CN-122000078-A - Assessment information generation method, device and equipment based on large model and storage medium

CN122000078ACN 122000078 ACN122000078 ACN 122000078ACN-122000078-A

Abstract

The application relates to the technical field of intelligent medical treatment, and discloses a method, a device, equipment and a storage medium for generating evaluation information based on a large model, wherein the method comprises the steps of mapping each case text into a vertex, and forming a plurality of different vertices into a vertex set; the method comprises the steps of determining the superside of each disease category, the superside of each semantic cluster, the superside of each logic mode and the superside of each length interval based on a vertex set, a first model, a second model, a third model and a fourth model, carrying out union processing on the superside of each disease category, the superside of each semantic cluster, the superside of each logic mode and the superside of each length interval to obtain a superside set, constructing a supergraph according to the superside set and the vertex set, processing the supergraph through a greedy screening algorithm to obtain a plurality of candidate case texts, and generating evaluation information of a case to be queried based on the plurality of candidate case texts and the large model. The method and the device are beneficial to the generation efficiency of the evaluation information of the case to be queried.

Inventors

TANG HONGKAI
LIANG WEI
Zuo Yuke
HUANG XUE
XU NUO
HU HANG

Assignees

湖南红普创新科技发展有限公司

Dates

Publication Date: 20260508
Application Date: 20251225

Claims (10)

1. A large model-based evaluation information generation method, characterized in that it is applied to an electronic device, the evaluation information generation method comprising: Obtaining a plurality of case texts from a database, mapping each case text into a vertex, and forming a vertex set by a plurality of different vertices; Determining the superside of each disease category, the superside of each semantic cluster, the superside of each logic mode and the superside of each length interval based on the vertex set, the first model, the second model, the third model and the fourth model; the method comprises the steps of performing union processing on the supersides of each disease category, the supersides of each semantic cluster, the supersides of each logic mode and the supersides of each length interval to obtain a superside set; Constructing a hypergraph according to the hyperedge set and the vertex set, and processing the hypergraph through a greedy screening algorithm to obtain a plurality of candidate case texts; when a case to be queried is received, based on the multiple candidate case texts and the large model, evaluation information of the case to be queried is generated.
2. The evaluation information generating method according to claim 1, wherein determining the superside of each disease category, the superside of each semantic cluster, the superside of each logical pattern, and the superside of each length section based on the vertex set, the first model, the second model, the third model, and the fourth model comprises: generating the superside of each disease category according to the vertex set and the preset first model, generating the superside of each disease category according to the vertex set and the preset second model, generating the superside of each logic mode according to the vertex set and the preset third model, and generating the superside of each length interval according to the vertex set and the preset fourth model.
3. The method for generating evaluation information according to claim 1, wherein, The method comprises the steps of constructing a hypergraph according to a hyperedge set and a vertex set, processing the hypergraph through a greedy screening algorithm to obtain a plurality of candidate case texts, and comprises the following steps: generating a weight coefficient of each superside in the superside set according to the superside set and the fifth model, and constructing a supergraph according to the weight coefficient of each superside in the superside set and the vertex set; And performing information extraction operation on the hypergraph through a greedy screening algorithm to obtain a plurality of candidate case texts.
4. The evaluation information generation method according to claim 1, wherein the generating evaluation information of the case to be queried based on the plurality of candidate case texts and the large model when the case to be queried is received, comprises: When a case to be queried is received, screening a plurality of candidate case texts by adopting an agent, and forming a text combination by the screened plurality of candidate case texts; And inputting the text combination into a large model, and generating evaluation information of the case to be queried by the large model.
5. The evaluation information generation method according to claim 1, characterized in that after the evaluation information of the case to be queried is generated based on a plurality of candidate case texts and a large model when the case to be queried is received, the evaluation information generation method comprises: Inputting the evaluation information into a text coding module, generating a characteristic vector of the evaluation information by the text coding module, and writing the characteristic vector of the evaluation information into a vector database.
6. The method for generating evaluation information according to claim 1, wherein, A first model, defined as follows: ; Represent the first Superb of individual disease categories, item The superside of each disease category comprises tag information which is mapped by a mapping function and belongs to the first Of individual disease categories ; Represent the first Label information of individual case text; Represent the first Vertex corresponding to each case text; representing a set of vertices; representing the total number of case texts; Mapping functions; Representing that the mapping function will be Tag information of individual case text is mapped to the first in disease classification space A disease category; a serial number indicating a disease type; a second model, defined as follows: ; Representing the first Superedge of semantic cluster, the first Superedge inclusion of semantic clusters belonging to the first Of semantic clusters ; Represent the first Vertex corresponding to each case text; representing a set of vertices; : Is a collection Elements of (a) and (b); Representing an encoding function; Represent the first Semantic coding vectors of individual case text; Representation of And Splicing the components to be spliced, Represent the first The word vector of the text of each case, Represent the first Sentence pattern vector of each case text; Represent the first Vectors of the clustering centers; Representing the distance between euclidean norms, measuring the similarity of the euclidean norms, wherein the smaller the value is, the more similar the euclidean norms are; Is shown in Take the value of Within this range, find the distance expression Obtaining an argument of a minimum value; a third model, defined as follows: ; Wherein, the Represent the first Superedge of the seed logic mode, the th The superside of the seed logic pattern includes that the logic relationship belongs to the first Of the logic mode ; Represent the first Vertex corresponding to each case text; representing a set of vertices; : Is a collection Elements of (a) and (b); Representing a logic feature extraction function for extracting causal, turning or time sequence connective words; Represent the first A logic mode is selected; Representing a similarity function; a preset threshold value; A fourth model, defined as follows: ; Wherein, the Represent the first Overrun of length interval, the first Super-edge inclusion of length intervals , In ] ; Represent the first Vertex corresponding to each case text; representing a set of vertices; : Is a collection Elements of (a) and (b); Represent the first The number of the lemmas after the segmentation of the individual case text; represents the lower limit of the mth length interval; The upper limit of the mth length section is indicated.
7. The evaluation information generation method according to claim 3, wherein the fifth model is defined as follows: ; the e-th superside weight coefficient in the superside set is represented; A weight value representing the category to which the e-th superside belongs in the superside set; Representing the total number of vertices contained in the hypergraph; Is the first in the hyperedge set The number of vertices that the bar contains, To prevent a smooth term with a denominator of zero.
8. An evaluation information generation apparatus based on a large model, characterized by being applied to an electronic device, comprising: the acquisition module is used for acquiring a plurality of case texts from the database, mapping each case text into a vertex, and forming a vertex set by a plurality of different vertices; The determining module is used for determining the superside of each disease category, the superside of each semantic cluster, the superside of each logic mode and the superside of each length interval based on the vertex set, the first model, the second model, the third model and the fourth model; the processing module is used for carrying out union processing on the supersides of each disease category, the supersides of each semantic cluster, the supersides of each logic mode and the supersides of each length interval to obtain a superside set; The construction module is used for constructing a hypergraph according to the hyperedge set and the vertex set, and processing the hypergraph through a greedy screening algorithm to obtain a plurality of candidate case texts; and the generation module is used for generating evaluation information of the case to be queried based on the plurality of candidate case texts and the large model when the case to be queried is received.
9. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the evaluation information generation method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the evaluation information generation method according to any one of claims 1 to 7.

