AMORE-UPF at SemEval-2018 Task 4: BiLSTM with Entity Library

TL;DR

This paper presents a simple BiLSTM model with an innovative entity library that significantly improves character identification in dialogues, especially for infrequent characters, demonstrating its potential for tasks with sparse data.

Contribution

The introduction of an entity library within a BiLSTM model for character identification in dialogues is a novel approach that enhances performance on infrequent characters.

Findings

Entity library greatly improves identification of infrequent characters

Model achieves winning results at SemEval-2018 Task 4

Potential applicability to other sparse data tasks

Abstract

This paper describes our winning contribution to SemEval 2018 Task 4: Character Identification on Multiparty Dialogues. It is a simple, standard model with one key innovation, an entity library. Our results show that this innovation greatly facilitates the identification of infrequent characters. Because of the generic nature of our model, this finding is potentially relevant to any task that requires effective learning from sparse or unbalanced data.