Omar Alonso is a Senior Science Manager at Amazon. He has extensively worked on a variety of topics in information retrieval, content analysis and NLP. One of his paper has won the ECIR Test-of-Time Award. Omar serves as Program Co-Chair of the 2025 ACM SIGIR Conference.
There is a lot of interest in knowledge graphs as a rich structure that can be used in many information retrieval applications like search, recommendations, and question-answering. While there are examples of knowledge graphs in industry and academia, there is little information on the practical aspects of design and construction. Accuracy evaluation of knowledge graphs has received limited attention in prior research. In this talk, we present a perspective on data quality and its implications for downstream applications. We introduce a utility-oriented accuracy evaluation framework capable of scaling with limited human annotations.