词项搭配的定量分析方法

文章阐明了语言学家如何将信息理论中的互信息熵(MI值)和统计学及概率理论中的零假设T检验值(T值)实际应用于测量词项共现中两词项间的搭配强弱(Collocability)和判断搭配强弱的置信度(Confidence Level).并通过例词分析,说明了利用MI值和T值可以从大量的词项共现中分辨出高频典型搭配,阐明了MI值和T值所提供量化标准的客观性和科学性,并指出藉此得到的有关高频典型搭配资料,对于提高外语学习者的搭配意识和搭配知识、优化外语教学效率和效果,将是不可或缺的资源.词项搭配研究对于语义识别、词典编撰、教材编写、ESP大纲设计、比较语言学研究等也有较大的指导意义.

Quantitative Approaches to Collocation Analysis

On the basis of probability theory and information theory,this paper illustrates how two measures of significance, namely MI-score and T-score,are employed by linguists to measure the collocability and the confidence level of the association between word combinations.By means of case study,the paper demonstrates the objectivity of qualitative and quantitative evalua- tions of the approaches and of the results produced by these approaches in two corpora.The paper also points out the potential ap- plications assisted by the two measures in language research areas,so as to cultivate language learners' collocation awareness,to accumulate their collocation knowledge and to improve the efficiency and effectiveness of collocation teaching and learning.The data obtained by these measures are also of guidance in the research areas like semantic categorization,textbook edition,ESP syl- labus design,language production and comparative language studies.With an optimistic view to corpus-based computational lin- guistics,this paper indicates a more rigorous and revealing future for language learning and linguistic research.