Video Game Chatbot

Objective

This repository represents the work accomplished by me (Anmol Saini) under the supervision of Dr. Cogan Shimizu during an undergraduate honors independent study in the Spring and Fall 2023 semesters. The primary goal of this project is to develop a knowledge graph-powered agent (or chatbot) capable of multiturn conversations about video game information with users.

Repository Structure

This section gives a brief overview of the layout of this repository.

• resources
  • datasets: Datasets used extensively for this project in some way
  • scripts
    • kg: Scripts related to creation of the knowledge graph using files in mapping and /resources/datasets
      • mapping: YAML files that delineate the relationships of the values in the rows of the CSVs in /resources/datasets
      • kg_output_parser.py: Script that takes TTL files generated by other scripts in kg and combines them into one TTL file
    • web_scrape: Scripts used to generate CSVs in resources/datasets
    • chatbot.py: The main chatbot code
  • conversational_scripts.md: A file containing sample snippets modeling what conversations between the agent and a user could look like
  • potential_datasets.md: A list of datasets found online that could be useful for this project. Confirmed/Integrated datasets were moved into resources/datasets
• schemas
  • convology_schema.graphml: A schema of Convology I created using yEd and the documentation found here
  • convology.owl: Convology's owl file downloaded from here
  • video_game_schema.graphml: A schema I created for this project

Current State and Next Steps

A functional but limited conversational agent has been created. Next steps would entail the following: integration of more data, utilizing more recognized entities and knowledge graph queries, retaining context of conversations and discerning user intent, and displaying responses in natural language.