• Stars
    star
    2,842
  • Rank 16,002 (Top 0.4 %)
  • Language
    C
  • License
    MIT License
  • Created over 7 years ago
  • Updated over 1 year ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

✨Fast Coreference Resolution in spaCy with Neural Networks

NeuralCoref 4.0: Coreference Resolution in spaCy with Neural Networks.

NeuralCoref is a pipeline extension for spaCy 2.1+ which annotates and resolves coreference clusters using a neural network. NeuralCoref is production-ready, integrated in spaCy's NLP pipeline and extensible to new training datasets.

For a brief introduction to coreference resolution and NeuralCoref, please refer to our blog post. NeuralCoref is written in Python/Cython and comes with a pre-trained statistical model for English only.

NeuralCoref is accompanied by a visualization client NeuralCoref-Viz, a web interface powered by a REST server that can be tried online. NeuralCoref is released under the MIT license.

Version 4.0 out now! Available on pip and compatible with SpaCy 2.1+.

Current Release Version spaCy Travis-CI NeuralCoref online Demo

  • Operating system: macOS / OS X · Linux · Windows (Cygwin, MinGW, Visual Studio)
  • Python version: Python 3.6+ (only 64 bit)
  • Package managers: [pip]

Install NeuralCoref

Install NeuralCoref with pip

This is the easiest way to install NeuralCoref.

pip install neuralcoref

spacy.strings.StringStore size changed error

If you have an error mentioning spacy.strings.StringStore size changed, may indicate binary incompatibility when loading NeuralCoref with import neuralcoref, it means you'll have to install NeuralCoref from the distribution's sources instead of the wheels to get NeuralCoref to build against the most recent version of SpaCy for your system.

In this case, simply re-install neuralcoref as follows:

pip uninstall neuralcoref
pip install neuralcoref --no-binary neuralcoref

Installing SpaCy's model

To be able to use NeuralCoref you will also need to have an English model for SpaCy.

You can use whatever english model works fine for your application but note that the performances of NeuralCoref are strongly dependent on the performances of the SpaCy model and in particular on the performances of SpaCy model's tagger, parser and NER components. A larger SpaCy English model will thus improve the quality of the coreference resolution as well (see some details in the Internals and Model section below).

Here is an example of how you can install SpaCy and a (small) English model for SpaCy, more information can be found on spacy's website:

pip install -U spacy
python -m spacy download en

Install NeuralCoref from source

You can also install NeuralCoref from sources. You will need to install the dependencies first which includes Cython and SpaCy.

Here is the process:

venv .env
source .env/bin/activate
git clone https://github.com/huggingface/neuralcoref.git
cd neuralcoref
pip install -r requirements.txt
pip install -e .

Internals and Model

NeuralCoref is made of two sub-modules:

  • a rule-based mentions-detection module which uses SpaCy's tagger, parser and NER annotations to identify a set of potential coreference mentions, and
  • a feed-forward neural-network which compute a coreference score for each pair of potential mentions.

The first time you import NeuralCoref in python, it will download the weights of the neural network model in a cache folder.

The cache folder is set by defaults to ~/.neuralcoref_cache (see file_utils.py) but this behavior can be overided by setting the environment variable NEURALCOREF_CACHE to point to another location.

The cache folder can be safely deleted at any time and the module will download again the model the next time it is loaded.

You can have more information on the location, downloading and caching process of the internal model by activating python's logging module before loading NeuralCoref as follows:

