|
--- |
|
license: afl-3.0 |
|
--- |
|
<p align="center"> |
|
<br> |
|
<img src="https://expressai-xlab.s3.amazonaws.com/rst/intro_rst.png" width="1000"/> |
|
<br> |
|
</p> |
|
|
|
# reStructured Pre-training (RST) |
|
official [repository](https://github.com/ExpressAI/reStructured-Pretraining), [paper](https://arxiv.org/pdf/2206.11147.pdf), [easter eggs](http://expressai.co/peripherals/emoji-eng.html) |
|
|
|
#### RST is a new paradigm for language pre-training, which |
|
* unifies **26** different types of signal from **10** data sources (Totten Tomatoes, Dailymail, Wikipedia, Wikidata, Wikihow, Wordnet, arXiv etc ) in the world structurally, being pre-trained with a monolithcal model, |
|
* surpasses strong competitors (e.g., T0) on **52/55** popular datasets from a variety of NLP tasks (classification, IE, retrieval, generation etc) |
|
* achieves superior performance in National College Entrance Examination **(Gaokao-English, 高考-英语)** achieves **40** points higher than the average scores made by students and 15 points higher than GPT3 with **1/16** parameters. In particular, Qin gets a high score of **138.5** (the full mark is 150) in the 2018 English exam |
|
|
|
In such a pre-training paradigm, |
|
* Data-centric Pre-training: the role of data will be re-emphasized, and model pre-training and fine-tuning of downstream tasks are viewed as a process of data storing and accessing |
|
* Pre-training over JSON instead of TEXT: a good storage mechanism should not only have the ability to cache a large amount of data but also consider the ease of access. |
|
|
|
|
|
## Model Description |
|
We release all models introduced in our [paper](https://arxiv.org/pdf/2206.11147.pdf), covering 13 different application scenarios. Each model contains 11 billion parameters. |
|
|
|
| Model | Description | Recommended Application |
|
| ----------- | ----------- |----------- | |
|
| **rst-all-11b** | **Trained with all the signals below except signals that are used to train Gaokao models** | **All applications below （specialized models are recommended first if high performance is preferred）** | |
|
| rst-fact-retrieval-11b | Trained with the following signals: WordNet meaning, WordNet part-of-speech, WordNet synonym, WordNet antonym, wikiHow category hierarchy, Wikidata relation, Wikidata entity typing, Paperswithcode entity typing | Knowledge intensive tasks, information extraction tasks,factual checker | |
|
| rst-summarization-11b | Trained with the following signals: DailyMail summary, Paperswithcode summary, arXiv summary, wikiHow summary | Summarization or other general generation tasks, meta-evaluation (e.g., BARTScore) | |
|
| rst-temporal-reasoning-11b | Trained with the following signals: DailyMail temporal information, wikiHow procedure | Temporal reasoning, relation extraction, event-based extraction | |
|
| rst-information-extraction-11b | Trained with the following signals: Paperswithcode entity, Paperswithcode entity typing, Wikidata entity typing, Wikidata relation, Wikipedia entity | Named entity recognition, relation extraction and other general IE tasks in the news, scientific or other domains| |
|
| rst-intent-detection-11b | Trained with the following signals: wikiHow goal-step relation | Intent prediction, event prediction | |
|
| rst-topic-classification-11b | Trained with the following signals: DailyMail category, arXiv category, wikiHow text category, Wikipedia section title | general text classification | |
|
| rst-word-sense-disambiguation-11b | Trained with the following signals: WordNet meaning, WordNet part-of-speech, WordNet synonym, WordNet antonym | Word sense disambiguation, part-of-speech tagging, general IE tasks, common sense reasoning | |
|
