# cdQA **Repository Path**: vax52/cdQA ## Basic Information - **Project Name**: cdQA - **Description**: No description available - **Primary Language**: Unknown - **License**: Apache-2.0 - **Default Branch**: master - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2021-07-01 - **Last Updated**: 2021-07-01 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # cdQA: Closed Domain Question Answering [![Build Status](https://travis-ci.com/cdqa-suite/cdQA.svg?branch=master)](https://travis-ci.com/cdqa-suite/cdQA) [![codecov](https://codecov.io/gh/cdqa-suite/cdQA/branch/master/graph/badge.svg)](https://codecov.io/gh/cdqa-suite/cdQA) [![PyPI Version](https://img.shields.io/pypi/v/cdqa.svg)](https://pypi.org/project/cdqa/) [![PyPI Downloads](https://img.shields.io/pypi/dm/cdqa.svg)](https://pypi.org/project/cdqa/) [![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/cdqa-suite/cdQA/master?filepath=examples%2Ftutorial-first-steps-cdqa.ipynb) [![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/cdqa-suite/cdQA/blob/master/examples/tutorial-first-steps-cdqa.ipynb) [![Contributor Covenant](https://img.shields.io/badge/Contributor%20Covenant-v1.4%20adopted-ff69b4.svg)](.github/CODE_OF_CONDUCT.md) [![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg)](http://makeapullrequest.com) ![GitHub](https://img.shields.io/github/license/cdqa-suite/cdQA.svg) An End-To-End Closed Domain Question Answering System. Built on top of the HuggingFace [transformers](https://github.com/huggingface/transformers) library. **⛔ [NOT MAINTAINED] This repository is no longer maintained, but is being kept around for educational purposes. If you want a maintained alternative to cdQA check out: https://github.com/deepset-ai/haystack** ## cdQA in details If you are interested in understanding how the system works and its implementation, we wrote an [article on Medium](https://towardsdatascience.com/how-to-create-your-own-question-answering-system-easily-with-python-2ef8abc8eb5) with a high-level explanation. We also made a presentation during the \#9 NLP Breakfast organised by [Feedly](feedly.com). You can check it out [here](https://blog.feedly.com/nlp-breakfast-9-closed-domain-question-answering/). ## Table of Contents - [Installation](#Installation) - [With pip](#With-pip) - [From source](#From-source) - [Hardware Requirements](#Hardware-Requirements) - [Getting started](#Getting-started) - [Preparing your data](#Preparing-your-data) - [Manual](#Manual) - [With converters](#With-converters) - [Downloading pre-trained models](#Downloading-pre-trained-models) - [Training models](#Training-models) - [Making predictions](#Making-predictions) - [Evaluating models](#Evaluating-models) - [Notebook Examples](#Notebook-Examples) - [Deployment](#Deployment) - [Manual](#Manual-1) - [Contributing](#Contributing) - [References](#References) - [LICENSE](#LICENSE) ## Installation ### With pip ```shell pip install cdqa ``` ### From source ```shell git clone https://github.com/cdqa-suite/cdQA.git cd cdQA pip install -e . ``` ### Hardware Requirements Experiments have been done with: - **CPU** 👉 AWS EC2 `t2.medium` Deep Learning AMI (Ubuntu) Version 22.0 - **GPU** 👉 AWS EC2 `p3.2xlarge` Deep Learning AMI (Ubuntu) Version 22.0 + a single Tesla V100 16GB. ## Getting started ### Preparing your data #### Manual To use `cdQA` you need to create a pandas dataframe with the following columns: | title | paragraphs | | ----------------- | ------------------------------------------------------ | | The Article Title | [Paragraph 1 of Article, ... , Paragraph N of Article] | #### With converters The objective of `cdqa` converters is to make it easy to create this dataframe from your raw documents database. For instance the `pdf_converter` can create a `cdqa` dataframe from a