicates any indicator at each level The weight coefficient of , D represents the indicator dimension of the corresponding level.

4. According to the multi-perspective scientific and technological data portrait system of claim 2, It is characterized in that The science and technology indicator portrait construction module uses a line graph to display the ranking change trend of different indicator values ​​of the same evaluation object over time, uses a radar chart to display the ranking comparison of different evaluation objects in the same indicator dimension, and uses a heat map to display the changes in different indicator dimensions of the evaluation object in recent time periods.

5. According to the multi-perspective scientific and technological data portrait system of claim 1, It is characterized in that The technology hotspot labeling module includes a technology theme word library construction module, a technology hotspot label prediction module and a technology hotspot word cloud labeling module, wherein: The technology thesaurus construction module first uses papers, patents, think tanks, and reports as data sources for technology texts, and constructs a technology text keyword relationship graph G in parallel. The technology text keyword relationship graph G uses technology keywords as nodes and the co-occurrence relationship of technology keywords in the text as edges. The co-occurrence relationship is only valid if the co-occurrence frequency of two technology keywords in the same text is greater than a threshold. A relationship edge is established when the word vector is generated; then the TextRank algorithm is used to rank the technology keywords in the technology text keyword relationship graph G to generate a (technology keyword, score) set; then, Word2Vec is used to encode the technology keywords in the technology text keyword relationship graph G, calculate the similarity between the technology keywords and the technology theme words, and replace the technology keywords with the technology theme words with the greatest similarity to achieve standardization; finally, the scores of different technology keywords with the same technology theme words are accumulated and aggregated, and the technology theme words with the top K scores are taken as technology hotspot tags to form a technology theme word library; The technology hotspot label prediction module is used to predict the labels that may become technology hotspots of the evaluated object in the future. The technology hotspot label prediction model consists of a three-layer LSTM network cascade, and a 32-node fully connected layer with a SoftMax activation function is added at the end. The input is the embedded coding sequence of the technology theme words in the previous m stages, and the output is the probability of the technology theme words in the next stage; the technology theme word library constructed by the technology texts of papers, patents, think tanks, and reports is used to construct the training sample set , fine-tuning the network structure of the technology hotspot label prediction model through incremental training, Optimize sample training loss to improve model performance, where is the serial number of the time period to which the sample belongs, and Q is the number of time period divisions; The technology hot word cloud annotation module adjusts the size and color of the label according to the technology hot label probability value output by the technology hot label prediction model, forming a cloud-like image with technology keywords as visual elements, showing the technology hot spots of the evaluation object in the future time period.

6. The multi-perspective scientific and technological data profiling system according to claim 1, It is characterized in that The technology map construction and analysis module includes a technology map construction module and a technology knowledge association mining module, wherein: The technology graph construction module first integrates three types of unstructured technology text data: technology projects, technology bulletins, and technology funds. It uses the BERT pre-training model to encode the input technology text, extract the semantic information of the text, and generate a semantic feature vector sequence for each word. , P is the number of words in the text; then, the output of the BERT pre-trained model is further processed using a bidirectional LSTM network , capturing the contextual dependencies between words and obtaining an enhanced semantic feature vector sequence , and the output of the bidirectional LSTM network As the input of the conditional random field, the boundaries and types of scientific and technological entities in the text are identified, and scientific and technological entities and attribute information are screened out according to the type, where scientific and technological entities include scientific and technological organizations, scientific and technological funds, scientific and technological projects, and scientific and technological talents; then, the scientific and technological entity relationships are extracted. The scientific and technological entity relationships are determined by the co-occurrence relationship of scientific and technological entities in the text and the similarity of the semantic feature vectors between scientific and technological entities. That is, if the frequency of co-occurrence of two scientific and technological entities in the same context exceeds the threshold And the similarity of semantic feature vectors exceeds the threshold , then there is a relationship edge between the two scientific and technological entities; finally, the identified scientific and technological entities and scientific and technological entity relationships are stored in a graph database or a relational database to build a complete scientific and technological graph; The science and technology knowledge association mining module discovers the potential relationship between science and technology entities and events from the science and technology graph through three methods: science and technology graph association analysis, science and technology graph time series analysis, and science and technology graph time-space analysis. Among them, the science and technology graph association analysis queries the multi-hop relationship of the science and technology graph to form a subgraph of the object of interest, and obtains the hidden association relationship between science and technology entities by checking other entities directly and indirectly related to the science and technology entity; the science and technology graph time series analysis dynamically presents the existence time of the science and technology entity and the start and end time of the science and technology entity relationship, and plays back the science and technology graph state at any time within the start and end time range, presenting the evolution process of the entire life cycle of the science and technology entity; the science and technology graph time-space analysis displays the time-space association relationship between science and technology entities by associating the science and technology graph with the geographical location information of the science and technology entity and displaying it on the three-dimensional earth.

