CHARM: Inferring personal attributes from conversations

Abstract

Personal knowledge about users' professions, hobbies, favorite food, and travel preferences, among others, is a valuable asset for individualized AI, such as recommenders or chatbots. Conversations in social media, such as Reddit, are a rich source of data for inferring personal facts. Prior work developed supervised methods to extract this knowledge, but these approaches can not generalize beyond attribute values with ample labeled training samples. This paper overcomes this limitation by devising CHARM: a zero-shot learning method that creatively leverages keyword extraction and document retrieval in order to predict attribute values that were never seen during training. Experiments with large datasets from Reddit show the viability of CHARM for open-ended attributes, such as professions and hobbies.