International Journal of Computer Processing of Languages

We describe a syntactic transformation model based on the probabilistic context-free grammar. This model is trained by using bilingual corpus and a broad coverage parser of the source language. Then we present two methods to solve the word-order problem using the transformational model. The first method deals with this problem in the preprocessing phase. There is no reordering in the decoding phase. The second method employs the syntactic transformation model in the decoding phase for phrase reordering within chunks. Speed is an advantage of this method. We considered translation from English to Vietnamese and from English to French. Our experiments showed significant BLEU-score improvements in comparison with Pharaoh, a state-of-the-art phrase-based SMT system.

Most example-based machine translation (EBMT) systems handle their translation examples using some heuristic measures based on human intuition. However, these heuristic rules are usually hard to be effectively organized to scale to incorporate diverse features to cover more language phenomenon and large domains. In this paper, we use machine learning approach for EBMT model design instead of human intuition. Maximum entropy (ME) model is introduced in order to adequately incorporate different kinds of features inherited in the translation examples effectively. At the same time, a multi-dimensional feature space is formally constructed to include various features of different aspects. In the experiments, the proposed model shows significant performance improvement.

In this paper, we propose a Japanese question-answering (QA) system to answer contextual questions using a Japanese non-contextual QA system. The contextual questions usually contain reference expressions to refer to previous questions and their answers. We address the reference resolution in contextual questions by finding the interpretation of references so as to maximize the cohesion with knowledge, i.e., information source like document collections. We utilize the appropriateness of the answer candidate obtained from the non-contextual QA system as the degree of the cohesion. The experimental results show that the proposed method is effective to disambiguate the interpretation of contextual questions.

We propose a mutual information approach to identify feature-based opinion expressions in customer reviews. Associations between opinion words and product feature categories are built by this approach. With the association set, we can identify what product feature a review unit refers to, even in the condition without explicit appearance of feature words. It can also be used to judge the semantic relatedness between a feature word and opinion words in its context. Thus it helps to decide which opinion word should contribute its polarity to the review feature. We also introduce the construction of a polarity lexicon, which is applied to identify opinion expressions in reviews. Using the approach proposed in this paper, we supply the polarity lexicon with the related product feature information. With the resource, we can get a better opinion mining result for a product review from different feature aspects.

A large-scale sentence pattern dictionary (SP-dictionary) for Japanese compound and complex sentences has been developed. The dictionary has been compiled based on the non-compositional language model. Sentences with 2 or 3 predicates are extracted from a Japanese-to-English parallel corpus of 1 million sentences, and the compositional constituents contained within them are generalized to produce a SP-dictionary containing a total of 215,000 pattern pairs. In evaluation tests, the SP-dictionary achieved a syntactic coverage of 92% and a semantic coverage of 70%.

Chinese news texts often contain a number of abbreviations without explicitly defining their full-forms. Therefore, expanding abbreviations to their original full-forms plays an important role in improving accuracy of the information extraction and retrieval systems for Chinese. In this paper, we present a hybrid approach to automatic expansion of abbreviations in Chinese news texts. Generally, Chinese abbreviations are produced from their original full-forms via reduction, elimination or generalization. To ensure every abbreviation can successfully be expanded, each abbreviation under expansion is assumed to be created by these three methods, respectively. Based on this assumption, a mapping table between shortened words and their matrix words, and a dictionary of short-form/full-form pairs are used to generate all possible expansions for abbreviations. For an ambiguous abbreviation with mutiple expansion candidates, then hidden Markov models are employed to rank all its expansion candidates and select a proper one with the maximum score. In order to further improve expansion performance, some linguistic knowledge like discourse information and abbreviation patterns are utilized to correct possible expansion errors. Evaluation on an abbreviation-expanded corpus built from the Peking University Corpus showed that our approach can achieve 86.3% and 83.8% on average in precision and recall respectively for various types of abbreviations in Chinese news texts.

This paper presents an approach to build a novel two-level collocation net, which enables calculation of the collocation relationship between any two words, from a large raw corpus. The first level consists of atomic classes (each atomic class consists of one word and feature bigram), which are clustered into the second level class set. Each class in both levels is represented by its collocation candidate distribution, extracted from the linguistic analysis of the raw training corpus, over possible collocation relation types. In this way, all the information extracted from the linguistic analysis is kept in the collocation net. Our approach applies to both frequently and less-frequently occurring words by providing a clustering mechanism and resolve the data sparseness problem through the collocation net. Experimentation shows that the collocation net is efficient and effective in solving the data sparseness problem and determining the collocation relationship between any two words.

Deep parsing of Chinese sentences is a very challenging task due to their complexity such as ambiguous word boundaries and meanings. An alternative mode of Chinese language processing is to perform shallow parsing of Chinese sentences in which chunk segmentation plays an important role. In this paper, we present a chunk segmentation algorithm using a combined statistical and rule-based approach (CSRA). The decision rules for refining chunk segmentation are generated from incorrectly segmented chunks from a statistical model which is built on a training corpus. Experimental results show that the CSRA works well and produces satisfactory chunk segmentation results for subsequent processes such as chunk tagging and chunk collocation extraction.