Skip to main content

Thai Word-Segmentation with LSTM in Tensorflow

Project description

# Cutkum ['คัดคำ']
Cutkum ('คัดคำ') is a python code for Thai Word-Segmentation using Recurrent Neural Network (RNN) based on Tensorflow library.

Cutkum is trained on BEST2010, a 5 Millions Thai words corpus by NECTEC ( It also comes with an already trained model, and can be used right out of the box. Cutkum is still a work-in-progress project. Evaluated on the 10% hold-out data from BEST2010 corpus (~600,000 words), the included trained model currently performs at

98.0% recall, 96.3% precision, 97.1% F-measure (character-level)
93.5% recall, 94.1% precision and 94.0% F-measure (word-level -- same evaluation method as BEST2010)

# Update
Feb 17, 2018 - add the training script

# Requirements
* python = 2.7, 3.0+
* tensorflow = 1.4+

# Installation

`cutkum` can be installed using `pip` and the trained model can be downloaded from github. The current included model (model/lstm.l6.d2.pb) is a stacked bi-directional LSTM neural network with 6 layers.

pip install cutkum

# then download the trained model (either from github) or with wget


# Usages

Once installed, you can use `cutkum` within your python code to tokenize thai sentences.


>>> from cutkum.tokenizer import Cutkum

>>> ck = Cutkum('lstm.l6.d2.pb')
>>> words = ck.tokenize("สารานุกรมไทยสำหรับเยาวชนฯ")

# python 3.0
>>> words
['สารานุกรม', 'ไทย', 'สำหรับ', 'เยาวชน', 'ฯ']

# python 2.7
>>> print("|".join(words))
# สารานุกรม|ไทย|สำหรับ|เยาวชน|ฯ


You can also use `cutkum` straight from the command line.

usage: cutkum [-h] [-v] -m MODEL_FILE
[-o OUTPUT_FILE | -od OUTPUT_DIR] [--max | --viterbi]

cutkum -m model/lstm.l6.d2.pb -s "สารานุกรมไทยสำหรับเยาวชนฯ"

# output as

`cutkum` can also be used to segment text within a file (with -i), or to segment all the files within a given directory (with -id).

cutkum -m model/lstm.l6.d2.pb -i input.txt -o output.txt
cutkum -m model/lstm.l6.d2.pb -id input_dir -od output_dir

## License

This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details

## To Do
* Improve performance, with better better model, and better included trained-model
* Improve the speed when processing big file

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for cutkum, version 2.4
Filename, size File type Python version Upload date Hashes
Filename, size cutkum-2.4.tar.gz (6.2 MB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page