Learning to Parse Bilingual Sentences Using Bilingual Corpus and Monolingual CFG

Learning to Parse Bilingual Sentences Using Bilingual Corpus and Monolingual CFG Chung-Chi Huang1 and Jason S. Chang2 1Dept. of Information Systems and Application/Taiwan International Graduate Program, National Tsing Hua University, HsinChu, Taiwan, u901571@alumni.nthu.edu.tw 1Taiwan International Graduate Program, Academia Sinica, Nankang, Taiwan 2 Dept. of Computer Science, National Tsing Hua University, HsinChu, Taiwan, Jason.jschang@gmail.com Abstract We present a new method for learning to parse a bilingual sentence using Inversion Transduction Grammar trained on a parallel corpus and a monolingual treebank. The method produces a parse tree for a bilingual sentence, showing the shared syntactic structures of individual sentence and the differences of word order within a syntactic structure. The method involves estimating lexical translation probability based on a word-aligning strategy and inferring probabilities for CFG rules. At runtime, a bottom-up CYK-styled parser is employed to construct the most probable bilingual parse tree for any given sentence pair. We also describe an implementation of the proposed method. The experimental results indicate the proposed model produces word alignments better than those produced by Giza++, a state-of-the-art word alignment system, in terms of alignment error rate and F-measure. The bilingual parse trees produced for the parallel corpus can be exploited to extract bilingual phrases and train a decoder for statistical machine translation. 1. Introduction 1.1. Background The amount of information available in English on the Internet has grown exponentially for the past few years. Although a myriad of data are at our disposal, non-native speakers often find it difficult to wade through all of it since they may not be familiar with the terms or idioms being used in the texts. To ease the situation, a number of online machine translation (MT) systems such as SYSTRAN and Google Translate provide translation of source text on demand. Moreover, online dictionaries have mushroomed to provide access at any time and everywhere for second language learners. 1.2. Motivation MT systems and bilingual dictionary are designed to provide the services for non-English speakers or to ease learning difficulties for second language learners. Both require a lexicon which can be derived from aligning words in a parallel corpus. Furthermore, second language learners can benefit by learning from example sentences with translations. By looking at bilingual examples, we acquire knowledge of the usage and meaning of word in context. With word alignment result of a sentence pair, it is much easier to grab the essential 2 concepts of unfamiliar foreign words in a sentence pair. For instance, consider the English sentence “These factors will continue to play a positive role after its return” with its segmented Chinese translation “香港 回歸 後 這些 條件 將會 繼續 發揮 積極 作用” shown in Figure 1, where the solid dark lines are word alignment results of them and ,e f stand for two sentences in two languages , E F respectively. If we don’t know the usage of “play” in the sense of “perform,” in this example sentence pair with the help of word alignment, we would quickly understand such meaning and learn useful …

Content truncated to 500 words