Korean to English TV Caption Translator: ""CaptionEye/KE""

In this paper, we present CaptionEye/KE, a Korean to English machine translation system that is applied to a practical TV caption translation. And its experimental evaluation is performed on actual TV news caption texts that are extracted by caption extractor. This system adopts a highly robust HMM (Hidden-Markov Model) POS (part-of-speech) tagger, a Korean dependency parser using verb phrase patterns and a pattern-based English generator. To overcome language-typical differences between Korean and English, we implement a data driven system that is based on a large quantity of verb phrase patterns. The current size of verb phrase patterns exceeds 80,000 entries. When the size of patterns grows up more and more, we can expect the quality of translation to improve. The result of experimental evaluation shows that the degree of understandability amounts up to 2.68 indicating that the meaning of translated English sentence is relatively clear to users.

• Type: Conference
• Year: 2001
• Where: NLPRS
• Authors: Seong-il Yang, Young Kil Kim, Young Ae Seo, Sung-Kwon Choi, Sang-Kyu Park
• Keywords: Korean Dependency Parser, Natural Language Processing, NLPRS 2001, TV Caption Translation, Verb Phrase Patterns