Sentence Modeling via Multiple Word Embeddings and Multi-level Comparison for Semantic Textual Similarity

TL;DR

This paper introduces M-MaxLSTM-CNN, a novel model that employs multiple word embeddings and multi-level comparison to improve semantic textual similarity assessment, outperforming state-of-the-art methods without relying on handcrafted features or uniform embedding dimensions.

Contribution

The paper proposes a new sentence modeling approach using multiple word embeddings and multi-level comparison, enhancing semantic similarity evaluation without handcrafted features.

Findings

Outperforms state-of-the-art methods on STS Benchmark and SICK datasets.

Does not require handcrafted features or uniform embedding dimensions.

Shows strong performance in paraphrase detection and textual entailment tasks.

Abstract

Different word embedding models capture different aspects of linguistic properties. This inspired us to propose a model (M-MaxLSTM-CNN) for employing multiple sets of word embeddings for evaluating sentence similarity/relation. Representing each word by multiple word embeddings, the MaxLSTM-CNN encoder generates a novel sentence embedding. We then learn the similarity/relation between our sentence embeddings via Multi-level comparison. Our method M-MaxLSTM-CNN consistently shows strong performances in several tasks (i.e., measure textual similarity, identify paraphrase, recognize textual entailment). According to the experimental results on STS Benchmark dataset and SICK dataset from SemEval, M-MaxLSTM-CNN outperforms the state-of-the-art methods for textual similarity tasks. Our model does not use hand-crafted features (e.g., alignment features, Ngram overlaps, dependency features) as well as does not require pre-trained word embeddings to have the same dimension.