A Quantitative Analysis Of The Structural Complexity Of English Sentence And Its Cross-register Comparison Based On Pattern Grammar

The description of sentence construction has great value in natural language processing(NLP)and language testing.However,the two prevailing rule-based syntactic analysis systems,phrase structure grammar and dependency grammar,have the deficiencies that they cannot completely reflect the real language.To remedy the deficiency,the study proposed pattern grammar as a new perspective to describe sentence construction due to its three features which are in accordance with the law of real natural language:the support of empirical evidence from the corpus of real language,the linear analysis of sentence construction in line with human psychology,and the close association of form and meaning and the greater flexibility of its semi-fixed expressions.Previous research on pattern grammar is still mainly at the lexis level.To complement its lack at the sentence level,the present study explored the feasibility of using pattern grammar to describe sentence construction,by conducting a preliminary verification of the feasibility through testing the capacity of pattern characteristics to reflect the structural complexity of sentenceBased on a corpus-based quantitative method,the present study attempted to confirm the feasibility of using pattern grammar to reflect the structural complexity of sentence in the English language by examining whether the universal laws of human language was reflected and whether the functional basis of register variation was met The employment of the British National Corpus Baby Edition(BNC Baby),which contains four registers and of which patterns were marked out into lists,and the establishment of five pattern measurement indicators,of which the two basic ones are the number of pattern flows(NPF)and the mean pattern length(MPL),reflect the degree of structural complexity of sentence from the two angles,"depth" and "length",provided a solid foundation for the quantitative analysis.Then,more related indicators were added for the comparison across registersThe results have shown that the universal laws of language,the economy principle of language and the limitation of working memory capacity,were both reflected by the degree of structural complexity of sentence shown through the five indicators.The functional basis of register variation was also met,since by comparing the pattern characteristics revealed through the related indicators across the four registers,conversation,fiction,newspaper,and academic,three summarized aspects of the variations in structural complexity of sentence,i.e.,degree,type(elaboration or compression),and information density could all be explained by the differences in situational characteristics.As thus,the feasibility of using pattern grammar to reflect the structural complexity of English sentence has been confirmed,which further implied the possibility of the application of pattern grammar to the description of sentence constructionThrough linking pattern grammar to the structural complexity of sentence and its variations across registers,on the basis of the NPF and MPL,the study preliminarily proposed a two-dimensional system of pattern measurement indicators for the description of sentence structure:The coverage of pattern grammar research has been enlarged from the word level to the sentence level.It has also evoked potential application significance in both NLP and language testing areas due to the consideration of the characteristics of patterns themselves and pattern configurations in sentences,and the concise and holistic property of the system of measurement indicators.