Description

Assessment information generation method, device and equipment based on large model and storage medium Technical Field The application relates to the technical field of intelligent medical treatment, in particular to an evaluation information generation method, device and equipment based on a large model and a storage medium. Background In the medical flow system, after the case to be queried enters the medical system, the medical system can distribute the case to be queried to doctors corresponding to special departments and corresponding qualification before diagnosis and treatment start based on the evaluation information of the case to be queried, so that the flow redundancy of the non-special doctors for consultation and re-consultation is avoided. However, the generation process of the evaluation information of the existing case to be queried is complicated, which is not beneficial to improving the generation efficiency of the evaluation information. The reason is that in the current technical application scenario, the main mode of generating the evaluation information of the case to be queried is that the staff completes the writing work through manual operation, and the manual operation mode consumes a great deal of manpower resources and time resources, so that the generation time of the evaluation information of the case to be queried is increased, and the evaluation information is easily influenced by manual intervention, thereby being unfavorable for improving the generation efficiency of the evaluation information. Disclosure of Invention The embodiment of the application provides a method, a device, equipment and a storage medium for generating evaluation information based on a large model, which are used for solving the technical problems that the generation process of the evaluation information of the existing case to be queried is complicated and the generation efficiency of the evaluation information is not beneficial to improvement. In a first aspect, an embodiment of the present application provides a method for generating evaluation information based on a large model, which is applied to an electronic device, where the method for generating evaluation information includes: Obtaining a plurality of case texts from a database, mapping each case text into a vertex, and forming a vertex set by a plurality of different vertices; Determining the superside of each disease category, the superside of each semantic cluster, the superside of each logic mode and the superside of each length interval based on the vertex set, the first model, the second model, the third model and the fourth model; the method comprises the steps of performing union processing on the supersides of each disease category, the supersides of each semantic cluster, the supersides of each logic mode and the supersides of each length interval to obtain a superside set; Constructing a hypergraph according to the hyperedge set and the vertex set, and processing the hypergraph through a greedy screening algorithm to obtain a plurality of candidate case texts; when a case to be queried is received, based on the multiple candidate case texts and the large model, evaluation information of the case to be queried is generated. In a possible implementation manner of the first aspect, the determining, based on the vertex set, the first model, the second model, the third model, and the fourth model, a hyperedge of each disease category, a hyperedge of each semantic cluster, a hyperedge of each logical mode, and a hyperedge of each length interval includes: generating the superside of each disease category according to the vertex set and the preset first model, generating the superside of each disease category according to the vertex set and the preset second model, generating the superside of each logic mode according to the vertex set and the preset third model, and generating the superside of each length interval according to the vertex set and the preset fourth model. In a possible implementation manner of the first aspect, the building a hypergraph according to the hyperedge set and the vertex set, processing the hypergraph by a greedy filtering algorithm to obtain a plurality of candidate case texts includes: generating a weight coefficient of each superside in the superside set according to the superside set and the fifth model, and constructing a supergraph according to the weight coefficient of each superside in the superside set and the vertex set; And performing information extraction operation on the hypergraph through a greedy screening algorithm to obtain a plurality of candidate case texts. In a possible implementation manner of the first aspect, when the case to be queried is received, generating evaluation information of the case to be queried based on a plurality of candidate case texts and a large model includes: When a case to be queried is received, screening a plurality of candidate case texts by adopting an agent,