import logging;
logging.basicConfig(level=logging.INFO)
import neuralcoref
>>> INFO:neuralcoref:Getting model from https://s3.amazonaws.com/models.huggingface.co/neuralcoref/neuralcoref.tar.gz or cache
>>> INFO:neuralcoref.file_utils:https://s3.amazonaws.com/models.huggingface.co/neuralcoref/neuralcoref.tar.gz not found in cache, downloading to /var/folders/yx/cw8n_njx3js5jksyw_qlp8p00000gn/T/tmp_8y5_52m
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 40155833/40155833 [00:06<00:00, 6679263.76B/s]
>>> INFO:neuralcoref.file_utils:copying /var/folders/yx/cw8n_njx3js5jksyw_qlp8p00000gn/T/tmp_8y5_52m to cache at /Users/thomaswolf/.neuralcoref_cache/f46bc05a4bfba2ae0d11ffd41c4777683fa78ed357dc04a23c67137abf675e14.7d6f9a6fecf5cf09e74b65f85c7d6896b21decadb2554d486474f63b95ec4633
>>> INFO:neuralcoref.file_utils:creating metadata file for /Users/thomaswolf/.neuralcoref_cache/f46bc05a4bfba2ae0d11ffd41c4777683fa78ed357dc04a23c67137abf675e14.7d6f9a6fecf5cf09e74b65f85c7d6896b21decadb2554d486474f63b95ec4633
>>> INFO:neuralcoref.file_utils:removing temp file /var/folders/yx/cw8n_njx3js5jksyw_qlp8p00000gn/T/tmp_8y5_52m
>>> INFO:neuralcoref:extracting archive file /Users/thomaswolf/.neuralcoref_cache/f46bc05a4bfba2ae0d11ffd41c4777683fa78ed357dc04a23c67137abf675e14.7d6f9a6fecf5cf09e74b65f85c7d6896b21decadb2554d486474f63b95ec4633 to dir /Users/thomaswolf/.neuralcoref_cache/neuralcoref

Loading NeuralCoref

Adding NeuralCoref to the pipe of an English SpaCy Language

Here is the recommended way to instantiate NeuralCoref and add it to SpaCY's pipeline of annotations:

# Load your usual SpaCy model (one of SpaCy English models)
import spacy
nlp = spacy.load('en')

# Add neural coref to SpaCy's pipe
import neuralcoref
neuralcoref.add_to_pipe(nlp)

# You're done. You can now use NeuralCoref as you usually manipulate a SpaCy document annotations.
doc = nlp(u'My sister has a dog. She loves him.')

doc._.has_coref
doc._.coref_clusters

Loading NeuralCoref and adding it manually to the pipe of an English SpaCy Language

An equivalent way of adding NeuralCoref to a SpaCy model pipe is to instantiate the NeuralCoref class first and then add it manually to the pipe of the SpaCy Language model.

# Load your usual SpaCy model (one of SpaCy English models)
import spacy
nlp = spacy.load('en')

# load NeuralCoref and add it to the pipe of SpaCy's model
import neuralcoref
coref = neuralcoref.NeuralCoref(nlp.vocab)
nlp.add_pipe(coref, name='neuralcoref')

# You're done. You can now use NeuralCoref the same way you usually manipulate a SpaCy document and it's annotations.
doc = nlp(u'My sister has a dog. She loves him.')

doc._.has_coref
doc._.coref_clusters

Using NeuralCoref

NeuralCoref will resolve the coreferences and annotate them as extension attributes in the spaCy Doc, Span and Token objects under the ._. dictionary.

Here is the list of the annotations:

Attribute Type Description
doc._.has_coref boolean Has any coreference has been resolved in the Doc
doc._.coref_clusters list of Cluster All the clusters of corefering mentions in the doc
doc._.coref_resolved unicode Unicode representation of the doc where each corefering mention is replaced by the main mention in the associated cluster.
doc._.coref_scores Dict of Dict Scores of the coreference resolution between mentions.
span._.is_coref boolean Whether the span has at least one corefering mention
span._.coref_cluster Cluster Cluster of mentions that corefer with the span
span._.coref_scores Dict Scores of the coreference resolution of & span with other mentions (if applicable).
token._.in_coref boolean Whether the token is inside at least one corefering mention
token._.coref_clusters list of Cluster All the clusters of corefering mentions that contains the token

A Cluster is a cluster of coreferring mentions which has 3 attributes and a few methods to simplify the navigation inside a cluster:

Attribute or method Type / Return type Description
i int Index of the cluster in the Doc
main Span Span of the most representative mention in the cluster
mentions list of Span List of all the mentions in the cluster
__getitem__ return Span Access a mention in the cluster
__iter__ yields Span Iterate over mentions in the cluster
__len__ return int Number of mentions in the cluster

Navigating the coreference cluster chains

You can also easily navigate the coreference cluster chains and display clusters and mentions.

