:hot_pepper: R²SQL: "Dynamic Hybrid Relation Network for Cross-Domain Context-Dependent Semantic Parsing." (AAAI 2021)

Overview

R²SQL

The PyTorch implementation of paper Dynamic Hybrid Relation Network for Cross-Domain Context-Dependent Semantic Parsing. (AAAI 2021)

Requirements

The model is tested in python 3.6 with following requirements:

torch==1.0.0
transformers==2.10.0
sqlparse
pymysql
progressbar
nltk
numpy
six
spacy

All experiments on SParC and CoSQL datasets were run on NVIDIA V100 GPU with 32GB GPU memory.

  • Tips: The 16GB GPU memory may appear out-of-memory error.

Setup

The SParC and CoSQL experiments in two different folders, you need to download different datasets from [SParC | CoSQL] to the {sparc|cosql}/data folder separately. Another related data file could be download from EditSQL. Then, download the database sqlite files from [here] as data/database.

Download Pretrained BERT model from [here] as model/bert/data/annotated_wikisql_and_PyTorch_bert_param/pytorch_model_uncased_L-12_H-768_A-12.bin.

Download Glove embeddings file (glove.840B.300d.txt) and change the GLOVE_PATH for your own path in all scripts.

Download Reranker models from [SParC reranker | CoSQL reranker] as submit_models/reranker_roberta.pt

Usage

Train the model from scratch.

./sparc_train.sh

Test the model for the concrete checkpoint:

./sparc_test.sh

then the dev prediction file will be appeared in results folder, named like save_%d_predictions.json.

Get the evaluation result from the prediction file:

./sparc_evaluate.sh

the final result will be appeared in results folder, named *.eval.

Similarly, the CoSQL experiments could be reproduced in same way.


You could download our trained checkpoint and results in here:

Reranker

If your want train your own reranker model, you could download the training file from here:

Then you could train, test and predict it:

train:

python -m reranker.main --train --batch_size 64 --epoches 50

test:

python -m reranker.main --test --batch_size 64

predict:

python -m reranker.predict

Improvements

We have improved the origin version (descripted in paper) and got more performance improvements 🥳 !

Compare with the origin version, we have made the following improvements:

  • add the self-ensemble strategy for prediction, which use different epoch checkpoint to get final result. In order to easily perform this strategy, we remove the task-related representation in Reranker module.
  • remove the decay function in DCRI, we find that DCRI is unstable with decay function, so we let DCRI degenerate into vanilla cross attention.
  • replace the BERT-based with RoBERTa-based model for Reranker module.

The final performance comparison on dev as follows:

SParC CoSQL
QM IM QM IM
EditSQL 47.2 29.5 39.9 12.3
R²SQL v1 (origin paper) 54.1 35.2 45.7 19.5
R²SQL v2 (this repo) 54.0 35.2 46.3 19.5
R²SQL v2 + ensemble 55.1 36.8 47.3 20.9

Citation

Please star this repo and cite paper if you want to use it in your work.

Acknowledgments

This implementation is based on "Editing-Based SQL Query Generation for Cross-Domain Context-Dependent Questions" EMNLP 2019.

Owner
huybery
Understanding & Generating Language.
huybery
Practical Machine Learning with Python

Master the essential skills needed to recognize and solve complex real-world problems with Machine Learning and Deep Learning by leveraging the highly popular Python Machine Learning Eco-system.

Dipanjan (DJ) Sarkar 2k Jan 08, 2023
CoNLL-English NER Task (NER in English)

CoNLL-English NER Task en | ch Motivation Course Project review the pytorch framework and sequence-labeling task practice using the transformers of Hu

Kevin 2 Jan 14, 2022
This repository has a implementations of data augmentation for NLP for Japanese.

daaja This repository has a implementations of data augmentation for NLP for Japanese: EDA: Easy Data Augmentation Techniques for Boosting Performance

Koga Kobayashi 60 Nov 11, 2022
A collection of GNN-based fake news detection models.

