Home

YATO: Yet Another deep learning based Text analysis Open toolkit

YATO, an open-source Python library for text analysis. In particular, YATO focuses on sequence labeling and sequence classification tasks, including extensive fundamental NLP tasks such as part-of-speech tagging, chunking, NER, CCG supertagging, sentiment analysis, and sentence classification. YATO can design both specific RNN-based and Transformer-based through user-friendly configuration and integrating the SOTA pre-trained language models, such as BERT.

Its previous version called NCRF++ has been accepted as a demo paper by ACL 2018. The in-depth experimental report based on NCRF++ was accepted as the best paper by COLING 2018. Compared with NCRF++, the highlight of YATO is the support for Pre-trained Language Model and sentence classification tasks.

Quick Links

• Introduction
  • Getting Started
  • Configurations
    • Train Configuration Parameters
    • Decode Configuration Parameters
  • YATO model
    • YATO Function
    • Source Code
  • Model
    • CharbiGRU
    • CharbiLSTM
    • Classification Head
    • CRF
    • Custom Model
    • Transformers
    • Classification Head
    • CRF
    • Custom Model
    • Transformers
    • Sentence Classification
    • Sequence Labeling
    • Word Representation
    • Word Sequence
  • Utils
    • Alphabet
    • Data
    • Functions
    • Metric
    • Tag-Scheme Converter
    • Text Attention

Quick Start: Develop YATO model in 10 lines

Installation

We provide an easy way to use the toolkit YATO from PyPI

pip install ylab-yato

Or directly install it from the source code

git clone https://github.com/jiesutd/YATO.git

Train

from yato import YATO
model = YATO(configuration file)
model.train()

Decode

from yato import YATO
decode_model = YATO(configuration file)
result_dict = decode_model.decode()