Here are some examples, try them out to test it for yourself.

import spacy
import neuralcoref
nlp = spacy.load('en')
neuralcoref.add_to_pipe(nlp)

doc = nlp(u'My sister has a dog. She loves him')

doc._.coref_clusters
doc._.coref_clusters[1].mentions
doc._.coref_clusters[1].mentions[-1]
doc._.coref_clusters[1].mentions[-1]._.coref_cluster.main

token = doc[-1]
token._.in_coref
token._.coref_clusters

span = doc[-1:]
span._.is_coref
span._.coref_cluster.main
span._.coref_cluster.main._.coref_cluster

Important: NeuralCoref mentions are spaCy Span objects which means you can access all the usual Span attributes like span.start (index of the first token of the span in the document), span.end (index of the first token after the span in the document), etc...

Ex: doc._.coref_clusters[1].mentions[-1].start will give you the index of the first token of the last mention of the second coreference cluster in the document.

Parameters

You can pass several additional parameters to neuralcoref.add_to_pipe or NeuralCoref() to control the behavior of NeuralCoref.

Here is the full list of these parameters and their descriptions:

Parameter Type Description
greedyness float A number between 0 and 1 determining how greedy the model is about making coreference decisions (more greedy means more coreference links). The default value is 0.5.
max_dist int How many mentions back to look when considering possible antecedents of the current mention. Decreasing the value will cause the system to run faster but less accurately. The default value is 50.
max_dist_match int The system will consider linking the current mention to a preceding one further than max_dist away if they share a noun or proper noun. In this case, it looks max_dist_match away instead. The default value is 500.
blacklist boolean Should the system resolve coreferences for pronouns in the following list: ["i", "me", "my", "you", "your"]. The default value is True (coreference resolved).
store_scores boolean Should the system store the scores for the coreferences in annotations. The default value is True.
conv_dict dict(str, list(str)) A conversion dictionary that you can use to replace the embeddings of rare words (keys) by an average of the embeddings of a list of common words (values). Ex: conv_dict={"Angela": ["woman", "girl"]} will help resolving coreferences for Angela by using the embeddings for the more common woman and girl instead of the embedding of Angela. This currently only works for single words (not for words groups).

How to change a parameter

import spacy
import neuralcoref

# Let's load a SpaCy model
nlp = spacy.load('en')

# First way we can control a parameter
neuralcoref.add_to_pipe(nlp, greedyness=0.75)

# Another way we can control a parameter
nlp.remove_pipe("neuralcoref")  # This remove the current neuralcoref instance from SpaCy pipe
coref = neuralcoref.NeuralCoref(nlp.vocab, greedyness=0.75)
nlp.add_pipe(coref, name='neuralcoref')

Using the conversion dictionary parameter to help resolve rare words

Here is an example on how we can use the parameter conv_dict to help resolving coreferences of a rare word like a name:

import spacy
import neuralcoref

nlp = spacy.load('en')

# Let's try before using the conversion dictionary:
neuralcoref.add_to_pipe(nlp)
doc = nlp(u'Deepika has a dog. She loves him. The movie star has always been fond of animals')
doc._.coref_clusters
doc._.coref_resolved
# >>> [Deepika: [Deepika, She, him, The movie star]]
# >>> 'Deepika has a dog. Deepika loves Deepika. Deepika has always been fond of animals'
# >>> Not very good...

