The code for our paper "NSP-BERT: A Prompt-based Zero-Shot Learner Through an Original Pre-training Task —— Next Sentence Prediction"

Overview

This is the code of our paper NSP-BERT: A Prompt-based Zero-Shot Learner Through an Original Pre-training Task —— Next Sentence Prediction. We use a sentence-level pre-training task NSP (Next Sentence Prediction) to realize prompt-learning and perform various downstream tasks, such as single sentence classification, sentence pair classification, coreference resolution, cloze-style task, entity linking, entity typing.

On the FewCLUE benchmark, our NSP-BERT outperforms other zero-shot methods (GPT-1-zero and PET-zero) on most of these tasks and comes close to the few-shot methods. We hope NSP-BERT can be an unsupervised tool that can assist other language tasks or models.

Guide

Environment

The environments are as follows:

Python 3.6
bert4keras 0.10.6
tensorflow-gpu 1.15.0

Downloads

Models

We should dowmload the checkpoints of different models. The vocab.txt and the config.json are already in our repository.

* We need to use UER's convert tool to convert UER pytorch to Original Tensorflow.

Datasets

We use FewCLUE datasets and DuEL2.0 (CCKS2020) in our experiments.

Put the datasets into the NSP-BERT/datasets/.

Use examples

We can run individual python files in the project directly to evaluate our NSP-BERT.

NSP-BERT
    |- datasets
        |- clue_datasets
           |- ...
        |- DuEL 2.0
           |- dev.json
           |- kb.json
    |- models
        |- uer_mixed_corpus_bert_base
           |- bert_config.json
           |- vocab.txt
           |- bert_model.ckpt...
           |- ...
    |- nsp_bert_classification.py             # Single Sentence Classification
    |- nsp_bert_sentence_pair.py              # Sentence Pair Classification
    |- nsp_bert_cloze_style.py                # Cloze-style Task
    |- nsp_bert_coreference_resolution.py     # Coreference Resolution
    |- nsp_bert_entity_linking.py             # Entity Linking and Entity Typing
    |- utils.py

Baselines

Reference FewCLUE, we choos 3 training scenarios, fine-tuning, few-shot and zero-shot. The baselines use Chineses-RoBERTa-Base and Chinses-GPT-1 as the backbone model.

Methods

Downloads

Model Comparison

Strategy Details

Discussion

• Sincce NSP-BERT is a sentence-level prompt-learning model, it is significantly superior to GPT-zero and PET-zero in terms of Single Sentence Classification tasks (TNEWS, CSLDCP and IFLYTEK). At the same time, it can solve the Entity Linking task (DuEL2.0), and the model is not limited by the non-fixed-length entity description, which GPT-zero and PET-zero cannot do this.
• However, it doesn't work as well on Token-Level tasks, such as Cloze-style task and Entity Typing.
• In future work, it is essential to extend NSP-BERT to the few-shot scenario.