CN-121998052-A - Method, equipment and computer program product for constructing knowledge graph in radio frequency domain

CN121998052ACN 121998052 ACN121998052 ACN 121998052ACN-121998052-A

Abstract

The application discloses a method, equipment and a computer program product for constructing a knowledge graph in the radio frequency field, and relates to the technical field of knowledge graph creation; the method comprises the steps of respectively carrying out sub-domain recognition on a plurality of semantic blocks by utilizing a pre-constructed domain dictionary, carrying out semantic matching on a prompt word template from a pre-constructed multi-level prompt word library based on the result of the sub-domain recognition, calling a pre-constructed large language model, carrying out entity relation extraction on the semantic blocks based on the prompt word template to obtain an entity relation extraction result, carrying out structure verification and fault tolerance mechanism processing on the entity relation extraction result to obtain a candidate knowledge triplet, and constructing a radio frequency domain knowledge graph based on the candidate knowledge triplet. The high-consistency and high-precision radio frequency domain knowledge graph construction is realized through semantic segmentation, sub-domain identification, multi-level prompt word library, structure verification and fault tolerance mechanism processing.

Inventors

LI JIANQIANG
Jin Xiongnan
YAN YAN
LI QINGQING
HUANG XINBIN
LIN JIANMING
ZHU QINGLING
CHEN QUANGANG
GE LEI

Assignees

深圳大学

Dates

Publication Date: 20260508
Application Date: 20251226

Claims (10)

1. The method for constructing the knowledge graph in the radio frequency field is characterized by comprising the following steps of: semantic segmentation is carried out on a pre-acquired radio frequency text data stream, and a plurality of semantic blocks are generated; Sub-domain recognition is respectively carried out on the plurality of semantic blocks by utilizing a pre-constructed domain dictionary, and a prompting word template is matched with the semantics from a pre-constructed multi-level prompting word library based on the result of the sub-domain recognition; Invoking a pre-constructed large language model, and extracting entity relations from the semantic blocks based on the prompt word template to obtain entity relation extraction results; Performing structure verification and fault tolerance mechanism processing on the entity relation extraction result to obtain a candidate knowledge triplet; And constructing a radio frequency domain knowledge graph based on the candidate knowledge triples.
2. The method for constructing a knowledge graph in a radio frequency domain according to claim 1, wherein the step of performing semantic segmentation on the pre-acquired radio frequency text data stream to generate a plurality of semantic blocks further comprises: collecting technical documents in the radio frequency field; performing code conversion and layout analysis on the technical document in the radio frequency field to obtain an original radio frequency text data stream; and filtering the original radio frequency text data stream to obtain the radio frequency text data stream.
3. The method for constructing a knowledge graph in a radio frequency domain according to claim 1, wherein the step of performing sub-domain recognition on the plurality of semantic blocks using a pre-constructed domain dictionary, and performing semantic matching on a hint word template from a pre-constructed multi-level hint word library based on a result of the sub-domain recognition further comprises: constructing a domain dictionary based on a predetermined radio frequency domain knowledge system; a task target is extracted based on the preset radio frequency domain knowledge to design multi-level prompt words, wherein the multi-level prompt words comprise task driving prompt words, domain guiding prompt words and format constraint prompt words, and each level of prompt words comprises a plurality of prompt word templates; and storing the multi-level prompt words in a classified mode, and constructing a multi-level prompt word library.
4. The method for constructing a knowledge graph in a radio frequency domain according to claim 1, wherein the step of performing sub-domain recognition on the plurality of semantic blocks by using a pre-constructed domain dictionary, respectively, and performing semantic matching on a hint word template from a pre-constructed multi-level hint word library based on a result of the sub-domain recognition comprises: Sub-domain recognition is respectively carried out on the plurality of semantic blocks by using the domain dictionary, so that a core sub-domain corresponding to each semantic block is obtained; Analyzing key features of the semantic block, and carrying out semantic matching in the multi-level prompt word library based on the core sub-field and the key features to obtain a candidate prompt word template set; And selecting an optimal prompting word template from the candidate prompting word template set by combining the key features, the historical extraction performance and the current task type.
5. The method for constructing a knowledge graph in a radio frequency domain according to claim 1, wherein the step of calling a pre-constructed large language model, extracting entity relations from the semantic blocks based on the prompt word template, and obtaining an entity relation extraction result comprises: Supplementing the context association of the prompt word template, and adding an example triplet and field explanation to the prompt word template to obtain an enhanced prompt word template: And calling a pre-built large language model, and inputting the enhanced prompt word template and the semantic block into the pre-built large language model to extract the entity relationship, so as to obtain an entity relationship extraction result.
6. The method for constructing a knowledge graph in a radio frequency domain according to claim 1, wherein the step of performing structure verification and fault tolerance mechanism processing on the entity relation extraction result to obtain a candidate knowledge triplet comprises: Screening a knowledge triplet which accords with a predefined triplet structure and a predefined semantic specification from the entity relation extraction result as a first candidate knowledge triplet; taking a knowledge triplet which does not accord with the predefined triplet structure or does not accord with the predefined semantic specification in the entity relation extraction result as a fourth candidate knowledge triplet; checking whether the first candidate knowledge triples belong to a preset allowed type set or not; If the first candidate knowledge triples belong to the allowed type set, carrying out format normalization and unit normalization on numerical value type parameters in the first candidate knowledge triples, and carrying out entity writing unification on the first candidate knowledge triples to obtain a second candidate knowledge triples; Deleting the conflict relation or incomplete relation in the second candidate knowledge triplet to obtain a third candidate knowledge triplet; repairing the fourth candidate knowledge triplet based on a preset fault tolerance mechanism to obtain a fifth candidate knowledge triplet; and obtaining the candidate knowledge triples based on the third candidate knowledge triples and the fifth candidate knowledge triples.
7. The method for constructing a knowledge-graph in a radio frequency domain according to claim 1, wherein the step of constructing the knowledge-graph in a radio frequency domain based on the candidate knowledge triples comprises: Writing the candidate knowledge triples and semantic blocks corresponding to the candidate knowledge triples into an incident frequency domain knowledge base to form a knowledge graph node and a knowledge graph relation; And constructing a knowledge graph of the radio frequency domain according to the knowledge graph nodes and the knowledge graph relationship.
8. The method for constructing a knowledge graph in a radio frequency domain according to claim 1, wherein the step of constructing the knowledge graph in a radio frequency domain based on the candidate knowledge triples further comprises: obtaining an error log and an extraction statistical result generated by processing the fault-tolerant mechanism; analyzing the error log and the extraction statistical result, and identifying a high-frequency error type and a weak subarea; adjusting the domain dictionary and the multi-level prompt word stock according to the high-frequency error type and the weak subarea; And returning to the execution step according to the adjusted domain dictionary and the multi-level prompt word library, namely respectively carrying out sub-domain recognition on the plurality of semantic blocks by utilizing a pre-built domain dictionary, and carrying out semantic matching on the prompt word templates from the pre-built multi-level prompt word library based on the result of the sub-domain recognition.
9. A radio frequency domain knowledge graph construction device, characterized in that the device comprises a memory, a processor and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the radio frequency domain knowledge graph construction method according to any one of claims 1 to 8.
10. A computer program product, characterized in that the computer program product comprises a computer program which, when being executed by a processor, implements the steps of the radio frequency domain knowledge graph construction method according to any one of claims 1 to 8.

