This paper proposes a new lexical selection algorithm for the utterance production from mental lexicon. Until now, lexical selection concerned with the state of mind was described as reference process based on information which was decided come up to conception. The algorithm in this study is based on multivariate numerical data processing and paid attention to context-dependency and waver-traits characteristics of human emotions. In the computation process on the point of synonyms or emotional contexts, and the performance of this system will have autonomous sense achieving much organic performance. Moreover, we have implemented this algorithm and made an evaluation. As the conclusion, we got that our new algorithm has much more organic and autonomous lexical selection behavior more than former models.
展开▼