Cooperation between Transfer and Analysis in Example-Based Framework

Transfer-Driven Machine Translation (TDMT) is presented as a method which drives the translation processes according to the nature of the input. In TDMT, transfer knowledge is the central knowledge of translation, and various kinds aml levels of knowledge are cooperatively applied to input sentences. TDMT effectively utilizes an example-based framework for transfer and analysis knowledge. A consistent framework of examples makes the cooperation between transfer and analysis effective, and efficient translation is achieved. The TDMT prototype system, which translates Japanese spoken dialogs into English, has shown great promise. 1 I n t r o d u c t i o n Many applications dealing with spoken-language, such as automatic telephone interpreting system, need efficient and robust processing. The system must be capable of handling many idiomatic expressions and spoken-language-specific expressions which deviate from conventional grammar. Also, spoken-hmguage has both easy and difficult expressions to translate. In human translation when translating an easy sentence, tile translated result is produced quickly using only surface-level knowledge. When translating a complex sentence, a more elaborate process is performed, using syntactic, semantic, and contextual knowledge. Thus, many strategies and various levels of knowledge are used to effectively translate spoken-language. This paper proposes a method called Transfer-Driven Machine Translation (TDMT), which carries out efficient translation processing by driving the necessary translation processes according to the nature of the input sentence. An example-based framework can achieve quick processing and consistently describe knowledge. The integration of transfer and analysis in an example-based framework is i)roposed as a method for achieving TDMT. In this method, transfer and analysis proceed autonomously and cooperatively. Also, a well-balanced load on each process can be achieved by employing this integrated processing mechanism. Sectiou 2 explains the idea of TDMT. Section 3 explains distance calculation and transfer in an examplebased framework. Section 4 explains analysis in an example-based framework. Section 5 reports on the TDMT prototype system, and Section 6 reports on tile experimental results. The explanations in the following sections use Japanese-to-English translation. 2 Transfer-Driven Machine Translation TD/vlT performs efficient and robust spoken-language translation using various kinds of strategies to be able to treat diverse input. Its characteristics are explained in the following sub-sections. 2.1 Transfer-centered cooperation mechanism Translation is essentially converting a source language expression into a target language expression. In TDMT, transfer knowledge consists of various levels of bilingual information. It is the primary knowledge used to solve translation problems. The transfer module retrieves the necess~u-y transfer knowledge ranging from global unit like sentence structures to local unit like words. The retrieval and application of transfer knowledge are flexibly controlled depending on the knowledge necessary to translate the input. Basically, translation is performed by using transfer knowledge. A transfer module uti l izes analys is knowledge (syntactic/szmantic infonnation) which helps to apply transfer L,~owledge to some part of the input. And generation and context knowledge are utilized for producing correct translatiou result. In other words, TDMT prodaces translation results by utilizing these dift'cK,~ut kinds of knowledge cooperatively and by centering on transfer, and achieves efficient translation according to the nature of the input. ACTES DE COL1NG-92, NAN'rES, 23-28 ^otrr 1992 6 4 5 l'~oc, oF COLING-92, N^rcrl~s, AU•. 23-28, 1992 2.2 Utilization of example-based framework Transfer knowledge is the basic data which is used for totally controlling the translation process. Most of the transfer knowledge in TDMT is described by the example-based framework. An example-based framework is useful for cortsistenfly describing transfer knowledge. The essence of the example-based framework is the distance calculation. This framework achieves the best-match based on the distance between the input and provided examples, and selects the most plausible target expression from many candidates. The distance is calculated quickly because of its simple mechanism. Through providing examples, various kinds and levels of knowledge can be described in the example-based franlework. 2.3 Mult i level knowledge TDMT provides multi-level transfer knowledge, which correspoods to each translation strategy. In the transfer knowledge of the TDMT prototype system, there is string-, patternand grammar-level knowledge. TDMT achieves efficient translation by utilizing multilevel knowledge effectively according to the nature of input. Some conventional machine translation systems also provide multiple levels of transfer knowledge for idioms, syntax, semantics, and so on, and try to apply these levels of that knowledge in a fixed order to cover diverse input [Ikehara et al. 87]. However, this method proceeds with the analysis lot deciding which level of knowledge should be applied for any given input sentence in a fixed order, placing heavy load on the analysis module. Also, the knowledge description is ratber more complicated than that of the example-based framework. Therefore, the lrauslation of a simple sentence is not always quick because the system tries to cover all translation strategies. 3 E x a m p l e b a s e d T r a n s f e r TDMT utilizes distance calculation to determine the most plausible target expression and structure in transfer.