CN-117350273-B - Data analysis method and device in answer scene, electronic equipment and storage medium

CN117350273BCN 117350273 BCN117350273 BCN 117350273BCN-117350273-B

Abstract

The invention provides a data analysis method, a device, electronic equipment and a storage medium in a response scene, which comprise the steps of obtaining a response text, determining a semantic network and a causal influence graph according to the response text, wherein the semantic network is a node network which takes words screened out in the response text as nodes and mutual information among the words as connecting edges, the causal influence graph is a node graph which takes words screened out in the response text as nodes and causal relations among the words as connecting edges, determining static characteristics according to the semantic network and a preset reference semantic network, determining dynamic characteristics according to the semantic network, determining logical characteristics according to the causal influence graph and the preset reference causal influence graph, determining an index value of an analysis index according to the static characteristics, the dynamic characteristics and the logical characteristics, and determining an analysis result according to the index value. The invention can realize specific and hierarchical identification of the key capability and the performance of the responder and complete comprehensive quantitative and qualitative evaluation of the performance of the responder.

Inventors

WANG CHEN
HONG YIFAN
LI LEFEI
WANG DINGDING
GUO HAORAN

Assignees

清华大学

Dates

Publication Date: 20260512
Application Date: 20230831

Claims (8)

1. A method for analyzing data in a response scene, comprising: Obtaining a response text, and determining a semantic network and a causal influence graph according to the response text, wherein the semantic network takes words screened out of the response text as nodes, determines a node network of connecting edges based on mutual information among the words, and the causal influence graph takes words screened out of the response text as nodes and takes causal relations among the words and the sentences as node diagrams of the connecting edges; Determining static characteristics according to the semantic network and a preset reference semantic network, wherein the static characteristics comprise information for representing the number of nodes in the semantic network and the connection between the nodes; Utilizing the motion process which is shown by the contents of the discussion of the topic views of respondents on the semantic network to construct a wandering path, wherein the wandering path can decompose the wandering direction, and the wandering direction refers to the direction between the nodes in the wandering path; Determining logic characteristics according to the causal influence graph and a preset reference causal influence graph, wherein the logic characteristics comprise information for representing directed connection among nodes in the causal influence graph; determining an index value of an analysis index according to the static feature, the dynamic feature and the logic feature, and determining an analysis result according to the index value, wherein the analysis result comprises results of a plurality of analysis items; wherein, the determining static features according to the semantic network and a preset reference semantic network includes: counting the number of nodes in the semantic network; Determining a first similarity distance between nodes according to mutual information between words in the semantic network and the reference semantic network, wherein the first similarity distance is semantic similarity between words; Clustering the semantic network, determining a node cluster, and calculating the modularity of the network according to the distance between each node in the node cluster and a first center node; calculating PMI matrix distance between the semantic network and the reference semantic network; calculating the skewness of node degree distribution in the semantic network, wherein the node degree is the number of edges connected by nodes, and the skewness is statistics for describing distribution characteristics; Wherein the determining logic features according to the causal effect graph and a preset reference causal effect graph comprises: determining a plurality of causal paths according to the causal influence graph; Determining the words and sentences corresponding to the nodes in the causal path and the matched words and sentences in the reference causal influence graph, and determining the corresponding causal path in the reference causal image graph based on the matched words and sentences; determining causal relationship pointing information between words and sentences in the two corresponding causal paths according to the reference causal influence graph; and determining the error rate of the causal path and the effective demonstration rate of the causal path on the answer theme based on the causal relationship pointing information.
2. The method for analyzing data in response scenario according to claim 1, wherein the determining dynamic characteristics according to the semantic network comprises: Identifying a plurality of travelling paths according to the semantic network, and determining the distance between adjacent nodes in the paths according to the shortest path in the plurality of travelling paths; determining the number of steps between nodes meeting the continuity condition, the stay number of the current node and the number of steps between nodes meeting the jumping condition according to the distance; And determining the number of nodes meeting the regression condition and the number of nodes meeting the divergence condition according to the moving direction of the nodes in the travelling path.
3. The method for analyzing data in response scenario according to claim 1, wherein the method further comprises: acquiring a voice file corresponding to the answer text; analyzing the voice file to divide a plurality of voice fragments; Determining the emotion characteristics of each voice segment, and determining the forward/reverse result and the infection result of each voice segment according to the emotion characteristics.
4. A method of analyzing data in a response scenario according to claim 3, the method further comprising: obtaining a response scene picture corresponding to the response text; identifying the answering scene picture to obtain morphological characteristics of an answer in the picture; And determining the etiquette specification compliance of the responder according to the morphological characteristics.
5. The method according to claim 1 or 4, wherein determining an analysis result from the index value comprises: acquiring the weight corresponding to each analysis index corresponding to each analysis item; And determining the analysis result of the analysis item according to the index value and the weight of the analysis index.
6. A data analysis device in response scene, comprising: The construction module is used for acquiring answer texts, determining a semantic network and a causal influence graph according to the answer texts, wherein the semantic network takes words screened out from the answer texts as nodes, determines node networks of connecting edges based on mutual information among the words, and the causal influence graph takes words screened out from the answer texts as nodes and takes causal relations among the words and the sentences as node diagrams of the connecting edges; The first acquisition module is used for determining static characteristics according to the semantic network and a preset reference semantic network, wherein the static characteristics comprise information for representing the number of nodes in the semantic network and the connection between the nodes, and the first acquisition module is specifically used for counting the number of the nodes in the semantic network, determining a first similar distance between the nodes according to the mutual information between the semantic network and the words in the reference semantic network, wherein the first similar distance is the semantic similarity between the words, clustering the semantic network, determining a node cluster, and calculating the modularity of the network according to the distance between each node in the node cluster and a first central node, wherein the first central node is the center of the node cluster, calculating the PMI matrix distance of the semantic network and the reference semantic network, and calculating the skewness of node degree distribution in the semantic network, wherein the skewness is the number of edges connected by the nodes, and the skewness is statistics for describing distribution characteristics; The second acquisition module is used for determining dynamic characteristics according to the semantic network, wherein the dynamic characteristics comprise information for representing directed paths among nodes in the semantic network; constructing a migration path by utilizing the motion process of the contents of the discussion of the topic views of respondents on the semantic network, wherein the migration path can decompose the migration direction, and the migration direction refers to the direction between nodes in the migration path; The system comprises a first acquisition module, a second acquisition module, a third acquisition module, a first judgment module, a second acquisition module, a third acquisition module, a first judgment module and a second judgment module, wherein the first acquisition module is used for determining a logic characteristic according to the causal influence graph and a preset reference causal influence graph, the logic characteristic comprises information for representing directed connection among nodes in the causal influence graph; And the analysis module is used for determining an index value of an analysis index according to the static characteristic, the dynamic characteristic and the logic characteristic, and determining an analysis result according to the index value, wherein the analysis result comprises results of a plurality of analysis items.
7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the data analysis method in the response scenario according to any one of claims 1 to 5 when the program is executed by the processor.
8. A non-transitory computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the data analysis method in a response scenario according to any one of claims 1 to 5.

