fproject Sentiment API

Arquitecture

Tensorflow Serving

Ones we have the trained model we need to load the model in the Tensorflow Serving Api and it will produce a gRPC server that we and interact with.

It's a little bit tricky and i recomend to use this tutorials:

• How to Deploy a Tensorflow Model to Production by Siraj Raval

• How to deploy Machine Learning models with TensorFlow by Vitaly Bezgachev

gRPC Api

Now that we have our tensorflow model handle by the TF Api we have to make a client to talk to it.

Dependencies

• python 3.6
• gensim
• tensorflow
• nltk
• grpcio
• grpcio-tools

Download spanish word2vec model:

$ wget http://cs.famaf.unc.edu.ar/~ccardellino/SBWCE/SBW-vectors-300-min5.bin.gz
$ gunzip SBW-vectors-300-min5.bin.gz

Simple set up

• Modify the Tensorflow Api host and port values

Usage

Clone this project.

$ cd fproject
$ python api.py

Full System Arquitecture

NODE SERVER

• https://github.com/si-m/fproject_node

FRONT-END REACT APP

• https://github.com/si-m/fproject_front