Sequence Classification with Human Attention (2018-10-01T00:00:00.000000Z)

TL;DR

Estimated human attention derived from eye-tracking corpora is used to regularize attention functions in recurrent neural networks and shows substantial improvements across a range of tasks, including sentiment analysis, grammatical error detection, and detection of abusive language.

Abstract

Learning attention functions requires large volumes of data, but many NLP tasks simulate human behavior, and in this paper, we show that human attention really does provide a good inductive bias on many attention functions in NLP. Specifically, we use estimated human attention derived from eye-tracking corpora to regularize attention functions in recurrent neural networks. We show substantial improvements across a range of tasks, including sentiment analysis, grammatical error detection, and detection of abusive language.

Authors

Anders Søgaard

14 papers

Joachim Bingel

2 papers

Marek Rei

6 papers

TL;DR

Abstract

Authors

References65 items

ZuCo, a simultaneous EEG and eye-tracking resource for natural sentence reading

Zero-Shot Sequence Labeling: Transferring Knowledge from Sentences to Tokens

Learning to Predict Readability Using Eye-Movement Data From Natives and Learners

Cognition-Cognizant Sentiment Analysis With Multitask Subjectivity Summarization Based on Annotators' Gaze Behavior

Pre-training Attention Mechanisms

Using Gaze Data to Predict Multiword Expressions

Auxiliary Objectives for Neural Error Detection Models

Learning Cognitive Features from Gaze Data for Sentiment and Sarcasm Classification using Convolutional Neural Network

Efficient Attention using a Fixed-Size Memory Representation

Semi-supervised Multitask Learning for Sequence Labeling

A Survey on Hate Speech Detection using Natural Language Processing

Identifying beneficial task relations for multi-task learning in deep neural networks

Cross-lingual Transfer of Correlations between Parts of Speech and Gaze Features

Are You a Racist or Am I Seeing Things? Annotator Influence on Hate Speech Detection on Twitter

Keystroke dynamics as signal for shallow syntactic parsing

Learning Sentence Representation with Guidance of Human Attention

Deep Multi-Task Learning with Shared Memory

Weakly Supervised Part-of-speech Tagging Using Eye-tracking Data

Leveraging Cognitive Features for Sentiment Analysis

Deep multi-task learning with low level tasks supervised at lower layers

Extracting token-level signals of syntactic processing from fMRI - with an application to PoS induction

Compositional Sequence Labeling Models for Error Detection in Learner Writing

Human Attention in Visual Question Answering: Do Humans and Deep Networks look at the same regions?

Hateful Symbols or Hateful People? Predictive Features for Hate Speech Detection on Twitter

Multilingual Part-of-Speech Tagging with Bidirectional Long Short-Term Memory Models and Auxiliary Loss

Improving sentence compression by learning to predict gaze

Transfer Learning for Low-Resource Neural Machine Translation

Multi-Way, Multilingual Neural Machine Translation with a Shared Attention Mechanism

Multi-task Sequence to Sequence Learning

The Human Kernel

Using reading behavior to predict grammatical functions

Reading metrics for estimating task efficiency with MT output

Multi-Task Learning for Multiple Language Translation

Reading behavior predicts syntactic categories

SemEval-2015 Task 10: Sentiment Analysis in Twitter

Looking hard: Eye tracking for detecting grammaticality of automatically compressed sentences

A Lexicon-based Approach for Hate Speech Detection

GloVe: Global Vectors for Word Representation

Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank

Unsupervised Dependency Parsing with Acoustic Cues

Sentiment analysis in Twitter

Making do with what we have: use your bootstraps.

Spectral Methods for Learning Multivariate Latent Tree Structure

Unsupervised Syntactic Chunking with Acoustic Cues: Computational Models for Prosodic Bootstrapping

A New Dataset and Method for Automatically Grading ESOL Texts

Inferring word relevance from eye-movements of readers

Low-cost gaze interaction: ready to deliver the promises

Data from eye-tracking corpora as evidence for theories of syntactic processing complexity

Nightmare at test time: robust learning by feature deletion

Integrating Probabilistic Extraction Models and Data Mining to Discover Relations and Patterns in Text

Fixation durations before word skipping in reading

Length, frequency, and predictability effects of words on eye movements in reading

Long Short-Term Memory

Eye Movements in Reading: Perceptual and Language Processes

Learning an Interactive Attention Policy for Neural Machine Translation

Using millions of emoji occurrences to pretrain any-domain models for detecting emotion, sentiment, and sarcasm

Alignment-Based Neural Machine Translation

Evaluation of Human Emotion from Eye Motions

Andrius Vilkauskas, Agne Paulauskaite-Taraseviciene, and Gintare Kersulyte-Raudone

Sentiment Analysis on Twitter

Eye movements in reading words and sentences

The dundee corpus

Can Relevance be Inferred from Eye Movements in Information Retrieval

BNC database and word frequency lists

On-line comprehension processes and eye movements in reading

Field of Study

Venue Information

Name

Type

URL

Alternate Names