Evaluation of Japanese-Chinese MT System using AAMT’s Test-Set
نویسندگان
چکیده
The methods for evaluating the quality of a machine translation (MT) system have two major problems. One is the establishment of criteria which are independent of evaluator objectivity and the other is reducing man-hour costs needed to evaluate huge sets of translation sentences. With these aims, JEIDA (Japan Electronic Industry Development Association) had developed JEIDA’s Test-Sets [1] which are applied to English-Japanese and Japanese-English translation evaluation. One of the features of their Test-Sets is translation examples that include grammatical checkpoints, which are answered “Yes” or “No” by evaluators. Compared with conventional evaluating methods, such as grading translation on its quality into 3 or 5 levels, the method based on answering yes/no questions is more objective. Since 2007, we have been developing the new Test-Set for evaluating quality of Japanese-Chinese MT systems, expanding JEIDA’s Test-Sets [2]. In June 2009, our new Test-Set was completed and made publicly available [3]. The work described in this paper has been developed by the Special Working Group of AAMT (Asia-Pacific Association for Machine Translation) (we call our Test-Set “AAMT’s Test-Set”). In 2008, we evaluated the quality of several actual Japanese-Chinese MT systems, applying AAMT’s Test-Set. In order to compare with conventional methods, we also carried out subjective evaluation focused on fluency and adequacy. In this paper we would like to describe the merits of the evaluation method using AAMT’s Test-Set comparing with conventional methods. In Section 2, we will overview AAMT’s Test-Set showing examples of yes/no questions. Section 3 describes the details of evaluation experiments we carried out. Chapter 4 describes variability of evaluation results depending on evaluators, and analyzes correlation among evaluating methods such as fluency evaluation, adequacy evaluation, yes/no evaluation and statistical evaluation (BLEU)[4]. Moreover, we will briefly discuss strong and weak points of MT systems, comparing Japanese-Chinese and Japanese-English MT systems.
منابع مشابه
Exploiting Shared Chinese Characters in Chinese Word Segmentation Optimization for Chinese-Japanese Machine Translation
Unknown words and word segmentation granularity are two main problems in Chinese word segmentation for ChineseJapanese Machine Translation (MT). In this paper, we propose an approach of exploiting common Chinese characters shared between Chinese and Japanese in Chinese word segmentation optimization for MT aiming to solve these problems. We augment the system dictionary of a Chinese segmenter b...
متن کاملUQAM's System Description for the NTCIR-10 Japanese and English PatentMT Evaluation Tasks
This paper describes the development of a Japanese-English and English-Japanese translation system for the NTCIR-10 Patent MT tasks. The MT system is based on the provided training data and Moses decoder. We report our first attempt on statistical machine translation for these pairs of languages and the Patent domain.
متن کاملGeneration for Multilingual MT
This paper presents an overview of the broad-coverage, application-independent natural language generation component of the NLP system being developed at Microsoft Research. It demonstrates how this component functions within a multilingual Machine Translation system (MSR-MT), using the languages that we are currently working on (English, Spanish, Japanese, and Chinese). Section 1 provides a sy...
متن کاملMSR-MT: The Microsoft Research Machine Translation System
MSR-MT is a data-driven MT system that combines rule-based and statistical techniques with example-based transfer. This hybrid, large-scale system is capable of learning all its knowledge of lexical and phrasal translations directly from data. MSR-MT has undergone rigorous evaluation showing that, trained on a corpus of technical data similar to the test corpus, its output surpasses the quality...
متن کاملEvaluating Machine Translation Performance on Chinese Idioms with a Blacklist Method
Idiom translation is a challenging problem in machine translation because the meaning of idioms is non-compositional, and a literal translation is likely to be wrong. In this paper, we assess the quality of idiom translation of a modern neural MT system. We introduce a new evaluation method based on an idiom-specific blacklist of literal translations, based on the insight that the occurrence of...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2009