Description

Data analysis method and device in answer scene, electronic equipment and storage medium Technical Field The present invention relates to the field of natural language processing technologies, and in particular, to a method and apparatus for analyzing data in response scenarios, an electronic device, and a storage medium. Background In the single interview, the respondent makes relevant discussion on the topic problem, and fully examines the capability of the respondent in multiple aspects. In past interviews, the methods of scale marking and the like are often adopted, and more experts and interviewee officers are relied on for evaluation. Therefore, the cognition, thinking and expression abilities of respondents cannot be evaluated more efficiently, comprehensively and fairly, and more comprehensive and comprehensive interview performance evaluation is difficult to be given. Disclosure of Invention Aiming at the problems existing in the prior art, the invention provides a data analysis method, a device, electronic equipment and a storage medium in a response scene. In a first aspect, the present invention provides a method for analyzing data in a response scenario, including: Obtaining a response text, and determining a semantic network and a causal influence graph according to the response text, wherein the semantic network is a node network which takes words screened out from the response text as nodes and mutual information among the words as connecting edges, and the causal influence graph is a node graph which takes words screened out from the response text as nodes and causal relations among the words as connecting edges; Determining static characteristics according to the semantic network and a preset reference semantic network, wherein the static characteristics comprise information for representing the number of nodes in the semantic network and the connection between the nodes; Determining dynamic characteristics according to the semantic network, wherein the dynamic characteristics comprise information for representing directed paths among nodes in the semantic network; Determining logic characteristics according to the causal influence graph and a preset reference causal influence graph, wherein the logic characteristics comprise information for representing directed connection among nodes in the causal influence graph; and determining an index value of an analysis index according to the static characteristic, the dynamic characteristic and the logic characteristic, and determining an analysis result according to the index value, wherein the analysis result comprises results of a plurality of analysis items. In one embodiment, the determining the static feature according to the semantic network and a preset reference semantic network includes: counting the number of nodes in the semantic network; Determining a first similarity distance between nodes according to mutual information between words in the semantic network and the reference semantic network, wherein the first similarity distance is semantic similarity between words; Clustering the semantic network, determining a node cluster, and calculating the modularity of the network according to the distance between each node in the node cluster and a first center node; calculating PMI matrix distance between the semantic network and the reference semantic network; And calculating the skewness of the node degree distribution in the semantic network, wherein the node degree is the number of edges connected by the nodes, and the skewness is statistics for describing the distribution characteristics. In one embodiment, the determining dynamic characteristics according to the semantic network includes: Identifying a plurality of travelling paths according to the semantic network, and determining the distance between adjacent nodes in the paths according to the shortest path in the plurality of travelling paths; determining the number of steps of the travelling among the nodes meeting the continuity condition, the stay times of the current nodes and the number of steps of the travelling among the nodes meeting the jumping condition according to the moving distance; And determining the number of nodes meeting the regression condition and the number of nodes meeting the divergence condition according to the moving direction of the nodes in the travelling path. In one embodiment, the determining the logic feature according to the causal effect map and a preset reference causal effect map includes: determining a plurality of causal paths according to the causal influence graph; Determining the words and sentences corresponding to the nodes in the causal path and the matched words and sentences in the reference causal influence graph, and determining the corresponding causal path in the reference causal image graph based on the matched words and sentences; determining causal relationship pointing information between words and sentences in the two corres