tf-multitask-pos-ner

Deep multi-task learning for POS-tagging and NER using tensorflow.

Training the Model

To train the model, a version of GloVe is required. Please download it here http://nlp.stanford.edu/data/glove.840B.300d.zip and unzip it. Additionally, you have to provide training and validation data for both, POS-Tagging and NER. Please see sample-data for the format of the files. Training is started by:

cd src/
mkdir model/
python3 main.py /path/to/glove.txt /path/to/pos_train.txt /path/to/pos_val.txt /path/to/ner_train.iob /path/to/ner_val.iob model/

Starting the webserver

cd src/
python3 server.py model/

Building the client

The client is written in Angular and can be built using

cd src/client/
ng build -op ../static/