Research On The Construction And Application Of Event-triggered Verb Data Set Of Classics Oriented To Digital Humanities

With the emergence of massive digital resources of ancient texts,as well as the continuous advancement of text mining technology and human computing tools,ancient Chinese information processing has begun to develop in a more intelligent and deep-semantic direction.In-depth mining and exploration of ancient texts appears to be more and more important.The more important.In the use of information and topic mining of classics,automatic event recognition and extraction is a new and important topic at present,and the recognition of event trigger words is a basic and key task.Trigger words refer to words that characterize the occurrence of an event.The process of trigger word recognition is essentially the process of judging the type of event through automatic extraction and classification of trigger words.At present,the recognition methods of event-triggered words are mainly based on statistics,rule-based and machine learning.They have achieved good results in modern text research in some specific fields.However,when facing the text of ancient books,due to the particularity of the writing structure and syntax of ancient books,there is a lack of general trigger word extraction rules.Therefore,further exploration in the construction of basic dictionaries is needed,and a complete classification system of trigger words in ancient books has not been established.There are certain difficulties in the recognition and classification of trigger words.Starting from the in-depth development of ancient books and documents,this paper uses the technology used in the research of trigger word recognition in specific fields,combined with the progress of trigger word recognition in modern texts,and explores the methods of automatic recognition and event extraction of event trigger words in ancient books.Strive to provide help for the research on the theme of ancient books,and be used to verify or assist related research in the humanities,and provide better methods for the research of ancient Chinese documents in the field of digital humanities.This article first uses natural language processing technology to preprocess the classic text to remove stop words and verb extraction;then use the LDA model to cluster the topics of the classic verbs,and summarize the topic categories through qualitative analysis,and construct Classical trigger verb classification system;then based on the results of the classification system and clustering,the trigger word seed word set is constructed,combined with the research results of the humanities and multiple classic dictionary resources,the dictionary semantic features and verb context features are extracted,and then the semantic similarity is used The text mining technology such as degree calculation expands the seed word set.Finally,this paper conducts empirical research on the proposed trigger word classification system and data set construction method,and conducts manual verification and consistency test on the results.In the empirical exploration and application research,this article uses Zuo Zhuan as the experimental object,constructs a trigger verb classification system with 10 categories and 26 subcategories,and builds a trigger word meaning data set on this basis.Then,according to the classification system and data set,the original event sentences were manually classified and labeled and structured,and the Kappa coefficient was used to check the consistency of the labeling results.Subsequently,the deep learning algorithm Bi-LSTM was used to design a classification experiment,and the application research of the data set was carried out;the trigger word data set was used for the automatic classification of the four different types of event sentence text in Zuo Zhuan,and the comparison was set by adjusting the experimental parameters.In the experiment,the standard was selected to evaluate the results of the experiment.The results show that the method proposed in this paper is feasible and effective,and the event-triggered word dataset constructed based on this has a high credibility.