7. The multi-perspective scientific and technological data profiling system according to claim 1, It is characterized in that The technology competition and cooperation situation awareness module includes a key field trend monitoring module, a key field resource competition and cooperation situation analysis module and a key field portrait construction module, wherein: The key science and technology field trend monitoring module collects think tank reports in key fields to form a think tank text library. It uses an open source large language model to automatically extract the content framework, summary and core viewpoints of think tank express reports, and displays the regional distribution of key fields in the form of a map. Hovering the mouse will display the core viewpoints of the latest think tank express reports in key fields. The module for analyzing resource competition and cooperation in key science and technology fields builds a trade competition and cooperation network based on resource reserves and production and trade exchanges. In the trade competition and cooperation network, the import dependence is calculated to show the degree of dependence on trade imports. The stability of trade is measured by calculating the trade stability index, i.e., the network structure entropy, to analyze the trade competition and cooperation situation of the industrial chain and supply chain. The module for constructing portraits of key scientific and technological fields conducts a network analysis of the competition and cooperation in the trade of important mineral resources involved in the integrated circuit and large-capacity battery industries, forms a resource portrait based on the output and reserve data of mineral resource demand, and displays the distribution of mineral resources in order to analyze the potential competition and cooperation between key minerals.

8. A multi-perspective scientific and technological data profiling method, It is characterized in that Based on the multi-perspective scientific and technological data portrait system described in any one of claims 1 to 7, a multi-perspective scientific and technological data portrait is realized, specifically: The science and technology indicator integrated evaluation module is used to build a science and technology indicator portrait, and a three-level science and technology indicator system is established from the dimensions of science and technology input, output and benefit. The relative importance of indicators at each level is determined through the "Delphi method", and the geometric analysis method is used to quantify and rank the indicators, so as to achieve multi-dimensional horizontal / vertical comparative analysis of science and technology competitiveness; Use the technology hotspot labeling module to build a technology hotspot portrait, extract technology keywords from technology text data in parallel through the TextRank algorithm, establish a technology theme word library through the similarity mapping of technology keywords and technology theme word vector encoding, and build a technology hotspot label prediction model and build a sample set training based on the technology theme word library to display the technology hotspots in the future in the form of a label word cloud; Use the technology graph construction analysis module to build a technology graph portrait. Through knowledge graph modeling and technology entity relationship extraction, establish the connection between technology entities. Through technology graph association analysis, technology graph time series analysis, and technology graph spatiotemporal analysis, determine the potential relationship between technology entities, present the evolution of the entire life cycle of technology entities, and show the spatiotemporal relationship between technology entities. Use the technology competition and cooperation situation awareness module to build a portrait of the technology field, extract the core content by analyzing think tank texts, monitor the latest trends in key technology fields, and calculate the import dependence and trade stability index by building a trade competition and cooperation situation network to reveal the competition and cooperation situation of resources in key fields; The multi-dimensional display module of science and technology portraits is used to unify the user portal for the comprehensive display of science and technology indicator portraits, science and technology hotspot portraits, science and technology map portraits, and science and technology field portraits.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the multi-perspective scientific and technological data portrait method of claim 8 is implemented to realize multi-perspective scientific and technological data portrait.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the multi-perspective scientific and technological data portrait method of claim 8 is implemented to realize multi-perspective scientific and technological data portrait.

Images