文章基本信息

标题：Distinguishing Different Classes of Utterances - the UC-PT Corpus
本地全文：下载
作者：Mariana Gaspar Fernandes ; Cátia Dias ; Luísa Coheur 等
期刊名称：OASIcs : OpenAccess Series in Informatics
电子版ISSN：2190-6807
出版年度：2019
卷号：74
页码：1-8
DOI：10.4230/OASIcs.SLATE.2019.14
出版社：Schloss Dagstuhl -- Leibniz-Zentrum fuer Informatik
摘要：Conversational bots are being used in many scenarios and we can find them playing museum guides or providing customer support, for instance. These bots base their answers in specific information related with their domain of expertise, but there is general information, presented in each user request that, when properly identified, could also be useful for the agent to decide what to answer. As an example, if the user is asking a question or uttering a statement, the bot's action in its search for a response will probably differ. In this paper we present three corpora for the Portuguese language - the UC-PT corpus - that can be used to help conversational bots to distinguish: a) questions from non questions, b) yes-no-questions from other types of questions; and c) personal from non-personal questions. With this information, the agent can decide, for instance, not to answer, redirect the question to a persona chatbot or decide to answer it with a simple "yes", "no" or "maybe". In addition, we benchmark the classification process in these corpora. This corpora will be made publicly available.
关键词：Corpora; Questions; Conversational Agents; Portuguese Language