| rst-natural-language-inference-11b | Trained with the following signals: ConTRoL dataset, DREAM dataset, LogiQA dataset, RACE & RACE-C dataset, ReClor dataset, DailyMail temporal information | Natural language inference, multiple-choice question answering, reasoning | |
|
| rst-sentiment-classification-11b | Trained with the following signals: Rotten Tomatoes sentiment, Wikipedia sentiment | Sentiment classification, emotion classification | |
|
| rst-gaokao-rc-11b | Trained with multiple-choice QA datasets that are used to train the [T0pp](https://huggingface.co/bigscience/T0pp) model | General multiple-choice question answering| |
|
| rst-gaokao-cloze-11b | Trained with manually crafted cloze datasets | General cloze filling| |
|
| rst-gaokao-writing-11b | Trained with example essays from past Gaokao-English exams and grammar error correction signals | Essay writing, story generation, grammar error correction and other text generation tasks | |
|
|
|
|
|
|
|
## Have a try? |
|
```python |
|
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM |
|
|
|
tokenizer = AutoTokenizer.from_pretrained("XLab/rst-all-11b") |
|
model = AutoModelForSeq2SeqLM.from_pretrained("XLab/rst-all-11b") |
|
|
|
inputs = tokenizer.encode("TEXT: this is the best cast iron skillet you will ever buy. QUERY: Is this review \"positive\" or \"negative\"", return_tensors="pt") |
|
outputs = model.generate(inputs) |
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True, clean_up_tokenization_spaces=True)) |
|
``` |
|
|
|
## Data for reStructure Pre-training |
|
This dataset is a precious treasure, containing a variety of naturally occurring signals. Any downstream task you can think of (e.g., the college entrance exam mentioned in the RST paper) can benefit from being pre-trained on some of our provided signals. We spent several months collecting the following 29 signal types, accounting for a total of 46,926,447 data samples. We hope this dataset will be a valuable asset for everyone in natural language processing research. |
|
|
|
We provide collected signals through [DataLab](https://github.com/ExpressAI/DataLab). For efficiency, we only provide 50,000 samples at most for each signal type. If you want all the samples we collected, please fill this [form](https://docs.google.com/forms/d/e/1FAIpQLSdPO50vSdfwoO3D7DQDVlupQnHgrXrwfF3ePE4X1H6BwgTn5g/viewform?usp=sf_link). More specifically, we collected the following signals. |
|
|
|
###### We will be happy :smiley: to know if the resource is helpful for your work, and please cite our [work](https://github.com/ExpressAI/reStructured-Pretraining/blob/main/README.md#Bib) :blush: |
|
|
|
| Mine | Signal | #Sample | Use in DataLab | Some Applications | |
|
| --- | --- | --- | --- | --- | |
|
| [Rotten Tomatoes](https://www.rottentomatoes.com/) | (review, rating) | 5,311,109 | `load_dataset("rst", "rotten_tomatoes_sentiment")` | Sentiment classification | |
|
| [Daily Mail](https://www.dailymail.co.uk/home/index.html) | (text, category) | 899,904 | `load_dataset("rst", "daily_mail_category")`| Topic classification | |
|
| [Daily Mail](https://www.dailymail.co.uk/home/index.html) | (title, text, summary) | 1,026,616 | `load_dataset("rst", "daily_mail_summary")` | Summarization; Sentence expansion| |
|
| [Daily Mail](https://www.dailymail.co.uk/home/index.html) | (text, events) | 1,006,412 | `load_dataset("rst", "daily_mail_temporal")` | Temporal reasoning| |
|
| [Wikidata](https://www.wikidata.org/wiki/Wikidata:Main_Page) | (entity, entity_type, text) | 2,214,274 | `load_dataset("rst", "wikidata_entity")` | Entity typing| |
|
| [Wikidata](https://www.wikidata.org/wiki/Wikidata:Main_Page) | (subject, object, relation, text) | 1,526,674 | `load_dataset("rst", "wikidata_relation")` | Relation extraction; Fact retrieval| |
|
