Research On The Construction And Application Of Ethnic Culture Knowledge Question And Answer Model Based On Knowledge Embedding Representation

Ethnic culture is an important carrier of national spirit development and inheritance,and an important content of traditional culture education.With the diversification of online learning forms and the digitization of resource contents,the fragmentation,diversity and unstructured characteristics of ethnic culture disseminated in different media forms have become more and more prominent,making it difficult to be effectively organized,tapped and deeply applied in more scenarios.How to use the new generation of information technology to enhance the vitality of national culture innovation and creation has become a key issue to be solved.In this paper,we realize the organization and integration of multi-source heterogeneous ethnic culture knowledge based on knowledge embedding representation technology,and realize the application of ethnic culture innovation through ethnic culture knowledge question and answer model.The main research work of this paper is as follows:(1)Knowledge graph construction of ethnic culture domain.Based on the multisource heterogeneous and diverse characteristics of ethnic culture knowledge,this paper collects semi-structured and unstructured ethnic culture data by crawling technology based on the knowledge system architecture of ethnic culture domain,constructs the ontology concept layer of ethnic culture knowledge map by top-down approach,and completes the extraction of ethnic culture knowledge by using predefined triad rules and deep learning method,and finally stores the triad by graph database Neo4 j stores the triads to realize the construction of the knowledge graph.(2)A deep learning-based knowledge graph embedding algorithm is proposed.To address the problems of diverse knowledge structures,complex relationships and modeling difficulties in the process of knowledge map construction,this paper proposes a deep learning-based knowledge map embedding algorithm.This algorithm maps entities and relationships into a two-dimensional Minkowski space using a deep learning model,defines relationships as rotations in space to learn the complex relationships existing in the knowledge map,and obtains a low-latitude vector representation of the knowledge map.And experiments on several publicly available datasets for link prediction tasks are conducted.The experimental results show that this model can effectively capture the complex relationships in the knowledge graph and outperforms the baseline models such as Trans E,Trans H and Trans R in terms of indicator ten hits and average ranking.(3)Unsupervised sentence embedding based on contrast learning.In order to solve the problems that traditional sentence embedding models cannot learn the hidden semantic connections between sentences and ignore the complex semantic relationships between sentences,this paper proposes an unsupervised sentence embedding model based on contrast learning,which uses a pre-trained model to obtain the vector representation of sentences,uses contrast learning to enhance the learning effect,and sets the agent task to data enhancement as well as R-Drop strategy.improve the accuracy of result search in the retrieval-based question-and-answer model.The experimental comparison of text semantic similarity tasks on public datasets STS12-16,STS B,and SICK-R shows that the sentence embedding model proposed in this paper outperforms the baseline model in terms of both Spearman correlation coefficients.Meanwhile,this paper also uses the Sen Eval toolkit to evaluate the quality of sentence embeddings generated by the model,and uses Sen Eval to use sentence embeddings as features of migration tasks MR,SUBJ,MPQA,TREC and MRPC for classification tasks,further validating the superiority of the sentence embedding model proposed in this paper for classification tasks.(4)Compound question-answer model construction for ethnic culture knowledge.In order to further improve the accuracy of the Q&A model under the limited data scale,this paper designs a composite Q&A model based on the knowledge embedding representation technology,which consists of two parts: the Q&A based on knowledge graph embedding and the retrieval-based Q&A based on sentence embedding.The former uses the knowledge map embedding algorithm proposed in this paper to embed the entities and relations in the knowledge map of ethnic culture,designs the head entity and relation learning model to parse the question and obtain the tail entity by calculation;the latter uses the sentence embedding model proposed in this paper to embed the representation of the ethnic culture question and answer corpus and constructs the vector index table using the HNSW algorithm,and obtains the vector representation after the question is input and After the question is input,the vector representation is obtained and the vector that is closest to the question vector is queried in the vector index table according to the greedy strategy and output.The questionanswer model makes full use of the advantages of the two retrieval models and achieves a higher accuracy rate in the multiple comparison experiments of the ethnic culture knowledge quiz task.(5)Design and implementation of the ethnic culture knowledge quiz system.In this paper,the proposed composite model is used in the actual Q&A scenario using Web technology to design and implement the ethnic culture knowledge Q&A system.The system is designed with three main modules,namely,the ethnic culture knowledge map visualization module,the ethnic culture knowledge Q&A module and the ethnic culture knowledge discovery module,which support the functions of map display based on entity query,knowledge Q&A based on user definition,and online knowledge automatic extraction,entity query and knowledge map completion.Compared with the traditional Q&A system,this system can not only provide users with more available information including the knowledge map,but also better understand the user’s intention and get more accurate answers by using the composite Q&A model of ethnic culture knowledge,and the knowledge discovery module can automatically collect text data related to ethnic culture knowledge and build a knowledge map to supplement the knowledge base,improving service efficiency and quality and providing better user experience.