Rethinking Generalization of Neural Models: A Named Entity Recognition Case Study

TL;DR

This paper investigates the generalization capabilities of neural models in Named Entity Recognition, analyzing their limitations and proposing measures to guide better model design and training.

Contribution

It introduces new measures for assessing NER model generalization and provides in-depth analysis of existing models' bottlenecks and dataset biases.

Findings

Neural NER models have identifiable bottlenecks in generalization.

Dataset bias significantly affects model performance.

Proposed measures help diagnose and improve NER models.

Abstract

While neural network-based models have achieved impressive performance on a large body of NLP tasks, the generalization behavior of different models remains poorly understood: Does this excellent performance imply a perfect generalization model, or are there still some limitations? In this paper, we take the NER task as a testbed to analyze the generalization behavior of existing models from different perspectives and characterize the differences of their generalization abilities through the lens of our proposed measures, which guides us to better design models and training methods. Experiments with in-depth analyses diagnose the bottleneck of existing neural NER models in terms of breakdown performance analysis, annotation errors, dataset bias, and category relationships, which suggest directions for improvement. We have released the datasets: (ReCoNLL, PLONER) for the future research at our project page: http://pfliu.com/InterpretNER/. As a by-product of this paper, we have open-sourced a project that involves a comprehensive summary of recent NER papers and classifies them into different research topics: https://github.com/pfliu-nlp/Named-Entity-Recognition-NER-Papers.