# Here are three ways we can add the conversion dictionary
nlp.remove_pipe("neuralcoref")
neuralcoref.add_to_pipe(nlp, conv_dict={'Deepika': ['woman', 'actress']})
# or
nlp.remove_pipe("neuralcoref")
coref = neuralcoref.NeuralCoref(nlp.vocab, conv_dict={'Deepika': ['woman', 'actress']})
nlp.add_pipe(coref, name='neuralcoref')
# or after NeuralCoref is already in SpaCy's pipe, by modifying NeuralCoref in the pipeline
nlp.get_pipe('neuralcoref').set_conv_dict({'Deepika': ['woman', 'actress']})

# Let's try agin with the conversion dictionary:
doc = nlp(u'Deepika has a dog. She loves him. The movie star has always been fond of animals')
doc._.coref_clusters
# >>> [Deepika: [Deepika, She, The movie star], a dog: [a dog, him]]
# >>> 'Deepika has a dog. Deepika loves a dog. Deepika has always been fond of animals'
# >>> A lot better!

Using NeuralCoref as a server

A simple example of server script for integrating NeuralCoref in a REST API is provided as an example in examples/server.py.

To use it you need to install falcon first:

pip install falcon

You can then start the server as follows:

cd examples
python ./server.py

And query the server like this:

curl --data-urlencode "text=My sister has a dog. She loves him." -G localhost:8000

There are many other ways you can manage and deploy NeuralCoref. Some examples can be found in spaCy Universe.

Re-train the model / Extend to another language

If you want to retrain the model or train it on another language, see our training instructions as well as our blog post

More Repositories

1

transformers

🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Python
133,705
star
2

pytorch-image-models

PyTorch image models, scripts, pretrained weights -- ResNet, ResNeXT, EfficientNet, NFNet, Vision Transformer (ViT), MobileNet-V3/V2, RegNet, DPN, CSPNet, Swin Transformer, MaxViT, CoAtNet, ConvNeXt, and more
Python
28,073
star
3

diffusers

🤗 Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch and FLAX.
Python
25,619
star
4

datasets

🤗 The largest hub of ready-to-use datasets for ML models with fast, easy-to-use and efficient data manipulation tools
Python
17,530
star
5

peft

🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.
Python
15,663
star
6

candle

Minimalist ML framework for Rust
Rust
15,011
star
7

trl

Train transformer language models with reinforcement learning.
Python
9,850
star
8

text-generation-inference

Large Language Model Text Generation Inference
Python
8,939
star
9

tokenizers

💥 Fast State-of-the-Art Tokenizers optimized for Research and Production
Rust
8,885
star
10

accelerate

🚀 A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support
Python
7,854
star
11

chat-ui

Open source codebase powering the HuggingChat app
TypeScript
7,113
star
12

lerobot

🤗 LeRobot: Making AI for Robotics more accessible with end-to-end learning
Python
6,522
star
13

alignment-handbook

Robust recipes to align language models with human and AI preferences
Python
4,474
star
14

parler-tts

Inference and training library for high-quality TTS models.
Python
4,027
star
15

autotrain-advanced

🤗 AutoTrain Advanced
Python
3,925
star
16

deep-rl-class

This repo contains the syllabus of the Hugging Face Deep Reinforcement Learning Course.
MDX
3,680
star
17

diffusion-models-class

Materials for the Hugging Face Diffusion Models Course
Jupyter Notebook
3,508
star
18

notebooks

Notebooks using the Hugging Face libraries 🤗
Jupyter Notebook
3,492
star
19

distil-whisper

Distilled variant of Whisper for speech recognition. 6x faster, 50% smaller, within 1% word error rate.
Python
3,455
star
20

safetensors

Simple, safe way to store and distribute tensors
Python
2,754
star
21

text-embeddings-inference

A blazing fast inference solution for text embeddings models
Rust
2,746
star
22

knockknock

🚪✊Knock Knock: Get notified when your training ends with only two additional lines of code
Python
2,682
star
23

speech-to-speech

Speech To Speech: an effort for an open-sourced and modular GPT4-o
Python
2,540
star
24

