Abstract
Named entities make up a bulk of documents. Extracting named entities is crucial to various applications of natural language processing. Although efforts to identify named entities within monolingual documents are numerous, aligning named entities in bilingual documents has not been investigated extensively due to the complexity of the task. In this dissertation, we introduce statistical phrase translation and transliteration models to align bilingual named entities in parallel corpora. In our approach, we model the process of translating an English named entity phrase into a Chinese equivalent using lexical translation/transliteration probabilities for word translation and alignment probabilities for word reordering. The method involves automatically learning phrase position alignment and acquiring word translation from a bilingual phrase dictionary and parallel corpora, and automatically discovering transliteration transformations from a training set of name-transliteration pairs. Unlike previous approaches, the proposed transliteration model does not involve the use of either a bilingual pronunciation dictionary for converting source words into phonetic symbols or manually assigned phonetic similarity scores between source and target words. The method for aligning bilingual named entities also involves language-specific knowledge functions, including abbreviation handling, Chinese person name recognition, and acronym expansion. At run time, the proposed models are applied to each source named entity in a pair of bilingual sentences to generate and evaluate the target named entity candidates, and the source and target named entities are aligned based on the computed probabilities. Experimental results demonstrate that the proposed approach, which integrates statistical models with extra knowledge sources, is highly feasible and offers significant improvement in performance. The proposed methodology is applicable to a wide range natural language processing, such as machine translation, cross-language information retrieval, and bilingual lexicon acquisition. Finally, we conclude the proposed approach with an emphasis on the main contributions of aligning bilingual named entities and some directions on future work.