Bots for Open Data Interactions (BODI) - Chatbot Generator for Open Data Sources

The objective of this project is to build a chatbot generator to interact with public open data sources such as .csv files so every citizen can get relevant information from a dataset in an easy way (via a chatbot interface).

The generated chatbots are built with Xatkit, an open-source chatbot development platform.

Quick Start

1 - Build the latest version of Xatkit

2 - Install xatkit-bot-platform/labs-bot-testing-tools and xatkit-bot-platform/xatkit-core-library-i18n

3 - Clone this repository

4 - Navigate to the root directory and build the project:

mvn clean compile

5 - You can run the bot generator!

mvn exec:java -Dexec.mainClass="mvn exec:java -Dexec.mainClass="bodi.generator.ui.Application""

Then, you can access the bodi-generator UI at http://localhost:8080/bodi-generator/

6 - We provide a solution for when a user utterance is not matched with a bot intent. The objective is to empower the chatbot with language models able to find answers to a wide range of inputs. A server running these models can be deployed and then, when no intent is matched, a request to this server is sent and a response is received. Our solution is implemented in opendata-for-all/bodi-nlp-server.

7 - Generated chatbots use opendata-for-all/bodi-drillbit to query the database containing the tabular data.

8 - You must use an intent recognition provider to detect user's intents. We recommend using our own implementation xatkit-nlu-server with its bot client, although it is possible to use other options such as Dialogflow or NLP.JS

9 - Once the chatbot is generated, you can run it:

cd <bot-folder>
mvn clean compile

To run the chatbot tests:

mvn test

To run the chatbot itself:

mvn exec:java -Dexec.mainClass="com.xatkit.bot.App"

📚 Check the Xatkit Wiki to learn more about Xatkit chatbots.