Knowledge graph-based case retrieval method, device and equipment, and storage medium

A knowledge graph-based case retrieval equipment, comprising a memory, a processor, and a computer program that is stored on the memory and can be run on the processor, wherein the processor implements a knowledge graph-based case retrieval method when executing the computer program, and the method comprises:

• constructing a legal case knowledge graph by analyzing text information based on a predetermined model, and constructing node set data by analyzing the legal case knowledge graph, wherein the text information comprises historical and/or real-time claim information of a plaintiff, defense information of a defendant, evidence information provided by the plaintiff, evidence information provided by the defendant, and historical legal case information;
• performing random-walk sampling on the node set data by using each node of the node set data as a starting point, so as to obtain a plurality of pieces of sequence data;
• training the model by using a word2vec algorithm based on the plurality of pieces of sequence data, so as to obtain an updated target model;
• obtaining target text information, and analyzing the target text information by using the target model, so as to construct a to-be-retrieved knowledge graph;
• retrieving case information and first word embedding vector data associated with the to-be-retrieved knowledge graph from the legal case knowledge graph, and obtaining second word embedding vector data of the to-be-retrieved knowledge graph;
• calculating a first similarity and a second similarity of the case information based on the first word embedding vector data and the second word embedding vector data, and adding the first similarity and the second similarity to obtain a target similarity, wherein the first similarity indicates a similarity of the case information in a content description, and the second similarity indicates a similarity of the case information in a logical relationship;
• and
• sorting the case information in descending order of values of the target similarity, and outputting the sorted case information;
• wherein the performing random-walk sampling on the node set data by using each node of the node set data as a starting point, so as to obtain a plurality of pieces of sequence data comprises:
• obtaining a weight of a next node of the starting point by using each node of the node set data as the starting point;
• analyzing the weight, so as to set a random walk step number, and analyzing the number of nodes of a branch that the starting point is located, so as to set a random walk step number threshold;
• when it is determined that the random walk step number is less than the random walk step number threshold, walking along a node connected to the next node of the starting point based on the random walk step number, generating random walk sequence data, and obtaining the plurality of pieces of sequence data;
• and
• when it is determined that the random walk step number is greater than or equal to the random walk step number threshold, walking along a node connected to the next node of the starting point based on the random walk step number threshold, generating random walk sequence data, and obtaining the plurality of pieces of sequence data.

A non-transitory computer-readable storage medium comprising instructions, wherein when the instructions are run on a computer, the computer is enabled to perform a knowledge graph-based case retrieval method, comprising:

• constructing a legal case knowledge graph by analyzing text information based on a predetermined model, and constructing node set data by analyzing the legal case knowledge graph, wherein the text information comprises historical and/or real-time claim information of a plaintiff, defense information of a defendant, evidence information provided by the plaintiff, evidence information provided by the defendant, and historical legal case information;
• performing random-walk sampling on the node set data by using each node of the node set data as a starting point, so as to obtain a plurality of pieces of sequence data;
• training the model by using a word2vec algorithm based on the plurality of pieces of sequence data, so as to obtain an updated target model;
• obtaining target text information, and analyzing the target text information by using the target model, so as to construct a to-be-retrieved knowledge graph;
• retrieving case information and first word embedding vector data associated with the to-be-retrieved knowledge graph from the legal case knowledge graph, and obtaining second word embedding vector data of the to-be-retrieved knowledge graph;
• calculating a first similarity and a second similarity of the case information based on the first word embedding vector data and the second word embedding vector data, and adding the first similarity and the second similarity to obtain a target similarity, wherein the first similarity indicates a similarity of the case information in a content description, and the second similarity indicates a similarity of the case information in a logical relationship;
• and
• sorting the case information in descending order of values of the target similarity, and outputting the sorted case information;
• wherein the performing random-walk sampling on the node set data by using each node of the node set data as a starting point, so as to obtain a plurality of pieces of sequence data comprises:
• obtaining a weight of a next node of the starting point by using each node of the node set data as the starting point;
• analyzing the weight, so as to set a random walk step number, and analyzing the number of nodes of a branch that the starting point is located, so as to set a random walk step number threshold;
• when it is determined that the random walk step number is less than the random walk step number threshold, walking along a node connected to the next node of the starting point based on the random walk step number, generating random walk sequence data, and obtaining the plurality of pieces of sequence data;
• and
• when it is determined that the random walk step number is greater than or equal to the random walk step number threshold, walking along a node connected to the next node of the starting point based on the random walk step number threshold, generating random walk sequence data, and obtaining the plurality of pieces of sequence data.