Exploring the Intersection of Machine Learning and Semantic Web Technologies for Model Explainability


Hatched by Glasp

Aug 25, 2023

3 min read


Exploring the Intersection of Machine Learning and Semantic Web Technologies for Model Explainability


In recent years, the field of machine learning (ML) has gained immense popularity for its predictive capabilities. However, one significant drawback of many ML models is their lack of explainability, especially in high-stakes domains such as healthcare and transportation. In order to address this challenge, researchers have turned to the use of Semantic Web Technologies (SWT) to enhance model explainability by incorporating reasoning on knowledge bases. This article aims to explore the common points between ML and SWT, examine the applications and tasks that benefit from this research field, and discuss the evaluation and presentation of model explanations to users.

ML and SWT: Enhancing Model Explainability:

The combination of ML and SWT offers the potential to create explainable outcomes, which is essential in domains where safety, ethics, and trade-offs are crucial. By leveraging SWT, ML models can provide semantically interpretable explanations that allow for reasoning on knowledge bases. Doran et al. argue that truly explainable systems must incorporate elements of reasoning to create human-understandable and unbiased explanations. This highlights the importance of not only the model itself but also the knowledge and skills of its users.

Supervised Classification and Unsupervised Embedding:

Semantic Web Technologies are primarily used to make two types of ML models explainable: supervised classification tasks using Neural Networks and unsupervised embedding tasks. Neural Networks are the dominant prediction model within the supervised classification group, often incorporating taxonomical information from knowledge bases. On the other hand, embedding methods commonly utilize knowledge graphs. Notably, the use of SWT alongside ML algorithms does not sacrifice performance, as these systems often achieve state-of-the-art results in their respective tasks.

Applications in Healthcare and Recommendation Systems:

The healthcare domain has seen the proposal of many interpretable ML models that utilize taxonomical knowledge to aid performance and interpretability. This is due to the high stakes nature of the field and the availability of medical ontologies. Recommendation systems are also a significant area of research, combining embedding models with knowledge graphs. However, most systems provide static explanations without much user interaction, indicating a need for more adaptive and interactive explanations.

Challenges and Future Directions:

One central challenge in the integration of ML and SWT is knowledge matching, which involves matching ML data with knowledge base entities. Future research needs to focus on developing automated and reliable methods for knowledge matching, as well as finding ways to overcome the potential lack of data interconnectedness and increased system complexity. Additionally, truly explainable systems should incorporate reasoning and external knowledge that is human-understandable, offering users the ability to scrutinize and interact with explanations.

Actionable Advice:

  • 1. Incorporate reasoning: When developing ML models, consider incorporating elements of reasoning that make use of knowledge bases. This can enhance the explainability of the model and provide more human-understandable explanations.
  • 2. Foster user interaction: Aim to create adaptive and interactive explanations that allow users to engage with the system. This can lead to greater user comprehension and overall benefit.
  • 3. Establish evaluation criteria: Work towards establishing common evaluation criteria for model explainability. This will enable more rigorous practices and objective assessments, replacing subjective judgments.


The combination of ML and SWT presents exciting opportunities for enhancing model explainability. By incorporating reasoning and knowledge bases, ML models can provide semantically interpretable explanations, particularly in domains like healthcare and recommendation systems. Future research should focus on overcoming challenges such as knowledge matching and developing adaptive and interactive explanations. By embracing common evaluation criteria and design patterns, meaningful progress can be made in the field of explainable AI.

Hatch New Ideas with Glasp AI 🐣

Glasp AI allows you to hatch new ideas based on your curated content. Let's curate and create with Glasp AI :)