Language models for name recognition in spanish spoken dialogue systems
Germán Tapia, Iván Meza, Luis A. Pineda · Research in Computing Science · 2010
Current advances on dialogue system require the development of language models for automatic speech recognition that are not only domain or task specific but also sub-task specific (e.g. name, age or price recognition). This paper presents a method for the creation of language models for name recognition at the greeting stage of a conversation in spoken Spanish. In particular, we focus on the introductory phrases (e.g., mi nombre es John/my name is John). The method relies in the construction of a corpus for names, and we study two variants; in one names are uniformly distributed; in the other, names are represented in a manner consistent with their frequency in web-searches.