Subword Mapping and Anchoring across Languages

TL;DR

SMALA is a novel method for constructing bilingual subword vocabularies that improves cross-lingual transfer and translation quality by addressing false positives and negatives in shared vocabularies.

Contribution

It introduces an unsupervised approach to align subwords across languages, enhancing multilingual models without requiring task-specific data.

Findings

Improves zero-shot transfer in cross-lingual NLP tasks.

Increases BLEU scores in neural machine translation.

Reduces false positives and negatives in shared vocabularies.

Abstract

State-of-the-art multilingual systems rely on shared vocabularies that sufficiently cover all considered languages. To this end, a simple and frequently used approach makes use of subword vocabularies constructed jointly over several languages. We hypothesize that such vocabularies are suboptimal due to false positives (identical subwords with different meanings across languages) and false negatives (different subwords with similar meanings). To address these issues, we propose Subword Mapping and Anchoring across Languages (SMALA), a method to construct bilingual subword vocabularies. SMALA extracts subword alignments using an unsupervised state-of-the-art mapping technique and uses them to create cross-lingual anchors based on subword similarities. We demonstrate the benefits of SMALA for cross-lingual natural language inference (XNLI), where it improves zero-shot transfer to an unseen language without task-specific data, but only by sharing subword embeddings. Moreover, in neural machine translation, we show that joint subword vocabularies obtained with SMALA lead to higher BLEU scores on sentences that contain many false positives and false negatives.