swift-coreml-diffusers

Swift app demonstrating Core ML Stable Diffusion
Swift
2,506
star
25

optimum

🚀 Accelerate training and inference of 🤗 Transformers and 🤗 Diffusers with easy to use hardware optimization tools
Python
2,469
star
26

blog

Public repo for HF blog posts
Jupyter Notebook
2,303
star
27

setfit

Efficient few-shot learning with Sentence Transformers
Jupyter Notebook
2,142
star
28

course

The Hugging Face course on Transformers
MDX
2,005
star
29

awesome-papers

Papers & presentation materials from Hugging Face's internal science day
1,996
star
30

datatrove

Freeing data processing from scripting madness by providing a set of platform-agnostic customizable pipeline processing blocks.
Python
1,909
star
31

evaluate

🤗 Evaluate: A library for easily evaluating machine learning models and datasets.
Python
1,825
star
32

cookbook

Open-source AI cookbook
Jupyter Notebook
1,660
star
33

transfer-learning-conv-ai

🦄 State-of-the-Art Conversational AI with Transfer Learning
Python
1,654
star
34

swift-coreml-transformers

Swift Core ML 3 implementations of GPT-2, DistilGPT-2, BERT, and DistilBERT for Question answering. Other Transformers coming soon!
Swift
1,543
star
35

pytorch-openai-transformer-lm

🐥A PyTorch implementation of OpenAI's finetuned transformer language model with a script to import the weights pre-trained by OpenAI
Python
1,464
star
36

huggingface.js

Utilities to use the Hugging Face Hub API
TypeScript
1,368
star
37

Mongoku

🔥The Web-scale GUI for MongoDB
TypeScript
1,313
star
38

huggingface_hub

All the open source things related to the Hugging Face Hub.
Python
1,311
star
39

gsplat.js

JavaScript Gaussian Splatting library.
TypeScript
1,302
star
40

llm-vscode

LLM powered development for VSCode
TypeScript
1,206
star
41

hmtl

🌊HMTL: Hierarchical Multi-Task Learning - A State-of-the-Art neural network model for several NLP tasks based on PyTorch and AllenNLP
Python
1,185
star
42

nanotron

Minimalistic large language model 3D-parallelism training
Python
1,071
star
43

pytorch-pretrained-BigGAN

🦋A PyTorch implementation of BigGAN with pretrained weights and conversion scripts.
Python
986
star
44

optimum-nvidia

Python
888
star
45

torchMoji

😇A pyTorch implementation of the DeepMoji model: state-of-the-art deep learning model for analyzing sentiment, emotion, sarcasm etc
Python
880
star
46

awesome-huggingface

🤗 A list of wonderful open-source projects & applications integrated with Hugging Face libraries.
853
star
47

optimum-quanto

A pytorch quantization backend for optimum
Python
738
star
48

llm.nvim

LLM powered development for Neovim
Lua
728
star
49

naacl_transfer_learning_tutorial

Repository of code for the tutorial on Transfer Learning in NLP held at NAACL 2019 in Minneapolis, MN, USA
Python
718
star
50

dataset-viewer

Backend that powers the dataset viewer on Hugging Face dataset pages through a public API.
Python
689
star
51

swift-transformers

Swift Package to implement a transformers-like API in Swift
Swift
647
star
52

exporters

Export Hugging Face models to Core ML and TensorFlow Lite
Python
587
star
53

llm-ls

LSP server leveraging LLMs for code completion (and more?)
Rust
586
star
54

ratchet

A cross-platform browser ML framework.
Rust
574
star
55

transformers-bloom-inference

Fast Inference Solutions for BLOOM
Python
557
star
56

lighteval

LightEval is a lightweight LLM evaluation suite that Hugging Face has been using internally with the recently released LLM data processing library datatrove and LLM training library nanotron.
Python
554
star
57

