Construction and Performance Evaluation of Parallel Corpora in Chinese and English Cross-Language Information Retrieval
[Purpose/significance]Corpora is a very important data source for translation in the field of cross-language information re-trieval.In CLIR,the performance evaluation of corpus,translation and extraction of bilingual dictionaries and semantic disambiguation can meet the needs of people to acquire knowledge and information.[Method/process]This paper collects Chinese and English web pages from news websites such as the Wall Street Journal,the Financial Times and the Hong Kong Government,uses the open-source software HTML Parser to filter out non-text content,converts the format and finally generates XML files,builds the parallel corpus by itself,uses CL-LSI and TDS models,and evaluates its performance.[Result/conclusion]In the establishment of CLIR evaluation cor-pus,it is verified that the TDS model can fully and objectively extract semantic bilingual subject features of semantic association in the process of bilingual paired search,and the performance of CLIR will exceed the retrieval efficiency of CL-LSI model through bilingual paired search.[Innovation/limitation]Aiming at in-depth research on corpora,this paper proposes a cross-language information re-trieval model(TDS)based on dual space in parallel corpora,and collects Chinese and English corpus for a given topic respectively.The obtained keywords are applied to the TDS model,and the co-occurrence semantic information of bilingual terms is analyzed.Fi-nally,the goal of parallel corpus construction and performance evaluation is realized.The disadvantage is that when the number of bi-lingual topics is small,the accuracy of translation is low,and when the number of topics is gradually increasing,the accuracy of trans-lation is higher.
cross-language information retrievalparallel corporadual spaceTDS modelCL-LSI model
NETL
NSTL
万方数据
