Bilingual Character Representation for Efficiently Addressing Out-of-Vocabulary Words in Code-Switching Named Entity Recognition

TL;DR

This paper introduces a bilingual character-based LSTM model with transfer learning and data normalization techniques to improve named entity recognition in code-switching Twitter data, achieving high accuracy without external gazetteers.

Contribution

The paper presents a novel hierarchical LSTM model utilizing bilingual character representations and transfer learning for out-of-vocabulary words in code-switching NER tasks.

Findings

Achieved 62.76% harmonic mean F1-score on English-Spanish data.

Outperformed baseline models without using gazetteers.

Demonstrated effectiveness of token normalization and replacement techniques.

Abstract

We propose an LSTM-based model with hierarchical architecture on named entity recognition from code-switching Twitter data. Our model uses bilingual character representation and transfer learning to address out-of-vocabulary words. In order to mitigate data noise, we propose to use token replacement and normalization. In the 3rd Workshop on Computational Approaches to Linguistic Code-Switching Shared Task, we achieved second place with 62.76% harmonic mean F1-score for English-Spanish language pair without using any gazetteer and knowledge-based information.