作者: Naho Orita , Eliana Vornov , Naomi Feldman , Hal Daumé III
DOI: 10.3115/V1/P15-1158
关键词: Language production 、 Linguistics 、 Computer science 、 Speech act 、 Word (computer architecture)
摘要: We propose a language production model that uses dynamic discourse information to account for speakers’ choices of referring expressions. Our extends previous rational speech act models (Frank and Goodman, 2012) more naturally distributed linguistic data, instead assuming controlled experimental setting. Simulations show close match between utterances predictions, indicating behavior can be modeled in principled way by considering the probabilities referents conveyed each word.