Design And Implementation Of Bilingual Information Processing System For Machine Translation

Along with the advance of information society and spread of Internet, information's quantity, content, format, storage and transfer way are all changed greatly. Improvement of network information's retrieval performance and translation quality is essential to promote Chinese Information Advance Progress. Including key problems of information processing, machine translation, and artificial intelligence, the research is not only provided with academic achievements, but also important for practical application.Present information system always concerns on some problems and short of holistic consideration. Without former knowledge and experience, system often repeats same dispose, which wastes huge resource and hard to advance without knowledge accumulation. Limited in analysis depth and quality, existing information is poor in retrieval language and retrieval performance. Practical MT system almost takes single rule-based strategy, which could not solve all the problems and multiple strategy integrated is necessary. What's more, translation by analogy is still not practical.To solve above problems, Network Bi-Lingual Information Processing System (NBLIPS), which integrated network techniques, machine translation and information processing, is designed and implemented. On base of valid information storage and management, system affords multiple effective retrieval methods to meet different query demands from last user and machine translation. Bi-lingual knowledge is not only the basis for IR refinement, but also the knowledge resource for Example-Based Machine Translation (EBMT) after inner alignment.Based on independent format recognition process and open knowledge base, the system can describe and dispose the latest format strategy in time. To improve the text analysis quality, some important heuristic information is kept during the pure text extraction. On basis of deep content analysis and understanding, comprehensive text features are recorded to generate relative index.With the Inverted Index and Term Co-Exist Table, the full text retrieval is achieved dynamically according to the terms number and operator category. On Terms Conception Relating Table extracted from bi-lingual dictionary, the keyword retrieval is achieved by concept expanding, which is practical and can...