directory containing `.pdf` files: ```python from cdqa.utils.converters import pdf_converter df = pdf_converter(directory_path='path_to_pdf_folder') ``` You will need to install [Java OpenJDK](https://openjdk.java.net/install/) to use this converter. We currently have converters for: - pdf - markdown We plan to improve and add more converters in the future. Stay tuned! ### Downloading pre-trained models and data You can download the models and data manually from the GitHub [releases](https://github.com/cdqa-suite/cdQA/releases) or use our download functions: ```python from cdqa.utils.download import download_squad, download_model, download_bnpp_data directory = 'path-to-directory' # Downloading data download_squad(dir=directory) download_bnpp_data(dir=directory) # Downloading pre-trained BERT fine-tuned on SQuAD 1.1 download_model('bert-squad_1.1', dir=directory) # Downloading pre-trained DistilBERT fine-tuned on SQuAD 1.1 download_model('distilbert-squad_1.1', dir=directory) ``` ### Training models Fit the pipeline on your corpus using the pre-trained reader: ```python import pandas as pd from ast import literal_eval from cdqa.pipeline import QAPipeline df = pd.read_csv('your-custom-corpus-here.csv', converters={'paragraphs': literal_eval}) cdqa_pipeline = QAPipeline(reader='bert_qa.joblib') # use 'distilbert_qa.joblib' for DistilBERT instead of BERT cdqa_pipeline.fit_retriever(df=df) ``` If you want to fine-tune the reader on your custom SQuAD-like annotated dataset: ```python cdqa_pipeline = QAPipeline(reader='bert_qa.joblib') # use 'distilbert_qa.joblib' for DistilBERT instead of BERT cdqa_pipeline.fit_reader('path-to-custom-squad-like-dataset.json') ``` Save the reader model after fine-tuning: ```python cdqa_pipeline.dump_reader('path-to-save-bert-reader.joblib') ``` ### Making predictions To get the best prediction given an input query: ```python cdqa_pipeline.predict(query='your question') ``` To get the N best predictions: ```python cdqa_pipeline.predict(query='your question', n_predictions=N) ``` There is also the possibility to change the weight of the retriever score versus the reader score in the computation of final ranking score (the default is 0.35, which is shown to be the best weight on the development set of SQuAD 1.1-open) ```python cdqa_pipeline.predict(query='your question', retriever_score_weight=0.35) ``` ### Evaluating models In order to evaluate models on your custom dataset you will need to annotate it. The annotation process can be done in 3 steps: 1. Convert your pandas DataFrame into a json file with SQuAD format: ```python from cdqa.utils.converters import df2squad json_data = df2squad(df=df, squad_version='v1.1', output_dir='.', filename='dataset-name') ``` 2. Use an annotator to add ground truth question-answer pairs: Please refer to our [`cdQA-annotator`](https://github.com/cdqa-suite/cdQA-annotator), a web-based annotator for closed-domain question answering datasets with SQuAD format. 3. Evaluate the pipeline object: ```python from cdqa.utils.evaluation import evaluate_pipeline evaluate_pipeline(cdqa_pipeline, 'path-to-annotated-dataset.json') ``` 4. Evaluate the reader: ```python from cdqa.utils.evaluation import evaluate_reader evaluate_reader(cdqa_pipeline, 'path-to-annotated-dataset.json') ``` ## Notebook Examples We prepared some notebook examples under the [examples](examples) directory. You can also play directly with these notebook examples using [Binder](https://gke.mybinder.org/) or [Google Colaboratory](https://colab.research.google.com/notebooks/welcome.ipynb): | Notebook | Hardware | Platform | | -------------------------------- | ------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | [1] First steps with cdQA | CPU or GPU | [![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/cdqa-suite/cdQA/master?filepath=examples%2Ftutorial-first-steps-cdqa.ipynb) [![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/cdqa-suite/cdQA/blob/master/examples/tutorial-first-steps-cdqa.ipynb) | | [2] Using the PDF converter | CPU or GPU | [![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/cdqa-suite/cdQA/master?filepath=examples%2Ftutorial-use-pdf-converter.ipynb) [![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/cdqa-suite/cdQA/blob/master/examples/tutorial-use-pdf-converter.ipynb) | | [3] Training the reader on SQuAD | GPU | [![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/cdqa-suite/cdQA/blob/master/examples/tutorial-train-reader-squad.ipynb) | Binder and Google Colaboratory provide temporary environments and may be slow to start but we recommend them if you want to get started with `cdQA` easily. ## Deployment ### Manual You can deploy a `cdQA` REST API by executing: ```shell export dataset_path=path-to-dataset.csv export reader_path=path-to-reader-model FLASK_APP=api.py flask run -h 0.0.0.0 ``` You can now make requests to test your API (here using [HTTPie](https://httpie.org/)): ```shell http localhost:5000/api query=='your question here' ``` If you wish to serve a user interface on top of your `cdQA` system, follow the instructions of [cdQA-ui](https://github.com/cdqa-suite/cdQA-ui), a web interface developed for `cdQA`. ## Contributing Read our [Contributing Guidelines](.github/CONTRIBUTING.md). ## References | Type | Title | Author | Year | | -------------------- | -------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------- | ---- | | :video_camera: Video | [Stanford CS224N: NLP with Deep Learning Lecture 10 – Question Answering](https://youtube.com/watch?v=yIdF-17HwSk) | Christopher Manning | 2019 | | :newspaper: Paper | [Reading Wikipedia to Answer Open-Domain Questions](https://arxiv.org/abs/1704.00051) | Danqi Chen, Adam Fisch, Jason Weston, Antoine Bordes | 2017 | | :newspaper: Paper | [Neural Reading Comprehension and Beyond](https://cs.stanford.edu/people/danqi/papers/thesis.pdf) | Danqi Chen | 2018 | | :newspaper: Paper | [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://arxiv.org/abs/1810.04805) | Jacob Devlin, Ming-Wei Chang, Kenton Lee, Kristina Toutanova | 2018 | | :newspaper: Paper | [Contextual Word Representations: A Contextual Introduction](https://arxiv.org/abs/1902.06006) | Noah A. Smith | 2019 | | :newspaper: Paper | [End-to-End Open-Domain Question Answering with BERTserini](https://arxiv.org/abs/1902.01718) | Wei Yang, Yuqing Xie, Aileen Lin, Xingyu Li, Luchen Tan, Kun Xiong, Ming Li, Jimmy Lin | 2019 | | :newspaper: Paper | [Data Augmentation for BERT Fine-Tuning in Open-Domain Question Answering](https://arxiv.org/abs/1904.06652) | Wei Yang, Yuqing Xie, Luchen Tan, Kun Xiong, Ming Li, Jimmy Lin | 2019 | | :newspaper: Paper | [Passage Re-ranking with BERT](https://arxiv.org/abs/1901.04085) | Rodrigo Nogueira, Kyunghyun Cho | 2019 | | :newspaper: Paper | [MRQA: Machine Reading for Question Answering](https://mrqa.github.io/) | Jonathan Berant, Percy Liang, Luke Zettlemoyer | 2019 | | :newspaper: Paper | [Unsupervised Question Answering by Cloze Translation](https://arxiv.org/abs/1906.04980) | Patrick Lewis, Ludovic Denoyer, Sebastian Riedel | 2019 | | :computer: Framework | [Scikit-learn: Machine Learning in Python](http://jmlr.csail.mit.edu/papers/v12/pedregosa11a.html) | Pedregosa et al. | 2011 | | :computer: Framework | [PyTorch](https://arxiv.org/abs/1906.04980) | Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan | 2016 | | :computer: Framework | [Transformers: State-of-the-art Natural Language Processing for TensorFlow 2.0 and PyTorch.](https://github.com/huggingface/transformers) | Hugging Face | 2018 | ## LICENSE [Apache-2.0](LICENSE)