Multiple Instance Learning Networks for Fine-Grained Sentiment Analysis

Stefanos Angelidis, Maria Lapata

Research output: Contribution to journalArticlepeer-review

Abstract

We consider the task of fine-grained sentiment analysis from the perspective of multiple instance learning (MIL). Our neural model is trained on document sentiment labels, and learns to predict the sentiment of text segments, i.e. sentences or elementary discourse units (EDUs), without segment-level supervision. We introduce an attention-based polarity scoring method for identifying positive and negative text snippets and a new dataset which we call SPOT (as shorthand for Segment-level POlariTy annotations) for evaluating MIL-style sentiment models like ours. Experimental results demonstrate superior performance against multiple baselines, whereas a judgement elicitation study shows that EDU-level opinion extraction produces more informative summaries than sentence-based alternatives.
Original languageEnglish
Pages (from-to)17-32
Number of pages16
JournalTransactions of the Association for Computational Linguistics
Volume6
Publication statusPublished - 1 Jan 2018

Fingerprint

Dive into the research topics of 'Multiple Instance Learning Networks for Fine-Grained Sentiment Analysis'. Together they form a unique fingerprint.

Cite this