pytorch_block_sparse

Fast Block Sparse Matrices for Pytorch
C++
523
star
58

node-question-answering

Fast and production-ready question answering in Node.js
TypeScript
459
star
59

large_language_model_training_playbook

An open collection of implementation tips, tricks and resources for training large language models
Python
452
star
60

swift-chat

Mac app to demonstrate swift-transformers
Swift
444
star
61

llm_training_handbook

An open collection of methodologies to help with successful training of large language models.
Python
437
star
62

text-clustering

Easily embed, cluster and semantically label text datasets
Python
422
star
63

cosmopedia

Python
416
star
64

optimum-intel

🤗 Optimum Intel: Accelerate inference with Intel optimization tools
Jupyter Notebook
393
star
65

controlnet_aux

Python
386
star
66

community-events

Place where folks can contribute to 🤗 community events
Jupyter Notebook
368
star
67

tflite-android-transformers

DistilBERT / GPT-2 for on-device inference thanks to TensorFlow Lite with Android demo apps
Java
368
star
68

nn_pruning

Prune a model while finetuning or training.
Jupyter Notebook
360
star
69

speechbox

Python
341
star
70

100-times-faster-nlp

🚀100 Times Faster Natural Language Processing in Python - iPython notebook
HTML
325
star
71

education-toolkit

Educational materials for universities
Jupyter Notebook
324
star
72

transformers.js-examples

A collection of 🤗 Transformers.js demos and example applications
JavaScript
323
star
73

open-muse

Open reproduction of MUSE for fast text2image generation.
Python
320
star
74

local-gemma

Gemma 2 optimized for your local machine.
Python
317
star
75

unity-api

C#
313
star
76

audio-transformers-course

The Hugging Face Course on Transformers for Audio
MDX
308
star
77

datablations

Scaling Data-Constrained Language Models
Jupyter Notebook
305
star
78

hf_transfer

Rust
287
star
79

dataspeech

Python
262
star
80

huggingface-llama-recipes

Jupyter Notebook
259
star
81

optimum-benchmark

🏋️ A unified multi-backend utility for benchmarking Transformers, Timm, PEFT, Diffusers and Sentence-Transformers with full support of Optimum's hardware optimizations & quantization schemes.
Python
245
star
82

diarizers

Python
238
star
83

hub-docs

Docs of the Hugging Face Hub
221
star
84

llm-swarm

Manage scalable open LLM inference endpoints in Slurm clusters
Python
216
star
85

sam2-studio

Swift
196
star
86

optimum-neuron

Easy, fast and very cheap training and inference on AWS Trainium and Inferentia chips.
Jupyter Notebook
193
star
87

data-is-better-together

Let's build better datasets, together!
Jupyter Notebook
192
star
88

instruction-tuned-sd

Code for instruction-tuning Stable Diffusion.
Python
189
star
89

simulate

🎢 Creating and sharing simulation environments for embodied and synthetic data research
Python
185
star
90

OBELICS

Code used for the creation of OBELICS, an open, massive and curated collection of interleaved image-text web documents, containing 141M documents, 115B text tokens and 353M images.
Python
184
star
91

diffusion-fast

Faster generation with text-to-image diffusion models.
Python
179
star
92

olm-datasets

Pipeline for pulling and processing online language model pretraining data from the web
Python
173
star
93

api-inference-community

Python
161
star
94

jat

General multi-task deep RL Agent
Python
154
star
95

workshops

Materials for workshops on the Hugging Face ecosystem
Jupyter Notebook
148
star
96

coreml-examples

Swift Core ML Examples
Jupyter Notebook
147
star
97

optimum-habana

Easy and lightning fast training of 🤗 Transformers on Habana Gaudi processor (HPU)
Python
147
star
98

chug

Minimal sharded dataset loaders, decoders, and utils for multi-modal document, image, and text datasets.
Python
140
star
99

sharp-transformers

A Unity plugin for using Transformers models in Unity.
C#
139
star
100

hf-hub

Rust client for the huggingface hub aiming for minimal subset of features over `huggingface-hub` python package
Rust
132
star