Sentence Embeddings in NLI with Iterative Refinement Encoders

Research output: Contribution to journalArticleScientificpeer-review

Abstract

Sentence-level representations are necessary for various NLP tasks. Recurrent neural networks have proven to be very effective in learning distributed representations and can be trained efficiently on natural language inference tasks. We build on top of one such model and propose a hierarchy of BiLSTM and max pooling layers that implements an iterative refinement strategy and yields state of the art results on the SciTail dataset as well as strong results for SNLI and MultiNLI. We can show that the sentence embeddings learned in this way can be utilized in a wide variety of transfer learning tasks, outperforming InferSent on 7 out of 10 and SkipThought on 8 out of 9 SentEval sentence embedding evaluation tasks. Furthermore, our model beats the InferSent model in 8 out of 10 recently published SentEval probing tasks designed to evaluate sentence embeddings' ability to capture some of the important linguistic properties of sentences.
Original languageEnglish
JournalNatural Language Engineering
Publication statusAccepted/In press - 30 Jan 2019
MoE publication typeA1 Journal article-refereed

Cite this

@article{f56f5d09183f4a7d868d1fcd4a53e4c0,
title = "Sentence Embeddings in NLI with Iterative Refinement Encoders",
abstract = "Sentence-level representations are necessary for various NLP tasks. Recurrent neural networks have proven to be very effective in learning distributed representations and can be trained efficiently on natural language inference tasks. We build on top of one such model and propose a hierarchy of BiLSTM and max pooling layers that implements an iterative refinement strategy and yields state of the art results on the SciTail dataset as well as strong results for SNLI and MultiNLI. We can show that the sentence embeddings learned in this way can be utilized in a wide variety of transfer learning tasks, outperforming InferSent on 7 out of 10 and SkipThought on 8 out of 9 SentEval sentence embedding evaluation tasks. Furthermore, our model beats the InferSent model in 8 out of 10 recently published SentEval probing tasks designed to evaluate sentence embeddings' ability to capture some of the important linguistic properties of sentences.",
author = "Talman, {Aarne Johannes} and Anssi Yli-Jyr{\"a} and J{\"o}rg Tiedemann",
year = "2019",
month = "1",
day = "30",
language = "English",
journal = "Natural Language Engineering",
issn = "1351-3249",
publisher = "Cambridge University Press",

}

Sentence Embeddings in NLI with Iterative Refinement Encoders. / Talman, Aarne Johannes; Yli-Jyrä, Anssi; Tiedemann, Jörg.

In: Natural Language Engineering, 30.01.2019.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Sentence Embeddings in NLI with Iterative Refinement Encoders

AU - Talman, Aarne Johannes

AU - Yli-Jyrä, Anssi

AU - Tiedemann, Jörg

PY - 2019/1/30

Y1 - 2019/1/30

N2 - Sentence-level representations are necessary for various NLP tasks. Recurrent neural networks have proven to be very effective in learning distributed representations and can be trained efficiently on natural language inference tasks. We build on top of one such model and propose a hierarchy of BiLSTM and max pooling layers that implements an iterative refinement strategy and yields state of the art results on the SciTail dataset as well as strong results for SNLI and MultiNLI. We can show that the sentence embeddings learned in this way can be utilized in a wide variety of transfer learning tasks, outperforming InferSent on 7 out of 10 and SkipThought on 8 out of 9 SentEval sentence embedding evaluation tasks. Furthermore, our model beats the InferSent model in 8 out of 10 recently published SentEval probing tasks designed to evaluate sentence embeddings' ability to capture some of the important linguistic properties of sentences.

AB - Sentence-level representations are necessary for various NLP tasks. Recurrent neural networks have proven to be very effective in learning distributed representations and can be trained efficiently on natural language inference tasks. We build on top of one such model and propose a hierarchy of BiLSTM and max pooling layers that implements an iterative refinement strategy and yields state of the art results on the SciTail dataset as well as strong results for SNLI and MultiNLI. We can show that the sentence embeddings learned in this way can be utilized in a wide variety of transfer learning tasks, outperforming InferSent on 7 out of 10 and SkipThought on 8 out of 9 SentEval sentence embedding evaluation tasks. Furthermore, our model beats the InferSent model in 8 out of 10 recently published SentEval probing tasks designed to evaluate sentence embeddings' ability to capture some of the important linguistic properties of sentences.

M3 - Article

JO - Natural Language Engineering

JF - Natural Language Engineering

SN - 1351-3249

ER -