Dipartimento di Ingegneria informatica, automatica e gestionale

Title: Addressing Model and Instance level Heterogeneities for Data Preparation under Knowledge Representation Principles

 

Abstract: Data preparation is the process of collecting, aggregating, transforming, and cleaning raw data to prepare it for future processing and analysis. It is a crucial process in all data-intensive applications, such as analytics or those involving machine learning approaches that are sensitive to low data quality, thus impacting the final result. However, data preparation often encounters challenges stemming from data-model heterogeneity, requiring the integration of sources beyond traditional structured databases. Additionally, data-level issues like entity duplication further complicate the process.

 

In this seminar, I will delve into approaches rooted in knowledge representation techniques and data integration principles to address these challenges. The discussion will revolve around two key topics:

 

1) Extending the data integration paradigm known as Ontology-based data access in order to incorporate semi-structured and unstructured sources (mainly raw text). This extension aims to enhance the integration of diverse data sources, facilitating a more comprehensive data preparation process.

 

2) Introducing a novel approach based on collective, formal, logical, and reasoning-based methods to tackle the complex problem of entity resolution while also focusing on the query answering task. This approach aims to ensure data accuracy and minimize redundancy by effectively addressing entity duplication.

 

The main aim of the techniques that I will present is to improve the efficiency and effectiveness of data preparation, ultimately enhancing the overall quality and reliability of data analysis outcomes.