Information-Based Machine Translation

Information-based Machine Translation Keiko HORIGUCHI Spoken Language Technology, Sony US Research Laboratories 3300 Zanker Road San Jose, CA 95134 keiko@slt.sel.sony.com Abstract This paper describes an approach to Machine Translation that places linguistic information at its foundation. The difficulty of translation from English to Japanese is illustrated with data that shows the influence of various linguistic contextual factors. Next, a method for natural language transfer is presented that integrates translation examples (represented as typed feature structures with source-target indices) with linguistic rules and constraints. The method has been implemented, and the results of an evaluation are presented. Introduction High-quality automatic translation requires the disambiguation of common, highly ambiguous verbs, such as to have, to take, or to get. It also requires the correct handling of non-compositional, idiomatic expressions with varying degrees of “fixedness”. We view Machine Translation in terms of linguistic information represented as typed feature structures. By integrating translation information represented as example pairs with other types of linguistic information represented as rules, our approach extends the capabilities of current machine translation methods, and solves a number of key problems. 1. Linguistic Context for Translation In translating different words, phrases, and expressions, different types and amounts of information from the context need to be considered. (Only the sentential context is considered here.) So far, a systematic solution to this problem has not been found. This section illustrates the extent of this problem, and the remainder of this paper describes our approach. 1.1. Expressions with to have We examined the problem of translating the English main verb to have into Japanese. The verb to have was selected because it is quite common in colloquial English, yet forms a large variety of senses, collocations, and idioms. 615 different expressions containing the English verb to have were extracted from a 7000-sentence corpus from the “international travel” domain. Each English expression was manually translated into Japanese in the most general way possible. 1.2. Target-language Distinctions The most general translation for the construction “X have Y” in this domain was found to be �� ���� (X-ni Y-ga aru): The copy shop next door has a fax machine. �������������������� tonari-no kopiiya-ni fakkusu-ga arimasu. next-ATT copy shop-LOC fax-NOM exist Other translations are often necessary when the target language imposes finer semantic distinction on the state or on the action that is described. For example, if the object noun phrase refers to one or more human beings, the Japanese verb aru is changed into iru. Similarly, the word pet or a pet animal as the object noun phrase triggers the translation of to have as katteiru, a Japanese verb for keeping an animal as a pet : We have two sons. ��������� musuko-ga futari imasu. son-NOM two-CONTR exist Do you have pets? ���������������� anata-wa petto-wo katte-imasu-ka you-TOP pet-ACC keep-ST-Q Other examples of finer target-language distinctions include a symptom/disease as the object of to have. While many physical symptoms and minor diagnoses (e.g. pain, cavity, fever, allergy) use the default translation (X-ga aru), a serious illness or diagnosis is …

Content truncated to 500 words