arXiv subjects

Riccardo Tommasini

Publications and source records attributed to Riccardo Tommasini.

5 recordsLinked to original sources

arXiv2025

Towards a Multimodal Stream Processing System

In this paper, we present a vision for a new generation of multimodal streaming systems that embed MLLMs as first-class operators, enabling real-time query processing across multiple modalities. Achieving this is non-trivial: while recent work has integrated MLLMs into databases for multimodal queries, streaming systems require fundamentally different approaches due to their strict latency and throughput requirements. Our approach proposes novel optimizations at all levels, including logical, physical, and semantic query transformations that reduce model load to improve throughput while preserving accuracy. We demonstrate this with Samsara, a prototype leveraging such optimizations to improve performance by more than an order of magnitude. Moreover, we discuss a research roadmap that outlines open research challenges for building a scalable and efficient multimodal stream processing systems.

cs.DB↗

arXiv2023

Multimodal and Explainable Internet Meme Classification

In the current context where online platforms have been effectively weaponized in a variety of geo-political events and social issues, Internet memes make fair content moderation at scale even more difficult. Existing work on meme classification and tracking has focused on black-box methods that do not explicitly consider the semantics of the memes or the context of their creation. In this paper, we pursue a modular and explainable architecture for Internet meme understanding. We design and implement multimodal classification methods that perform example- and prototype-based reasoning over training cases, while leveraging both textual and visual SOTA models to represent the individual cases. We study the relevance of our modular and explainable models in detecting harmful memes on two existing tasks: Hate Speech Detection and Misogyny Classification. We compare the performance between example- and prototype-based methods, and between text, vision, and multimodal models, across different categories of harmfulness (e.g., stereotype and objectification). We devise a user-friendly interface that facilitates the comparative analysis of examples retrieved by all of our models for any given meme, informing the community about the strengths and limitations of these explainable methods.

cs.AI↗

arXiv2022

PAPyA: Performance Analysis of Large RDF Graphs Processing Made Easy

Prescriptive Performance Analysis (PPA) has shown to be more useful than traditional descriptive and diagnostic analyses for making sense of Big Data (BD) frameworks' performance. In practice, when processing large (RDF) graphs on top of relational BD systems, several design decisions emerge and cannot be decided automatically, e.g., the choice of the schema, the partitioning technique, and the storage formats. PPA, and in particular ranking functions, helps enable actionable insights on performance data, leading practitioners to an easier choice of the best way to deploy BD frameworks, especially for graph processing. However, the amount of experimental work required to implement PPA is still huge. In this paper, we present PAPyA 1, a library for implementing PPA that allows (1) preparing RDF graphs data for a processing pipeline over relational BD systems, (2) enables automatic ranking of the performance in a user-defined solution space of experimental dimensions; (3) allows user-defined flexible extensions in terms of systems to test and ranking methods. We showcase PAPyA on a set of experiments based on the SparkSQL framework. PAPyA simplifies the performance analytics of BD systems for processing large (RDF) graphs.We provide PAPyA as a public open-source library under an MIT license that will be a catalyst for designing new research prescriptive analytical techniques for BD applications.

cs.DB↗

arXiv2021

Semantic Foundations of Seraph Continuous Graph Query Language

The scientific community has been studying graph data models for decades. Their high expressiveness and elasticity led the scientific community to design a variety of graph data models and graph query languages, and the practitioners to use them to model real-world cases and extract useful information. Recently, property graphs and, in particular, Cypher 9 (the first open version of the well-known Neo4j Inc.'s language) are gaining popularity. Practitioners find Cypher useful and applicable in many scenarios. However, we are living in a streaming world where data continuously flows. A growing number of Cypher's users show interest in continuously querying graph data to act in a timely fashion. Indeed, Cypher lacks the features for dealing with streams of (graph) data and continuous query evaluation. In this work, we propose Seraph, an extension of Cypher, as a first attempt to introduce streaming features in the context of property graph query languages. Specifically, we define Seraph semantics, we propose a first version of Seraph syntax, and we discuss the potential impacts from a user perspective.

cs.DB↗

arXiv2020

The Future is Big Graphs! A Community View on Graph Processing Systems

Graphs are by nature unifying abstractions that can leverage interconnectedness to represent, explore, predict, and explain real- and digital-world phenomena. Although real users and consumers of graph instances and graph workloads understand these abstractions, future problems will require new abstractions and systems. What needs to happen in the next decade for big graph processing to continue to succeed?

cs.DC↗