Ontology-based question answering over corporate structured data
This work addresses the need for explainable and data-efficient question answering systems in corporate settings, though it appears incremental by combining existing ontology and graph-based techniques.
The paper tackles the problem of improving question answering quality in dialogue systems by using an ontology-based approach to transform user input into SPARQL queries for knowledge graphs, enabling fact extraction from corporate structured data without requiring large datasets. It describes a chat bot engine that maintains conversation context and asks clarifying questions, offering better explainability compared to neural net methods.
Ontology-based approach to the Natural Language Understanding (NLU) processing allows to improve questions answering quality in dialogue systems. We describe our NLU engine architecture and evaluate its implementation. The engine transforms user input into the SPARQL SELECT, ASK or INSERT query to the knowledge graph provided by the ontology-based data virtualization platform. The transformation is based on the lexical level of the knowledge graph built according to the Ontolex ontology. The described approach can be applied for graph data population tasks and to the question answering systems implementation, including chat bots. We describe the dialogue engine for a chat bot which can keep the conversation context and ask clarifying questions, simulating some aspects of the human logical thinking. Our approach uses graph-based algorithms to avoid gathering datasets, required in the neural nets-based approaches, and provide better explainability of our models. Using question answering engine in conjunction with data virtualization layer over the corporate data sources allows extracting facts from the structured data to be used in conversation.