PRIDE: Predicting Relationships in Conversations

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 4636–4650 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 4636 PRIDE: Predicting Relationships in Conversations Anna Tigunova, Paramita Mirza, Andrew Yates, Gerhard Weikum Max Planck Institute for Informatics Saarbrücken, Germany {tigunova, paramita, ayates, weikum}@mpi-inf.mpg.de Abstract Automatically extracted interpersonal relation- ships of conversation interlocutors can enrich personal knowledge bases to enhance person- alized search, recommenders and chatbots. To infer speakers’ relationships from dialogues we propose PRIDE, a neural multi-label clas- sifier, based on BERT and Transformer for cre- ating a conversation representation. PRIDE utilizes the dialogue structure and augments it with external knowledge about speaker fea- tures and conversation style. Unlike prior works, we address multi-label prediction of fine-grained relationships. We release large- scale datasets, based on screenplays of movies and TV shows, with directed relationships of conversation participants. Extensive experi- ments on both datasets show superior perfor- mance of PRIDE compared to the state-of-the- art baselines. 1 Introduction Motivation and Problem. Personal knowledge about individual users is a valuable asset for per- sonalizing downstream applications, such as intel- ligent assistants, recommender systems and search engines. However, such personalized services are commonly achieved with end-to-end learning ap- proaches, where user information is bound to be in latent representation and inaccessible to users. Explicit Personal Knowledge Bases (PKBs) (Balog and Kenter, 2019), which are built independently of any downstream application, serve as background knowledge for personalization. PKBs are crucial for empowering users with control over what can be learned from their data collected by big tech com- panies. Such PKBs will also provide transparency and explainability to end users about inferred per- sonal knowledge and any personalized decisions made by the systems. With the ubiquity of social media and online forums, user-generated content is available in abun- dance. Mining personal knowledge from user- User A 2 h tough day.. can't wait for a relaxed weekend 2 comments User B 1 h but.. dad said we'll go somewhere exciting! User A 1 h of course, sweetheart, it's your birthday after all ;) but make sure to finish your homework first! User B 3 h Yallll @justinbieber just announced World Tour 2022!! Super excited! #Belieber 15 comments User B 10 h Decided to post this recording :D #music #singersongwriter #13yearoldgirl #guitar #sing #singing #talent #musicians #RT #retweet #like 27 comments Figure 1: Example of conversation in social media. generated content to populate PKBs, or user profil- ing, is a long-standing topic in NLP (e.g., Flekova et al., 2016; Basile et al., 2017; Tigunova et al., 2019). While users’ demographic attributes and interests can be learned from their profile descrip- tions and posts, interpersonal relationships with other users are rarely mentioned explicitly and may only be inferred from their interactions and con- versations. In this work, we develop an automatic method for predicting fine-grained relationships between two speakers, given their logged conversa- tion history. Consider the example in Figure 1. From the ex- cerpt of interactions between A and B, the reader …

Content truncated to 500 words