qiuyuew / openmatch

An Open-Source Package for Information Retrieval.

OpenMatch

An Open-Source Package for Information Retrieval.

😃 What's New

• Top Spot on TREC-COVID Challenge (May 2020, Round2)

  The twin goals of the challenge are to evaluate search algorithms and systems for helping scientists, clinicians, policy makers, and others manage the existing and rapidly growing corpus of scientific literature related to COVID-19, and to discover methods that will assist with managing scientific information in future global biomedical crises.
  >> Reproduce Our Submit >> About COVID-19 Dataset >> Our Paper

Overview

OpenMatch integrates excellent neural methods and technologies to provide a complete solution for deep text matching and understanding.

1/ Document Retrieval

Document Retrieval refers to extracting a set of related documents from large-scale document-level data based on user queries.

* Sparse Retrieval

Sparse Retriever is defined as a sparse bag-of-words retrieval model.

* Dense Retrieval

Dense Retriever performs retrieval by encoding documents and queries into dense low-dimensional vectors, and selecting the document that has the highest inner product with the query

2/ Document Reranking

Document reranking aims to further match user query and documents retrieved by the previous step with the purpose of obtaining a ranked list of relevant documents.

* Neural Ranker

Neural Ranker uses neural network as ranker to reorder documents.

* Feature Ensemble

Feature Ensemble can fuse neural features learned by neural ranker with the features of non-neural methods to obtain more robust performance

3/ Domain Transfer Learning

Domain Transfer Learning can leverages external knowledge graphs or weak supervision data to guide and help ranker to overcome data scarcity.

* Knowledge Enhancemnet

Knowledge Enhancement incorporates entity semantics of external knowledge graphs to enhance neural ranker.

* Data Augmentation

Data Augmentation leverages weak supervision data to improve the ranking accuracy in certain areas that lacks large scale relevance labels.

Stage	Model	Paper
1/ Sparse Retrieval	BM25	Best Match25 ~Tool
1/ Dense Retrieval	ANN	Approximate nearest neighbor ~Tool
2/ Neural Ranker	K-NRM	End-to-End Neural Ad-hoc Ranking with Kernel Pooling ~Paper
2/ Neural Ranker	Conv-KNRM	Convolutional Neural Networks for Soft-Matching N-Grams in Ad-hoc Search ~Paper
2/ Neural Ranker	TK	Interpretable & Time-Budget-Constrained Contextualization for Re-Ranking ~Paper
2/ Neural Ranker	BERT	BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding ~Paper
2/ Feature Ensemble	Coordinate Ascent	Linear feature-based models for information retrieval. Information Retrieval ~Paper
3/ Knowledge Enhancement	EDRM	Entity-Duet Neural Ranking: Understanding the Role of Knowledge Graph Semantics in Neural Information Retrieval ~Paper
3/ Data Augmentation	ReInfoSelect	Selective Weak Supervision for Neural Information Retrieval ~Paper

Note that the BERT model is following huggingface's implementation - transformers, so other bert-like models are also available in our toolkit, e.g. electra, scibert.

Installation

* From PyPI

pip install git+https://github.com/thunlp/OpenMatch.git

* From Source

git clone https://github.com/thunlp/OpenMatch.git
cd OpenMatch
python setup.py install

Quick Start

* Detailed examples are available here.

import torch
import OpenMatch as om

query = "Classification treatment COVID-19"
doc = "By retrospectively tracking the dynamic changes of LYM% in death cases and cured cases, this study suggests that lymphocyte count is an effective and reliable indicator for disease classification and prognosis in COVID-19 patients."

* For bert-like models:

from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("allenai/scibert_scivocab_uncased")
input_ids = tokenizer.encode(query, doc)
model = om.models.Bert("allenai/scibert_scivocab_uncased")
ranking_score, ranking_features = model(torch.tensor(input_ids).unsqueeze(0))

* For other models:

tokenizer = om.data.tokenizers.WordTokenizer(pretrained="./data/glove.6B.300d.txt")
query_ids, query_masks = tokenizer.process(query, max_len=16)
doc_ids, doc_masks = tokenizer.process(doc, max_len=128)
model = om.models.KNRM(vocab_size=tokenizer.get_vocab_size(),
                       embed_dim=tokenizer.get_embed_dim(),
                       embed_matrix=tokenizer.get_embed_matrix())
ranking_score, ranking_features = model(torch.tensor(query_ids).unsqueeze(0),
                                        torch.tensor(query_masks).unsqueeze(0),
                                        torch.tensor(doc_ids).unsqueeze(0),
                                        torch.tensor(doc_masks).unsqueeze(0))

* The GloVe can be downloaded using:

wget http://nlp.stanford.edu/data/glove.6B.zip -P ./data
unzip ./data/glove.6B.zip -d ./data

Experiments

* Ad-hoc Search

Model	ClueWeb09	Robust04	ClueWeb12
KNRM	0.1880	0.3016	0.0968
Conv-KNRM	0.1894	0.2907	0.0896
EDRM	0.2015	0.2993	0.0937
TK	0.2306	0.2822	0.0966
BERT Base	0.2701	0.4168	0.1183
ELECTRA Base	0.2861	0.4668	0.1078

* MS MARCO Passage Ranking

Model	eval	dev
BERT Base	0.345	0.349
ELECTRA Base	0.344	0.352
RoBERTa Large	0.375	0.386
ELECTRA Large	0.376	0.388

* MS MARCO Document Ranking

Contribution

Thanks to all the people who contributed to OpenMatch!

Kaitao Zhang, Si Sun, Zhenghao Liu, Aowei Lu

Project Organizers

• Zhiyuan Liu
  • Tsinghua University
  • Homepage
• Chenyan Xiong
  • Microsoft Research AI
  • Homepage
• Maosong Sun
  • Tsinghua University
  • Homepage