[Paper Review] Question Answering Over Biological Knowledge Graph via Amazon Alexa

Md Rezaul Karim, Hussain Ali+3 more|arXiv (Cornell University)|Oct 12, 2022

Genomics and Rare Diseases4 citations

TL;DR

This paper presents a proof-of-concept integration of Amazon Alexa’s voice interface with the DisGeNET knowledge graph to enable natural language question answering over biological data. By leveraging SPARQL endpoints and custom Alexa skills, the system allows users to query gene-disease associations via voice, achieving functional QA for simple and multi-layered queries, though scalability is limited by execution timeouts in the Alexa console beyond two query layers.

ABSTRACT

Structured and unstructured data and facts about drugs, genes, protein, viruses, and their mechanism are spread across a huge number of scientific articles. These articles are a large-scale knowledge source and can have a huge impact on disseminating knowledge about the mechanisms of certain biological processes. A knowledge graph (KG) can be constructed by integrating such facts and data and be used for data integration, exploration, and federated queries. However, exploration and querying large-scale KGs is tedious for certain groups of users due to a lack of knowledge about underlying data assets or semantic technologies. A question-answering (QA) system allows the answer of natural language questions over KGs automatically using triples contained in a KG. Recently, the use and adaption of digital assistants are getting wider owing to their capability at enabling users to voice commands to control smart systems or devices. This paper is about using Amazon Alexa's voice-enabled interface for QA over KGs. As a proof-of-concept, we use the well-known DisgeNET KG, which contains knowledge covering 1.13 million gene-disease associations between 21,671 genes and 30,170 diseases, disorders, and clinical or abnormal human phenotypes. Our study shows how Alex could be of help to find facts about certain biological entities from large-scale knowledge bases.

Motivation & Objective

• To enable natural language question answering over large-scale biological knowledge graphs using a voice-enabled interface.
• To evaluate the feasibility and limitations of deploying a QA system on Amazon Alexa for biomedical data retrieval.
• To integrate the DisGeNET knowledge graph—containing 1.13 million gene-disease associations—into a voice-based QA pipeline via SPARQL endpoints.
• To identify technical bottlenecks in executing complex, multi-layered queries within the Alexa development environment.
• To explore future enhancements through integration with multiple knowledge graphs and improved NLU support.

Proposed method

• The system uses Amazon Alexa’s custom skill framework to process natural language queries and route them to a SPARQL endpoint hosting the DisGeNET knowledge graph.
• Simple intents (e.g., disease definition, causation) are implemented as single-round-trip queries to the KG via SPARQL, with results formatted for voice output.
• For complex queries, multi-layered reasoning is implemented by chaining multiple API calls: the result of one query is used as input for the next, simulating graph traversal.
• The system leverages RDF triples and SPARQL queries to extract structured facts from DisGeNET, including gene-disease associations and associated scores.
• Custom handlers in the Alexa skill manage intent recognition, slot filling (e.g., disease name), and result formatting for voice delivery.
• The implementation is tested both locally (where complex queries work) and in the Alexa developer console (where timeouts and crashes occur beyond two query layers).

Experimental results

Research questions

• RQ1 Can a voice-enabled interface like Amazon Alexa effectively support natural language question answering over large-scale biological knowledge graphs?
• RQ2 How does the complexity of SPARQL queries affect the reliability and performance of QA systems when deployed on the Alexa platform?
• RQ3 What are the technical limitations of executing multi-layered queries in the Alexa developer console compared to local execution environments?
• RQ4 To what extent can existing biomedical knowledge graphs like DisGeNET be effectively exposed through voice-based QA systems?
• RQ5 How can the scalability and robustness of voice-based QA over KGs be improved for real-world biomedical research applications?

Key findings

• The system successfully implements voice-based QA for simple queries, such as disease definitions and gene-disease causation, using SPARQL endpoints and Alexa’s intent framework.
• Simple queries executed via a single API call to the KG work reliably both locally and in the Alexa console, demonstrating functional feasibility.
• Multi-layered queries—where the result of one query is used as input for another—work correctly in local testing environments, enabling deep reasoning over the KG.
• However, when deployed in the Alexa developer console, the system fails to execute queries beyond two layers due to execution timeouts and unhandled crashes.
• The root cause appears to be platform-specific limitations in code formatting and execution environment, not logic errors, as the same code runs without issues locally.
• The study identifies a critical gap in the Alexa platform’s support for complex, chained queries, which limits the system’s scalability for advanced biomedical queries.