Pietro Totis

Pietro Totis

Senior AI Research Scientist

J.P. Morgan, Madrid

Biography

I am currently part of the AI Research team at JP Morgan Chase & Co. as Senior Research Associate. I obtained my PhD in Computer Science at the DTAI Lab of the Department of Computer Science of KU Leuven, under the supervision of professor Luc De Raedt. My research was also supervised by professors Jesse Davis and Angelika Kimmig (DTAI Lab).

My main PhD project was Solving Combinatorial and Probabilistic Problems in Natural Language, which aims to build a system capable of understanding and solving probability and combinatorial problems expressed in natural language. Other lines of research included Probabilistic Logic Programming, Knowledge Compilation and Argumentation.

My Bachelor and Master education took place in Udine (Italy), my hometown, with an Erasmus period of 6 months at the University of Potsdam.

Interests
  • (Probabilistic) Logic Programming (PLP)
  • Knowledge Representation and Reasoning (KRR)
  • Lifted Inference
  • Natural Language Processing (NLP)
  • Argumentation
Education

  • PhD in Artificial Intelligence, Nov 2018 - Mar 2023

    KU Leuven

  • MSc in Computer Science, Oct 2016 - Oct 2018

    Università degli Studi di Udine

  • BSc in Computer Science, Oct 2013 - Oct 2016

    Università degli Studi di Udine

Academic

 
 
 
 
 
PhD in Computer Science
DTAI Lab @KU Leuven
Nov 2018 – Present Belgium

Also Teaching Assistant for the courses:

  • Databases (BSc)
  • Advanced Programming Languages in AI (MSc)
 
 
 
 
 
Erasmus
Potsdam Universität
Sep 2017 – Feb 2018 Germany

Courses from the Master in Cognitive Systems:

  • Advanced Problem Solving Techniques
  • Argumentation Mining
  • Introduction to Deep Learning
  • Natural Language Processing
 
 
 
 
 
MSc Computer Science
Università degli studi di Udine
Oct 2016 – Oct 2018 Italy

Languages and Algorithms path:

  • Advanced Algorithms
  • Algorithms and Complexity
  • Applied Statistics and Data Analisys
  • Advanced Databases
  • Automated Reasoning
  • Automated System Verification: Theory and Applications
  • Models and Algorithms for Decisions
  • Network science
 
 
 
 
 
BSc Computer Science
Università degli studi di Udine
Oct 2013 – Oct 2016 Italy

Projects

Solving Combinatorial and Probabilistic Problems in Natural Language
This project aims to build a system capable of solving probabilistic and combinatorial problems from questions formulated in natural language. We are working on a dataset composed of common textbook problems that cover a wide range of situations and structures.
Solving Combinatorial and Probabilistic Problems in Natural Language

Publications

*
All Journal Conference KRR NLP PLP Argumentation
Lifted Reasoning for Combinatorial Counting
In this paper we tackle the problem of automating the resolution of combinatorics math word problems. We introduce CoLa, a novel declarative language to express combinatorics math word problems, and CoSo, a solver based on lifted reasoning, which outperforms alternative frameworks.
Pietro Totis, Jesse Davis, Luc De Raedt, Angelika Kimmig
PDF Code DOI URL
Efficient Knowledge Compilation Beyond Weighted Model Counting
We introduce second level algebraic model counting (2AMC) problems, a framework generalizing several probabilistic inference task. We present a novel Knowledge Compilation technique to address the increased complexity of a 2AMC task with respect to first-level AMC problems.
Rafael Kiesel, Pietro Totis, Angelika Kimmig
PDF Cite Code DOI URL
Mapping Probability Word Problems to Executable Representations
We analyze different neural models to solve probability math word problems in two ways. First, to predict directly the answer in an end-to-end fashion. Second, to map the text to a formal representation used by a probabilistic programming system to compute the answer.
Simon Suster, Pieter Fivez, Pietro Totis, Angelika Kimmig, Jesse Davis, Luc De Raedt, Walter Daelemans
PDF Cite Project Video DOI URL
smProbLog: Stable Model Semantics in ProbLog for Probabilistic Argumentation
We model beliefs in argumentation problems with probabilistic logic programs and show that traditional probabilistic logic programming (PLP) systems cannot reason on this type of programs. We thus present smProblog, a novel PLP system based on ProbLog, where inference and learning over such probabilistic argumentation problems are possible.
Pietro Totis, Angelika Kimmig, Luc De Raedt
PDF Cite Code DOI URL
Mining Italian Short Argumentative Texts
We present the first model for argumentation mining for Italian short argumentative texts. We adapted to Italian the software …
Ivan Namor, Pietro Totis, Samuele Garda, Manfred Stede
PDF Cite
Towards Distributed Computation of Answer Sets
Answer Set Programming (ASP) is a logic programming language widely used in non-monotonic automated reasoning. Thanks to its …
Marco De Bortoli, Federico Igne, Fabio Tardivo, Pietro Totis, Agostino Dovier, Enrico Pontelli
PDF Cite
Classifying Italian newspaper text: news or editorial?
We present a text classifier that can distinguish Italian news stories from editorials. Inspired by earlier work on English, we built a …
Pietro Totis, Manfred Stede
PDF Cite

Recent & Upcoming Talks

Declarative Modelling and Reasoning for Combinatorial Problem Solving and Argumentation under Uncertainty
Mar 8, 2023 1:30 PM — 3:30 PM aula van de Tweede Hoofdwet, 01.02
PDF Slides
Reasoning on Arguments and Beliefs with Probabilistic Logic Programs
Dec 8, 2022 3:00 PM — 4:00 PM CLArg Group, Department of Computing
PDF Slides
Lifted Reasoning for Combinatorial Counting
A presentation on my research about Lifted Reasoning for solving combinatorics math word problems.
Oct 25, 2022 4:30 PM — 5:30 PM Room 01.27
PDF Slides
Stable Model Semantics in ProbLog and its Applications in Argumentation
A presentation on my research about Probabilistic Logic Programming and Argumentation
Feb 21, 2022 4:00 PM — 5:00 PM Online event
PDF

Contact

  • Mail
  • P.º de la Castellana, 31, Edificio Pirámide, Madrid, 28046