Natural-Language-To-SPARQL

Translate a query in natural language to SPARQL query with an interactive web based interface.

Neural SPARQL Machines

Installation guide for spacey

pip install -U spacey

python -m venv .env
source .env/bin/activate

pip install spacey

Installation guide for flask

mkdir myproject

cd myproject

python3 -m venv venv

.venv/bin/activate

pip install Flask

Data preparation

To generate the training data, execute the following commands

mkdir data/monument_300

python generator.py --templates data/annotations_monument.csv  --output data/monument_300

Training

Now go back to the initial directory and launch train.sh to train the model. The first parameter is the prefix of the data directory and the second parameter is the number of training epochs.

sh train.sh data/monument_300 120000

This command will create a model directory called data/monument_300_model.

Inference

Predict the SPARQL sentence for a given question with a given model. In this, we wil run back.py. After that go to http://localhost:5000/result on any web browser. This has front end. Enter your query and sparql sentence will be generated in next page.

Execution of the application - Run back.py and then go to http://localhost:5000/ .