Shiho Ogino

An interactive Japanese parser for machine translation

In this paper, we describe a working system for interactive Japanese syntactic analysis. A human user can intervene during parsing to help the system to produce a correct parse tree. Human interactions are limited to the very simple task of indicating the modifiee (governor) of a phrase, and thus a non-expert native speaker can use the system. The user is free to give any information in any order, or even to provide no information. The system is being used as the source language analyzer of a Japanese-to-English machine translation system currently under development.

A stochastic parser based on a structural word prediction model

In this paper, we present a stochastic language model using dependency. This model considers a sentence as a word sequence and predicts each word from left to right. The history at each step of prediction is a sequence of partial parse trees covering the preceding words. First our model predicts the partial parse trees which have a dependency relation with the next word among them and then predicts the next word from only the trees which have a dependency relation with the next word. Our model is a generative stochastic model, thus this can be used not only as a parser but also as a language model of a speech recognizer. In our experiment, we prepared about 1,000 syntactically annotated Japanese sentences extracted from a financial newspaper and estimated the parameters of our model. We built a parser based on our model and tested it on approximately 100 sentences of the same newspaper. The accuracy of the dependency relation was 89.9%, the highest accuracy level obtained by Japanese stochastic parsers.

Context-Based Text Mining for Insights in Long Documents

In this paper, we consider long documents and try to find differences between document collections. In the analysis of document collections such as project status reports or annual reports, each document and each sentence tend to be relatively long. Therefore, it can be difficult to derive insights by looking only for representative concepts in the selected document collection based on a divergence metric. In this paper, we propose an analysis approach based on contextual information. By extracting pairs of a topic word and a keyword and assessing their representativeness in the selected document collection, we are developing a method to extract insights from these long documents. Applying the proposed method for the analysis between the annual reports of bankrupt companies and those of sound companies, we were able to derive insights that could not be extracted with the conventional methods.

Technical Evaluation of MT Systems from the Developer's Point of View

機械翻訳システムの開発者がシステムの技術的評価を翻訳品質に注目して客観的に行う手法を開発した。評価過程の客観性と評価結果の解釈の客観性を維持するために、本手法では単なる評価用例文集ではなく、システムの出力を評価するための設問と、その設問がどのような言語現象を対象としているかについての解説とを各例文に付与したテストセットを用いている。各例文は基本的な言語現象と現在の機械翻訳システムにおいて処理が困難である言語現象のそれぞれを出来る限り網羅するように収集された。今回、英日機械翻訳システム、日英機械翻訳システムのそれぞれについての評価用テストセットを作成した。これらを用いて商用の機械翻訳システムでの評価実験を繰り返すことにより、機械翻訳システムの能力の差異を提示できることが示された。