Exploring multiple features for sense prediction of Chinese unknown words

Word sense disambiguation is a crucial problem in natural language processing. While sense disambiguation of in-vocabulary words is well studied to date, few research findings are yet available concerning the prediction of unknown words' sense. In this paper, we attempt to exploit multiple features for predicting sense of Chinese out-of-vocabulary words in real text. To this end, we first take morpheme as the basic component units of Chinese words and thus investigate the relationship between Chinese unknown words' senses and their internal morphological structures. Then, we explore both word internal cues and word external contextual features, and combine them for sense prediction of Chinese unknown words using maximum entropy modeling. Our experimental results show that the incorporation of multiple features, especially the word-internal morphological features are of great value to Chinese unknown word sense prediction.