PostDoc Researcher · BIFOLD, TU Berlin
Guido Rocchietti
I work in the Data Integration and Data Preparation (D2IP) group, on Information Integration and Data Cleansing. My research centres on Natural Language Processing, Large Language Models and Information Retrieval, and extends to time-series modelling, in particular for medical data. I hold a Ph.D. in AI for Society from the University of Pisa and ISTI-CNR, with a thesis on efficient methods for conversational search and summarization.
Large Language ModelsConversational SearchInformation RetrievalData IntegrationText SummarizationTime Series
Research
My research focuses on Artificial Intelligence and Natural Language Processing. I was involved in the following projects:
Tech stack
PythonPyTorchTensorFlowHugging Face Transformerspandas
Publications
-
Quantifying Retriever-Generator Alignment in RAG with Local Explanations
Findings of ACL: EMNLP 2026 PDF
-
FoodSafeSum: Enabling Natural Language Processing Applications for Food Safety Document Summarization and Analysis
Findings of ACL: EMNLP 2025 PDF
-
Efficient and Effective Methods for Conversational Search and Summarization
PhD Thesis, University of Pisa & ISTI-CNR PDF
-
ChatGPT Versus Modest Large Language Models: An Extensive Study on Benefits and Drawbacks for Conversational Search
IEEE Access, 13 DOI
-
Efficient Conversational Search via Topical Locality in Dense Retrieval (short paper)
SIGIR '25 DOI
-
Human Mobility Datasets Enriched With Contextual and Social Dimensions
arXiv:2510.02333 arXiv
-
From Geolocated Images to Urban Region Identification and Description: a Large Language Model Approach (short paper)
ACM SIGSPATIAL '24 DOI
-
LongDoc Summarization using Instruction-tuned Large Language Models for Food Safety Regulations
IIR 2024, 14th Italian Information Retrieval Workshop
-
Enhancing Conversational Search with Large Language Models
ERCIM News, 136
-
Rewriting Conversational Utterances with Instructed Large Language Models
IEEE/WIC WI-IAT 2023 DOI
-
Commonsense Injection in Conversational Systems: An Adaptable Framework for Query Expansion
IEEE/WIC WI-IAT 2023 DOI
-
FANCY: A Diagnostic Data-Set for NLI Models
CLiC-it 2021, Eighth Italian Conference on Computational Linguistics DOI
Research software
pip install rag-exp
EMNLP 2026 (Findings)
efficient_retrieval
Algorithms supporting efficient conversational retrieval.
SIGIR '25
llms4queryrewriting
Extensive study on the benefits and drawbacks of using LLMs for query rewriting.
IEEE Access 2025
UrbanRegionDescription
NLP methodology for the automatic identification of urban regions from geolocated images.
SIGSPATIAL '24
conv-common-sense
Adaptable framework for query expansion using commonsense knowledge.
WI-IAT 2023
FANCY
Diagnostic dataset for evaluating Natural Language Inference (NLI) models.
CLiC-it 2021
Contact
Write to guido.rocchietti@tu-berlin.de or guido.rocchietti@gmail.com.