A Morpho-Syntactically Informed LSTM-CRF Model for Named Entity Recognition

TL;DR

This paper introduces a morphologically informed LSTM-CRF model for Bulgarian named entity recognition, demonstrating that coarse POS tags significantly improve performance in morphologically rich languages.

Contribution

It shows that combining POS tags with word and character embeddings enhances NER, especially in morphologically rich languages, and that coarse POS tags are sufficient for substantial gains.

Findings

Significant performance improvements over state-of-the-art in Bulgarian NER

POS tags contribute more than detailed morphological info for NER

Coarse POS tags are effective for morphologically rich languages

Abstract

We propose a morphologically informed model for named entity recognition, which is based on LSTM-CRF architecture and combines word embeddings, Bi-LSTM character embeddings, part-of-speech (POS) tags, and morphological information. While previous work has focused on learning from raw word input, using word and character embeddings only, we show that for morphologically rich languages, such as Bulgarian, access to POS information contributes more to the performance gains than the detailed morphological information. Thus, we show that named entity recognition needs only coarse-grained POS tags, but at the same time it can benefit from simultaneously using some POS information of different granularity. Our evaluation results over a standard dataset show sizable improvements over the state-of-the-art for Bulgarian NER.