The Natural Language Decathlon: Multitask Learning as Question Answering

TL;DR

The paper introduces decaNLP, a multitask question answering framework covering ten diverse NLP tasks, and proposes MQAN, a unified model that learns all tasks jointly without task-specific modules, achieving strong transfer and zero-shot learning capabilities.

Contribution

It presents a new multitask question answering paradigm and a unified neural network model that handles multiple NLP tasks simultaneously without task-specific components.

Findings

MQAN improves transfer learning for translation and NER.

MQAN enables domain adaptation for sentiment analysis and NLI.

MQAN achieves zero-shot classification and state-of-the-art results on semantic parsing.

Abstract

Deep learning has improved performance on many natural language processing (NLP) tasks individually. However, general NLP models cannot emerge within a paradigm that focuses on the particularities of a single metric, dataset, and task. We introduce the Natural Language Decathlon (decaNLP), a challenge that spans ten tasks: question answering, machine translation, summarization, natural language inference, sentiment analysis, semantic role labeling, zero-shot relation extraction, goal-oriented dialogue, semantic parsing, and commonsense pronoun resolution. We cast all tasks as question answering over a context. Furthermore, we present a new Multitask Question Answering Network (MQAN) jointly learns all tasks in decaNLP without any task-specific modules or parameters in the multitask setting. MQAN shows improvements in transfer learning for machine translation and named entity recognition, domain adaptation for sentiment analysis and natural language inference, and zero-shot capabilities for text classification. We demonstrate that the MQAN's multi-pointer-generator decoder is key to this success and performance further improves with an anti-curriculum training strategy. Though designed for decaNLP, MQAN also achieves state of the art results on the WikiSQL semantic parsing task in the single-task setting. We also release code for procuring and processing data, training and evaluating models, and reproducing all experiments for decaNLP.