Fileread Fork of ABSA-BERT-PAIR

Setting up the project

You need to declare two environment variables in your project. BERT_MODEL_PATH, the path to the trained Semeval 2014 NLI M model and BERT_CONFIG_PATH, the path to the directory that holds the vocab and bert json configuration.

export BERT_MODEL_PATH=$(pwd)/models/semeval_absa.bin
export BERT_CONFIG_PATH=$(pwd)/models

Contents of BERT_CONFIG_PATH

bert_config.json  semeval_absa.bin  vocab.txt

Installing th deps

First setup virtualenv and the pip environment

# setup virtual env and pip
$ python3 -m venv .venv
$ source .venv/bin/activate
$ pip install --upgrade pip
$ pip install -e .

Using the command line interface

# the first sentence is the main context
# the second sentence is the aspect
$ python absa "The food was fantastic. Too bad about the service though" "service"

Using the class

Use the Predictor.predict method to generate labels. Examples of how to use it is in the absa/__main__.py file.