Norm It! Lexical Normalization for Italian and Its Downstream Effects for Dependency Parsing

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 610–614 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 610 MPDD: A Multi-Party Dialogue Dataset for Analysis of Emotions and Interpersonal Relationships Yi-Ting Chen1, Hen-Hsen Huang2,3, Hsin-Hsi Chen1,3 1Department of Computer Science and Information Engineering, National Taiwan University, Taipei, Taiwan 2Department of Computer Science, National Chengchi University, Taipei, Taiwan 3MOST Joint Research Center for AI Technology and All Vista Healthcare, Taipei, Taiwan ytchen@nlg.csie.ntu.edu.tw, hhhuang@nccu.edu.tw, hhchen@ntu.edu.tw Abstract A dialogue dataset is an indispensable resource for building a dialogue system. Additional information like emotions and interpersonal relationships labeled on conversations enables the system to capture the emotion flow of the participants in the dialogue. However, there is no publicly available Chinese dialogue dataset with emotion and relation labels. In this paper, we collect the conversions from TV series scripts, and annotate emotion and interpersonal relationship labels on each utterance. This dataset contains 25,548 utterances from 4,142 dialogues. We also set up some experiments to observe the effects of the responded utterance on the current utterance, and the correlation between emotion and relation types in emotion and relation classification tasks. Keywords: Multi-party dialogue, Interpersonal relationships, emotions, dataset 1. Introduction The social dialogue system and the task-oriented dialogue system are the two major types of dialogue systems. Unlike the latter, which needs to focus on confirming user’s purpose and usually targets a specific domain, the former is built in the open domain. The social dialogue system aims to understand both users’ intents and their feelings, and creates an appropriate response to the users. To design a dialogue system that understands the users’ feelings, a dataset with the labels of the users’ mental state is indispensable. However, those researches about the emotion detection are usually based on the datasets crawled from comment (Feng et al, 2010) and social media (Sun et al., 2010). There is few publicly available Chinese dataset containing conversation content and emotion labels similar to EmotionLines (Chen et al., 2018). In addition to the users’ emotion, we aim to investigate other factors that may affect the conversation. Interpersonal relationship between a speaker and a listener plays an important role. When people talk to friends, the wording would be much different from that talking to a stranger. If a dialogue system can predict which relation type the wording of a user belongs to, the system will be able to generate a better response. In this paper, we release a dataset, Multi-Party Dialogue Dataset (MPDD),1 with both emotion and relation labels on each utterance. All the dialogues are collected from TV series scripts. To the best of our knowledge, this dataset is very rare Chinese multi-party dialogue dataset with both emotion and relation labels. We also make some experiments to observe the effects of the context on the emotion and relation classification, and the correlation between emotion and relation types. 2. Related Work Several dialogue datasets have been released in recent years. Some focus on the …

Content truncated to 500 words