Pre-processing of bilingual corpora for Mandarin-English EBMT

Generation of Noun-Noun Compounds in the Spanish-English Machine Translation System SPANAM Julia Aymerich Pan American Health Organization 525 23rd Street, N.W. Washington, DC 20037 (USA) aymericj@paho.org Abstract The translation of Spanish Noun + preposition + Noun (NPN) constructions into English Noun-Noun (NN) compounds in many cases produces output with a higher level of fluency than if the NPN ordering is preserved. However, overgeneration of NN compounds can be dangerous because it may introduce ambiguity in the translation. This paper presents the strategy implemented in SPANAM to address this issue. The strategy involves dictionary coding of key words and expressions that allow or prohibit NN formation as well as an algorithm that generates NN compounds automatically when no dictionary coding is present. Certain conditions specified in the algorithm may also override the dictionary coding. The strategy makes use of syntactic and lexical information. No semantic coding is required. The last step in the strategy involves post-editing macros that allow the posteditor to quickly create or undo NN compounds if SPANAM did not generate the desired result. Keywords Noun-Noun compounds, Spanish-English machine translation and SPANAM system SPANAM today SPANAM has been operational at the Pan American Health Organization since 1980. The program, along with its English-Spanish counterpart Engspan, was ported from the mainframe to the PC in 1992 and then to the Windows environment in 2000 (León, 2000). The basic architecture of the program hasn't changed since 1985: it is a transfer system with an ATN that generates a top-down, left-to-right sequential parse with chronological and explicit backtracking (Vasconcellos and León, 1988; Amores, 1996). Dictionary entries are rich morphological, syntactic, and semantic information. Thein grammar and dictionaries have grown considerably in the past 15 years. SPANAM dictionaries currently contain over 95,000 source entries and 85,000 translations. Source entries include single words (65,000 stem forms), multiple word entries or SUs1 (11,000), analysis rules or AUs (8,500), and context-sensitive translation rules (10,000). In several MT evaluation experiments sponsored by the Advanced Research Projects Agency (ARPA/SISTO) between 1992 and 1994, where 3 research systems and 5 commercial systems were evaluated for adequacy, comprehension, and fluency, SPANAM consistently received the highest scores for all three criteria. The fluency scores were slightly lower than the other two. The problem Spanish makes extensive use of NPN constructions (abuso de sustancias, abastecimiento de agua) but does not use NN compounds. If the NPN structure is maintained in the English noun phrase, the resulting translation, although understandable, is usually too literal (supply water/water supply, abuse of subtances/substance abuse),of especially when several nouns are involved. Compare center of control of hurricanes with hurricane control center. Adequacy and comprehension may be acceptable, but fluency suffers. 1 The concept of SU and AU in the PAHO systems is explained below. A robust Spanish-English MT program should be able to generate NN compounds in English from NPN constructions in Spanish. It is crucial, however, that MT not overgenerate NN compounds and that it not create ambiguity or change the meaning. It is preferable to …

Content truncated to 500 words