projects

DTALab develops trustworthy data-centric AI systems that connect models, data, evidence and explanations. Our research focuses on NLP/LLMs, evidence-grounded reasoning, explainable and auditable decision support, semantic robustness, reproducible benchmarking, data integration, fairness-aware learning and financial AI. We design AI pipelines whose behaviour can be inspected, stress-tested and validated by domain experts, with applications in finance, compliance, education, information integrity, public-sector services and data-intensive organisational workflows.

interests

Evidence-grounded NLP and LLMs

NLP and LLM systems that connect answers, claims and decisions to explicit evidence.

Interpretable Time-series Analysis

Methods for interpretable clustering, novelty detection, self-supervised learning and forecasting of multivariate time series.

Data Integration and Explainable Entity Matching

Methods for matching, linking, cleaning and explaining heterogeneous records in data integration pipelines.

Fairness, Auditability and Responsible AI

Methods and tools for evaluating, mitigating and auditing fairness harms in machine-learning systems.

Semantic Robustness and Red-teaming

Methods for testing the robustness of NLP and LLM systems against adversarial triggers, semantic perturbations and manipulative formulations.

Financial AI and Decision Support

Reproducible benchmarking and portfolio-oriented evaluation for stock market prediction models.

projects

Panacea

PRIN PROGETTI DI RICERCA DI RILEVANTE INTERESSE NAZIONALE Bando 2022

Resist0

PR FESR 2021-27 EMILIA-ROMAGNA

past

KEYSTONE

COST Action IC1302

Re-search Alps

CEF Telecom