Natural-Language-To-SPARQL

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

Go to Untitled1.ipynb

The below steps are optional

Neural SPARQL Machines

Create a virtual env -

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

Install requirements -

pip install spacey

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.

Execution of the application -

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.