Description

Method, equipment and computer program product for constructing knowledge graph in radio frequency domain Technical Field The present application relates to the field of knowledge graph creation technology, and in particular, to a method, an apparatus, and a computer program product for knowledge graph construction in the radio frequency field. Background Because of the high complexity and multi-sub-field difference of the radio frequency technical document, if the radio frequency domain knowledge graph construction is carried out by adopting the prior knowledge extraction method driven by the unified Prompt or the static Prompt template, the self-adaptive adjustment cannot be carried out according to the semantic feature difference of the multi-sub-field of the radio frequency domain, the problems of entity omission, relation error, incomplete structure and the like occur, so that the accurate radio frequency domain knowledge graph cannot be constructed, in addition, the prior knowledge extraction method lacks semantic segmentation and context maintaining means, so that the parameter link, structural relation or flow relation in the radio frequency domain is broken in the extraction process, the semantic integrity of the complex document in the radio frequency domain is damaged, knowledge breakage is caused, and the unification is difficult. Therefore, how to construct a high-consistency and high-precision knowledge graph in the radio frequency domain becomes a technical problem to be solved in the application. Disclosure of Invention The application mainly aims to provide a method, equipment and a computer program product for constructing a radio frequency domain knowledge graph, which aim to solve the technical problem of how to construct a radio frequency domain knowledge graph with high consistency and high precision. In order to achieve the above purpose, the present application provides a method for constructing a knowledge graph in a radio frequency domain, which comprises: semantic segmentation is carried out on a pre-acquired radio frequency text data stream, and a plurality of semantic blocks are generated; Sub-domain recognition is respectively carried out on the plurality of semantic blocks by utilizing a pre-constructed domain dictionary, and a prompting word template is matched with the semantics from a pre-constructed multi-level prompting word library based on the result of the sub-domain recognition; Invoking a pre-constructed large language model, and extracting entity relations from the semantic blocks based on the prompt word template to obtain entity relation extraction results; Performing structure verification and fault tolerance mechanism processing on the entity relation extraction result to obtain a candidate knowledge triplet; And constructing a radio frequency domain knowledge graph based on the candidate knowledge triples. In an embodiment, the step of performing semantic segmentation on the pre-acquired radio frequency text data stream to generate a plurality of semantic blocks further includes: collecting technical documents in the radio frequency field; performing code conversion and layout analysis on the technical document in the radio frequency field to obtain an original radio frequency text data stream; and filtering the original radio frequency text data stream to obtain the radio frequency text data stream. In an embodiment, the step of using a pre-built domain dictionary to perform sub-domain recognition on the plurality of semantic blocks, and based on the result of the sub-domain recognition, performing semantic matching on the alert word templates from a pre-built multi-level alert word library further includes: constructing a domain dictionary based on a radio frequency domain knowledge system; The method comprises the steps of extracting task targets based on radio frequency domain knowledge, designing multi-level prompt words, wherein the multi-level prompt words comprise task driving prompt words, domain guiding prompt words and format constraint prompt words, and each level of prompt words comprises a plurality of prompt word templates; and storing the multi-level prompt words in a classified mode, and constructing a multi-level prompt word library. In an embodiment, the step of performing sub-domain recognition on the plurality of semantic blocks by using a pre-built domain dictionary, and performing semantic matching on a hint word template from a pre-built multi-level hint word library based on a result of the sub-domain recognition includes: Sub-domain recognition is respectively carried out on the plurality of semantic blocks by using the domain dictionary, so that a core sub-domain corresponding to each semantic block is obtained; Analyzing key features of the semantic block, and carrying out semantic matching in the multi-level prompt word library based on the core sub-field and the key features to obtain a candidate prompt word template set; And selec