Jesus A. Rodriguez Perez

3Machine learning papers tracked by Lacuna
Is this you?

Research Overview

Jesus A. Rodriguez Perez focuses on information retrieval, specifically addressing the challenges of microblog search and query performance prediction. During his observed active period, his work explored methods to bridge vocabulary mismatches in short text and improve the reliability of automatic query expansion. His most recent contributions involve real-time event detection and visualization interfaces for social media streams.

Research Trajectory

The available evidence for Rodriguez Perez's research trajectory is sparse, concentrated entirely in a brief period between 2013 and 2014. During this time, his work focused tightly on the intersection of core information retrieval mechanics and the unique challenges posed by social media streams.

Rodriguez Perez's early work centers on the difficulties of ad-hoc retrieval in microblog environments, where high vocabulary diversity creates significant mismatches between user queries and relevant posts. As the first author of On using inter-document relations in microblog retrieval (2013), he proposed a re-ranking approach that leverages inter-document relations to improve search effectiveness. This work contributes to the broader research direction of Bridging the Lexical Gap in Information Retrieval, demonstrating that incorporating structural relationships between documents can significantly enhance precision and mean average precision (MAP) when the initial baseline retrieval is sufficiently strong.

Building on the challenges of microblog search, his subsequent research investigated how to make retrieval enhancements more reliable. In Predicting query performance in microblog retrieval (2014), Rodriguez Perez led an investigation into Query Performance Prediction (QPP) specifically tailored for microblogs. The study addressed the unreliability of Automatic Query Expansion (AQE) when the initial set of retrieved documents is of poor quality. By estimating retrieval success prior to expansion, this approach aligns with the area of Robust Pseudo Relevance Feedback for Information Retrieval and themes such as Query Adaptive Retrieval Strategy Selection. The goal was to selectively apply pseudo-relevance feedback only when the initial results were robust enough to support it, thereby improving the overall stability of the retrieval system.

In addition to core retrieval algorithms, Rodriguez Perez participated in collaborative efforts to make social media data more accessible to end users. As a middle author on An interactive interface for visualizing events on Twitter (2014), he helped develop a system designed to detect, track, and summarize breaking news and ongoing events in real-time. While his primary focus remained on the underlying retrieval mechanics, this contribution highlights an application-oriented extension of his work, addressing information overload through interactive categorization and temporal distribution visualizations.

Current Directions

The most recent papers in the provided evidence date to 2014, reflecting a concentrated focus on optimizing search and event detection in microblog environments. Based on this period, the work investigates the intersection of query performance prediction and real-time social media analysis. Open problems addressed in these final observed publications include the selective application of query expansion techniques to prevent semantic drift, and the development of scalable interfaces to manage information overload during breaking news events.

Profile updated

See something wrong? Report a mistake