A Unified Tagging Solution: Bidirectional LSTM Recurrent Neural Network with Word Embedding

TL;DR

This paper presents a unified approach using Bidirectional LSTM RNNs with word embeddings for multiple tagging tasks, achieving near state-of-the-art results without task-specific features.

Contribution

The study introduces a task-independent BLSTM-RNN framework that simplifies tagging across different NLP tasks without specialized feature engineering.

Findings

Achieves near state-of-the-art performance in POS tagging, chunking, and NER.

Uses only one set of task-independent features and learned representations.

Eliminates the need for task-specific feature engineering.

Abstract

Bidirectional Long Short-Term Memory Recurrent Neural Network (BLSTM-RNN) has been shown to be very effective for modeling and predicting sequential data, e.g. speech utterances or handwritten documents. In this study, we propose to use BLSTM-RNN for a unified tagging solution that can be applied to various tagging tasks including part-of-speech tagging, chunking and named entity recognition. Instead of exploiting specific features carefully optimized for each task, our solution only uses one set of task-independent features and internal representations learnt from unlabeled text for all tasks.Requiring no task specific knowledge or sophisticated feature engineering, our approach gets nearly state-of-the-art performance in all these three tagging tasks.