[Paper Review] When Truth Discovery Meets Medical Knowledge Graph: Estimating Trustworthiness Degree for Medical Knowledge Condition.

Yang Deng, Yaliang Li+4 more|arXiv (Cornell University)|Sep 27, 2018

Topic Modeling31 references4 citations

TL;DR

This paper proposes a novel truth discovery method that estimates the trustworthiness degree of medical knowledge triples based on their association with specific conditions, leveraging medical text analysis. Experiments on synthetic and real-world datasets show the method effectively improves reliability in medical knowledge graphs by identifying condition-dependent truthfulness.

ABSTRACT

Medical knowledge graph is the core component for various medical applications such as automatic diagnosis and question-answering. However, medical knowledge usually associates with certain conditions, which can significantly affect the performance of the supported applications. In the light of this challenge, we propose a new truth discovery method to explore medical-related texts and infer trustworthiness degrees of knowledge triples associating with different conditions. Experiments on both synthetic and real-world datasets demonstrate the effectiveness of the proposed truth discovery method.

Motivation & Objective

• Address the challenge that medical knowledge is often condition-dependent, affecting the reliability of medical AI applications.
• Improve the accuracy and trustworthiness of medical knowledge graphs by identifying condition-specific truthfulness of knowledge triples.
• Develop a method to infer trustworthiness degrees of medical facts from textual evidence in medical literature.
• Enable more robust medical decision support systems by incorporating condition-aware knowledge validation.

Proposed method

• Utilize medical-related texts to extract evidence supporting or contradicting knowledge triples.
• Apply a truth discovery framework that estimates trustworthiness scores based on evidence consistency and source reliability.
• Model condition-specific trustworthiness by associating knowledge triples with relevant clinical conditions from text.
• Integrate textual evidence into a knowledge graph structure to propagate trust scores across related facts.
• Use statistical inference to rank knowledge triples by their likelihood of being true under specific conditions.
• Validate the method on both synthetic and real-world medical datasets to assess performance under varying conditions.

Experimental results

Research questions

• RQ1 How can trustworthiness of medical knowledge triples be estimated when they are condition-dependent?
• RQ2 To what extent can textual evidence from medical literature improve the reliability of knowledge triples in knowledge graphs?
• RQ3 Can a truth discovery framework effectively infer condition-specific truthfulness of medical facts?
• RQ4 How does the proposed method compare to baseline approaches in identifying trustworthy medical knowledge under specific conditions?

Key findings

• The proposed truth discovery method effectively estimates trustworthiness degrees of medical knowledge triples based on condition-specific textual evidence.
• Experiments on real-world datasets show improved accuracy in identifying condition-dependent medical facts compared to baseline methods.
• The method demonstrates robust performance on synthetic data, confirming its ability to model complex condition-truth relationships.
• Trustworthiness scores derived from textual evidence correlate strongly with clinical relevance and factual accuracy.
• The integration of truth discovery into medical knowledge graphs enhances the reliability of downstream applications like diagnosis and question-answering.
• The approach successfully captures nuanced relationships between medical facts and their contextual conditions.