Free CSS Website Template
SKOPE is an endeaver to develop Korean language processing engine which can be used in both speech and NLP applications. Currently, SKOPE is composed of 3 different parts: morphological analysis and tagging, syntax analysis, and semantic analysis.
This project presents a hybrid model for Korean part-of-speech tagging which integrates both statistical and rule-based method. The narrow windows provided by stochastic model cannot cover necessary lexical and long-distance dependencies for POS disambiguation in Korean. On the other hand, rule-based approaches are not accurate and flexible enough to new tag-sets and languages. To overcome the limitations of two independent approaches, hybrid statistical/rule-based architecture is presented for Korean part-of-speech tagging. Our automatic rule learner follows Eric Brill's style. It is flexible to new tag-sets and corpus, and needs small space for rule model. These rules handle relations of morphemes between Eojeols. Our experiments show that this hybrid approach is good for especially morphologically complex agglutinative languages such as Korean.
This project is to develop a probabilistic chart parsing method for Korean using categorial grammar formalism.
This project is to develop Korean semantic analyzer that can interleave syntax analysis and to generate QLFs (quasi-logical form) as semantic structures. We are also working on the Korean WSD problems using English WordNet and Korean-English bilingual dictionary. Korean semantic parser demo is currently not available.
We are developing information extraction system for continuing education domain.
SKOPE engine is being developed together with the possible applications such as natural language interface agent, automatic web indexing, text-to-speech systems, etc.