Question Answering over Multiple and Heterogeneous Knowledge Bases

                                       

The MuHeQA (Multiple and Heterogeneous Question-Answering) system creates natural language answers from natural language questions using knowledge base from both structured (KG) and unstructured (documents) data sources.

Quick Start!

Install the muheqa package:

pip install muheqa

Create a new connection to Wikidata, or DBpedia, or D4C (Drugs4Covid). The first time it may take a few minutes to download the required models:

import muheqa.connector as mhqa

wikidata = mhqa.connect(wikidata=True)

And finally, make a question in natural language!:

response = wikidata.query("Who is the father of Barack Obama")
print("Response:",response)

Preparation

1. Prepare a Python 3 environment and install the Conda framework.
2. Clone this repo:

   git clone https://github.com/librairy/MuHeQA.git
   

3. Move into the root directory:

   cd MuHeQA
   

4. Create an environment (if it does not already exist):

   conda create --name .muheqa python=3.9
   

5. Activate the environment:

   conda activate .muheqa
   

6. Download the answer classifier and unzip into the root project directory. The folder resources_dir/ is created.

   wget -O resources.zip https://delicias.dia.fi.upm.es/nextcloud/index.php/s/Jp5FeoBn57c8k4M/download
   unzip resources.zip
   

7. Install dependencies

   pip install -r requirements.txt
   

M1 Environments (only for Apple's M1 devices)

1. Install TensorFlow dependencies

   conda install -c apple tensorflow-deps
   

2. Install base TensorFlow

   pip install tensorflow-macos
   

3. Install tensorflow-metal plugin

   pip install tensorflow-metal