Revolutionizing Named Entity Recognition with Efficient Bidirectional Transformer Models

Author(s): Chien Vu

Originally published on Towards AI.

The light NER model outperforms both ChatGPT and fine-tuned LLMs in zero-shot evaluations on various NER benchmarks.
Image by author

Named Entity Recognition (NER) is an important task in Natural Language Processing (NLP) that involves identifying and classifying named entities, such as names of people, organizations, locations, etc., in text. Traditional NER models are effective but they have some limitations:

Predefined set of entities: It is necessary to have a predefined set of entity types such as person, location, and organization. The model does not have the ability to recognize new entities that do not exist in the training dataset.Waste resources: we need to train various models using domain-specific datasets, and we cannot use the trained models in a new, unrelated domain. This process requires significant resources for training from scratch and model storage 😔

Large Language Models (LLMs) now make it possible to easily extract different types of information from documents or text [1]. The problems of the traditional NER models are solved but other issues have arisen:

Size and Resource Limitations: Powerful LLMs typically have billions of parameters, making them computationally expensive and impractical in resource-limited scenarios. Accessing closed-source LLMs via APIs incurs high costs and leakage of internal data.Scalability: Autoregressive language models commonly used for NER can be slow due to token-by-token generation.​ Generating entities in multiple decoding… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI