FastNLP

Introduction

fastNLP, a light NLP toolkit, aims at implement NLP tasks quickly and build complex models.
Features
- Provide automatic download of some datasets and Pre-trained models, provide built-in Loader and Pipe of various datasets
- simplify the data preprocessing process by using a unified Tabular data loader
- Provide various NLP tools, such as Embedding loading and intermediate data cache
- Provide various components and reproduction models of Neural Network. Including Chinese Word Segment, Named Entity Recognition, Syntactic Analysis, Text Classification, Text Matching, Co-reference Resolution, Text Summarization, etc.
- Provide various build-in callback functions.

Structure

fastNLP	Fuctions
`fastNLP.core`	core functions, such as data process components, trainer, tester, etc
`fastNLP.models`	Neural Network models
`fastNLP.modules`	components used to build Neural Network models
`fastNLP.embeddings`	functions tuning setence index into vector index
`fastNLP.io`	Read and write functions

FastHan

fastHan, a Chinese NLP tool based on fastNLP and pytorch, has two versions: base and large. Its kernel is a joint model based on BERT. it is trained in 13 corpus and can handle four tasks: Chinese word segmentation, part-of-speech tagging, dependency Parsing, and named entity recognition.