Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
NCRF++, a Neural Sequence Labeling Toolkit. Easy use to any sequence labeling tasks (e.g. NER, POS, Segmentation). It includes character LSTM/CNN, word LSTM/CNN and softmax/CRF components.
| Date | Stars |
|---|---|
| 2026-07-24 | 1901 |
| 2026-07-25 | 1901 |
| 2026-07-28 | 1901 |
| 2026-07-30 | 1901 |
| 2026-08-06 | 1901 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
 # NCRF++: An Open-source Neural Sequence Labeling Toolkit * [1. Introduction](#Introduction) * [2. Requirement](#Requirement) * [3. Advantages](#Advantages) * [4. Usage](#Usage) * [5. Data Format](#Data-Format) * [6. Performance](#Performance) * [7. Add Handcrafted Features](#Add-Handcrafted-Features) * [8. Speed](#Speed) * [9. N best Decoding](#N-best-Decoding) * [10. Reproduce Paper Results and Hyperparameter Tuning](#Reproduce-Paper-Results-and-Hyperparameter-Tuning) * [11. Report Issue or Problem](#Report-Issue-or-Problem) * [12. Cite](#Cite) * [13. Future Plan](#Future-Plan) * [13. Update](#Update) ## Introduction Sequence labeling models are quite popular in many NLP tasks, such as Named Entity Recognition (NER), part-of-speech (POS) tagging and word segmentation. State-of-the-art sequence labeling models mostly utilize the CRF structure with input word features. LSTM (or bidirectional LSTM) is a popular deep learning based feature extractor in sequence labeling task. And CNN can also be used due to faster computation. Besides, features within word are also useful to represent word, which can be captured by character LSTM or character CNN structure or human-defined neural features. NCRF++ is a PyTorch based framework with flexiable choices of input features and output structures. The design of neural sequence labeling models with NCRF++ is fully configurable through a configuration file, which does not require any code work. NCRF++ can be regarded as a neural network version of [CRF++](http://taku910.github.io/crfpp/), which is a famous statistical CRF framework. This framework has been accepted by [ACL 2018](https://arxiv.org/abs/1806.05626) as demonstration paper. And the detailed experiment report and analysis using NCRF++ has been accepted at [COLING 2018](https://arxiv.org/abs/1806.04470) as the best paper. NCRF++ supports different structure combinations of on three levels: character sequence representation, word sequence representation and inference layer. * Character sequence representation: character LSTM, character GRU, character CNN and handcrafted word features. * Word sequence representation: word LSTM, word GRU, word CNN. * Inference layer: Softmax, CRF. Welcome to star this repository! ## Requirement Python: 2 or 3 PyTorch: 1.0 [PyTorch 0.3 compatible version is here.](https://github.com/jiesutd/NCRFpp/tree/PyTorch0.3) ## Advantages * Fully configurable: all the neural model structures can be set with a configuration file. * State-of-the-art system performance: models build on NCRF++ can give comparable or better results compared with state-of-the-art models. * Flexible with features: user can define their own features and pretrained feature embeddings. * Fast running speed: NCRF++ utilizes fully batched operations, making the system efficient with the help of GPU (>1000sent/s for training and >2000sents/s for decoding). * N best output: NCRF++ support `nbest` decoding (with their probabilities). ## Usage NCRF++ supports designing the neural network structure through a configuration file. The program can run in two status; ***training*** and ***decoding***. (sample configuration and data have been included in this repository) In ***training*** status: `python main.py --config demo.train.config` In ***decoding*** status: `python main.py --config demo.decode.config` The configuration file controls the network structure, I/O, training setting and hyperparameters. ***Detail configurations and explanations are listed [here](readme/Configuration.md).*** NCRF++ is designed in three layers (shown below): character sequence layer; word sequence layer and inference layer. By using the configuration file, most of the state-of-the-art models can be easily replicated ***without coding***. On the other hand, users can extend each layer by designing their own modules (for example, they may want to design their
Excerpt of 10,652 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:7dc705e7354fa410, topic:natural-language-processing, topic:named-entity-recognition, readme:named entity recognition
matched fp:7dc705e7354fa410, topic:pytorch