This repo includes the Pytorch-Geometric implementation of a series of Graph Neural Network (GNN) based fake news detection models. All GNN models are implemented and evaluated under the User Prefere

SafeGraph 251 Jan 01, 2023
test

Lidar-data-decode In this project, you can decode your lidar data frame(pcap file) and make your own datasets(test dataset) in Windows without any hug

46 Dec 05, 2022
SimBERT升级版(SimBERTv2)!

RoFormer-Sim RoFormer-Sim,又称SimBERTv2,是我们之前发布的SimBERT模型的升级版。 介绍 https://kexue.fm/archives/8454 训练 tensorflow 1.14 + keras 2.3.1 + bert4keras 0.10.6 下载

317 Dec 23, 2022
The proliferation of disinformation across social media has led the application of deep learning techniques to detect fake news.

Fake News Detection Overview The proliferation of disinformation across social media has led the application of deep learning techniques to detect fak

Kushal Shingote 1 Feb 08, 2022
Chinese NewsTitle Generation Project by GPT2.带有超级详细注释的中文GPT2新闻标题生成项目。

GPT2-NewsTitle 带有超详细注释的GPT2新闻标题生成项目 UpDate 01.02.2021 从网上收集数据,将清华新闻数据、搜狗新闻数据等新闻数据集,以及开源的一些摘要数据进行整理清洗,构建一个较完善的中文摘要数据集。 数据集清洗时,仅进行了简单地规则清洗。

logCong 785 Dec 29, 2022
A spaCy wrapper of OpenTapioca for named entity linking on Wikidata

spaCyOpenTapioca A spaCy wrapper of OpenTapioca for named entity linking on Wikidata. Table of contents Installation How to use Local OpenTapioca Vizu

Universitätsbibliothek Mannheim 80 Jan 03, 2023
keras implement of transformers for humans

keras implement of transformers for humans

苏剑林(Jianlin Su) 4.8k Jan 03, 2023
Residual2Vec: Debiasing graph embedding using random graphs

Residual2Vec: Debiasing graph embedding using random graphs This repository contains the code for S. Kojaku, J. Yoon, I. Constantino, and Y.-Y. Ahn, R

SADAMORI KOJAKU 5 Oct 12, 2022
A python package to fine-tune transformer-based models for named entity recognition (NER).

nerblackbox A python package to fine-tune transformer-based language models for named entity recognition (NER). Resources Source Code: https://github.

Felix Stollenwerk 13 Jul 30, 2022
A simple Flask site that allows users to create, update, and delete posts in a database, as well as perform basic NLP tasks on the posts.

A simple Flask site that allows users to create, update, and delete posts in a database, as well as perform basic NLP tasks on the posts.

Ian 1 Jan 15, 2022
Unofficial Parallel WaveGAN (+ MelGAN & Multi-band MelGAN & HiFi-GAN & StyleMelGAN) with Pytorch

Parallel WaveGAN implementation with Pytorch This repository provides UNOFFICIAL pytorch implementations of the following models: Parallel WaveGAN Mel

Tomoki Hayashi 1.2k Dec 23, 2022
Python api wrapper for JellyFish Lights

Python api wrapper for JellyFish Lights The hope is to make this a pip installable package Current capabalilities: Connects to a local JellyFish Light

10 Dec 18, 2022
The PyTorch based implementation of continuous integrate-and-fire (CIF) module.

CIF-PyTorch This is a PyTorch based implementation of continuous integrate-and-fire (CIF) module for end-to-end (E2E) automatic speech recognition (AS

Minglun Han 24 Dec 29, 2022
Contains links to publicly available datasets for modeling health outcomes using speech and language.

speech-nlp-datasets Contains links to publicly available datasets for modeling various health outcomes using speech and language. Speech-based Corpora

Tuka Alhanai 77 Dec 07, 2022
Persian Bert For Long-Range Sequences

ParsBigBird: Persian Bert For Long-Range Sequences The Bert and ParsBert algorithms can handle texts with token lengths of up to 512, however, many ta

Sajjad Ayoubi 63 Dec 14, 2022
🦆 Contextually-keyed word vectors

sense2vec: Contextually-keyed word vectors sense2vec (Trask et. al, 2015) is a nice twist on word2vec that lets you learn more interesting and detaile

Explosion 1.5k Dec 25, 2022
Python bot created with Selenium that can guess the daily Wordle word correct 96.8% of the time.

Wordle_Bot Python bot created with Selenium that can guess the daily Wordle word correct 96.8% of the time. It will log onto the wordle website and en

Lucas Polidori 15 Dec 11, 2022