| [wikiHow](https://www.wikihow.com/Main-Page) | (text, category) | 112,109 | `load_dataset("rst", "wikihow_text_category")` | Topic classification | |
|
| [wikiHow](https://www.wikihow.com/Main-Page) | (low_category, high_category) | 4,868 | `load_dataset("rst", "wikihow_category_hierarchy")` | Relation extraction; Commonsense reasoning| |
|
| [wikiHow](https://www.wikihow.com/Main-Page) | (goal, steps) | 47,956 | `load_dataset("rst", "wikihow_goal_step")` | Intent detection| |
|
| [wikiHow](https://www.wikihow.com/Main-Page) | (text, summary) | 703,278 | `load_dataset("rst", "wikihow_summary")` | Summarization; Sentence expansion | |
|
| [wikiHow](https://www.wikihow.com/Main-Page) | (goal, first_step, second_step) | 47,787 | `load_dataset("rst", "wikihow_procedure")` | Temporal reasoning | |
|
| [wikiHow](https://www.wikihow.com/Main-Page) | (question, description, answer, related_questions) | 47,705 | `load_dataset("rst", "wikihow_question")` | Question generation| |
|
| [Wikipedia](https://www.wikipedia.org/) | (text, entities) |22,231,011 | `load_dataset("rst", "wikipedia_entities")` | Entity recognition| |
|
[Wikipedia](https://www.wikipedia.org/) | (texts, titles) | 3,296,225 | `load_dataset("rst", "wikipedia_sections")` | Summarization| |
|
| [WordNet](https://wordnet.princeton.edu/) | (word, sentence, pos) | 27,123 | `load_dataset("rst", "wordnet_pos")` | Part-of-speech tagging| |
|
| [WordNet](https://wordnet.princeton.edu/) | (word, sentence, meaning, possible_meanings) | 27,123 | `load_dataset("rst", "wordnet_meaning")` | Word sense disambiguation| |
|
| [WordNet](https://wordnet.princeton.edu/) | (word, sentence, synonyms) | 17,804 | `load_dataset("rst", "wordnet_synonym")`| Paraphrasing| |
|
| [WordNet](https://wordnet.princeton.edu/) | (word, sentence, antonyms) | 6,408 | `load_dataset("rst", "wordnet_antonym")` |Negation | |
|
| [ConTRoL]() | (premise, hypothesis, label) | 8,323 | `load_dataset("rst", "qa_control")` | Natural language inference| |
|
|[DREAM](https://transacl.org/ojs/index.php/tacl/article/view/1534)| (context, question, options, answer) | 9,164 | `load_dataset("rst", "qa_dream")` | Reading comprehension| |
|
| [LogiQA](https://doi.org/10.24963/ijcai.2020/501) | (context, question, options, answer) | 7,974 | `load_dataset("rst", "qa_logiqa")` | Reading comprehension| |
|
| [ReClor](https://openreview.net/forum?id=HJgJtT4tvB) | (context, question, options, answer) | 5,138 | `load_dataset("rst", "qa_reclor")` |Reading comprehension | |
|
| [RACE](https://doi.org/10.18653/v1/d17-1082) | (context, question, options, answer) | 44,880 | `load_dataset("rst", "qa_race")` | Reading comprehension| |
|
| [RACE-C](http://proceedings.mlr.press/v101/liang19a.html) | (context, question, options, answer) | 5,093 | `load_dataset("rst", "qa_race_c")` | Reading comprehension| |
|
| [TriviaQA](https://doi.org/10.18653/v1/P17-1147) | (context, question, answer) | 46,636 | `load_dataset("rst", "qa_triviaqa")` |Reading comprehension | |
|
| [Arxiv](https://arxiv.org/) | (text, category) | 1,696,348 | `load_dataset("rst", "arxiv_category")` |Topic classification| |
|
| [Arxiv](https://arxiv.org/) | (text, summary) | 1,696,348 | `load_dataset("rst", "arxiv_summary")` | Summarization; Sentence expansion| |
|
| [Paperswithcode](https://paperswithcode.com/) | (text, entities, datasets, methods, tasks, metrics) | 4,731,233 | `load_dataset("rst", "paperswithcode_entity")` | Entity recognition| |
|
| [Paperswithcode](https://paperswithcode.com/) | (text, summary) | 120,924 | `load_dataset("rst", "paperswithcode_summary")` | Summarization; Sentence expansion| |
|
|
|
|
|
## Bibtext for Citation Info |
|
``` |
|
@article{yuan2022restructured, |
|
title={reStructured Pre-training}, |
|
author={Yuan, Weizhe and Liu, Pengfei}, |
|
journal={arXiv preprint arXiv:2206.11147}, |
|
year={2022} |
|
} |
|
``` |
