DeepAI AI Chat
Log In Sign Up

Explicit Pairwise Word Interaction Modeling Improves Pretrained Transformers for English Semantic Similarity Tasks

by   Yinan Zhang, et al.

In English semantic similarity tasks, classic word embedding-based approaches explicitly model pairwise "interactions" between the word representations of a sentence pair. Transformer-based pretrained language models disregard this notion, instead modeling pairwise word interactions globally and implicitly through their self-attention mechanism. In this paper, we hypothesize that introducing an explicit, constrained pairwise word interaction mechanism to pretrained language models improves their effectiveness on semantic similarity tasks. We validate our hypothesis using BERT on four tasks in semantic textual similarity and answer sentence selection. We demonstrate consistent improvements in quality by adding an explicit pairwise word interaction module to BERT.


page 1

page 2

page 3

page 4


Temporal Attention for Language Models

Pretrained language models based on the transformer architecture have sh...

Character-based Neural Networks for Sentence Pair Modeling

Sentence pair modeling is critical for many NLP tasks, such as paraphras...

Screen2Vec: Semantic Embedding of GUI Screens and GUI Components

Representing the semantics of GUI screens and components is crucial to d...

Integrating Knowledge Graph embedding and pretrained Language Models in Hypercomplex Spaces

Knowledge Graphs, such as Wikidata, comprise structural and textual know...

Improving word mover's distance by leveraging self-attention matrix

Measuring the semantic similarity between two sentences is still an impo...

DABERT: Dual Attention Enhanced BERT for Semantic Matching

Transformer-based pre-trained language models such as BERT have achieved...