Exploring Local Grammar Of Evaluation In Media Discourse With NLP Techniques

Local grammar focuses on particular word usage or structure and relates to specific textual categories and language usage.Based on the theory framework of local grammar of evaluation,this thesis takes adjective evaluative patterns as the research object,and combines with the emerging natural language processing(NLP)techniques to discuss the frequency distribution differences of adjective evaluative patterns and its evaluation orientation in Chinese and American media discourses.Natural language processing techniques can replace the manual way to collect and classify the massive data on the internet scientifically and efficiently.This thesis adopts a self-designed crawler program to collect and sort out the corpus,and constructs a corpus including China Daily and USA Today media discourses texts respectively.This thesis uses the four adjective evaluative patterns “IT + LINK VERB ADJECTIVE GROUP CLAUSE”,“THERE LINK VERB + SOMETHING/ANYTHING/NOTHING ADJECTIVE GROUP + ABOUT/IN + NOUN GROUP/-ING CLAUSE”,“LINK VERB ADJECTIVE GROUP + TO-INFINITIVE CLAUSE”,“LINK VERB + ADJECTIVE GROUP + THAT CLAUSE” proposed by Hunston and Sinclair(2000)as the research object and combines the NLP techniques developed in recent years,using regular expressions and POS tagging in NLTK as the critical criteria,and then the adjective evaluative patterns are automatically retrieved in the two sub-corpora.Based on the classification of adjective evaluation semantics by Groom(2005)and the functional category of adjective evaluation patterns proposed by Hunston and Sinclair(2000),three questions are addressed in this thesis:1.What is the role of NLP techniques in this study?2.What are the distribution differences between Chinese and American media discourse in using adjective evaluative patterns?3.How are the evaluation orientation of adjective evaluative patterns reflected in Chinese and American media discourses?We find that(1)NLP techniques help a lot in collecting and sorting out the research data in this study,saving a lot of time,manpower and material resources.Additionally,it also contributes to automatic annotation,word segmentation,syntactic parse,and retrieval,judgement,and quantity statistics of evaluative patterns in texts.(2)The frequency of these four adjective evaluative patterns in American media discourses is higher than that in Chinese media discourses.Additionally,the frequency of the first pattern is the highest in both Chinese and American media discourse.(3)Researches on the evaluation orientation of the four patterns of the adjective evaluative patterns in Chinese media discourse and American media discourses can be concluded: In the first pattern,the Chinese media discourses have an evaluation orientation of Necessity,Possibility,and Difficulty,and in American media discourses have an evaluation orientation that expresses Certainty,Difficulty,and Necessity.In the second pattern,both the Chinese media discourse and the American media discourses have an evaluation orientation that expresses predictability.In the third pattern,Chinese media discourses have an evaluation orientation of Possibility,Necessity,and Difficulty and American media discourses have an evaluation orientation that expresses Difficulty,Desirability,and Possibility.In the fourth pattern,both Chinese media discourses and American media discourses have an evaluation orientation of expressing Certainty,Possibility,and Necessity,which still convey individual emotional responses to the situation.This thesis is instructive for the four adjective evaluative patterns in Chinese and American media discourses and provides a new perspective on studying local grammar of evaluation.NLP techniques can not only help people collect and manage huge data scientifically and effectively,and obtain some implied rules in the data,but also realize the functions of automatic annotation,lemmatization,syntax parse of text,etc.