Andrea Galassi Home Page
Andrea Galassi Home Page
Andrea Galassi 3rd year PhD student DISI, UniversitĂ di Bologna email: a<dot>galassi<at>unibo<dot>it |
Profiles: |
I am a second year PhD student at the Computer Science and Engineer department (DISI) of the University of Bologna. My supervisor is Professor Paolo Torroni.
My PhD is focused on the application of Artificial Intelligence and Machine Learning techniques to Argumentation Mining and similar Natural Language Processing tasks.
My research interests involve also the investigation of what Deep Networks can achieve by themselves and possible ways to combine symbolic and sub-symbolic techniques. So far, we have conducted our investigation on board games and Constraint Satisfaction Problems.
I've been teaching assistant ("Tutor Didattico") of professors Paola Mello, Federico Chesani, and Paolo Torroni. I am one of the responsible of the Board Game Students Challenge. I am also co-supervisor in some bachelor thesis, masters thesis, and course projects, most of them regarding machine learning application.
I have visited Stanford University during summer 2018, working under the supervision of Margaret Hagan in the context of the MIREL project.
Recent works and news:
- Dec 2020: Our paper "Cross-lingual Annotation Projection in Legal Texts" has been presented at COLING 2020. Our poster can be found here.
- Sep 2020: Our paper "Neural-Symbolic Argumentation Mining: an Argument in Favor of Deep Learning and Reasoning" will be presented in the Sister and Journal Track of the ICLP2020 conference.
- Aug 2020: Our paper "Attention in Natural Language Processing" has been accepted for publication in IEEE Transactions on Neural Networks and Learning Systems. A presentation on the topic is available here.
- Jan 2020: Our paper "Neural-Symbolic Argumentation Mining: an Argument in Favor of Deep Learning and Reasoning" has been published on "Frontiers in Big Data". It is available for free on the website of the journal.
- Jun 2019: Our abstract "Imitation or Understanding? Deep Learning to the Test on Constraint Satisfaction Problems" has been presented at EURO19.
- Mar 2019: Our paper "HABITAT: An IoT Solution for Independent Elderly" has been published on "Sensors". It is freely available on the website of the journal.
- Feb 2019: Our paper "Attention, please! A Critical Review of Neural Attention Models in Natural Language Processing" is under peer review. A preliminary version is available on arXiv. A presentation on the topic is available here.
- Jun 2018: Our short paper "Model agnostic solution of CSPs via Deep Learning: a preliminary study" has been presented at CPAIOR 2018. The free pre-publish version can be found here.
- Feb 2018: Our extended abstract "Deep Neural Networks for Constraint Satisfaction Problems: An Initial Investigation for the N-Queens" has been presented at OLA18. A presentation on the topic can be found here.
- Feb 2018: Our paper "Can Deep Networks Learn to Play by the Rules? A Case Study on Nine Men's Morris" has been published on "IEEE Transaction on Games". The accepted version can be found here.
- Nov 2017: Our paper "A Game-Based Competition as Instrument for Teaching Artificial Intelligence" has been presented at AI*IA 2017. The free pre-publish version can be found here.
Projects in which I am or I've been involved:
- Claudette: Machine Learning Powered Analysis of Consumer Contracts and Privacy Policies
- Argumentation Mining: extraction of arguments from unstructured textual documents
- Games and AI: application of Artificial Intelligence techniques in the context of games
- Habitat: Home Assistance Based on the Internet of Things for the AuTonomy
- DeepOpt
Teaching
TeachingI am currently covering the role of teaching assistant for the following courses:
"Languages and Algorithms for Artificial Intelligence", prof. Paolo Torroni.
"Natural Language Processing", prof. Maurizio Gabbrielli.
In the past, I have covered the role of teaching assistant for the following courses:
"Real Time OS M - Real Time Systems for Automation M", Prof. Paolo Torroni, material on the website: https://iol.unibo.it/course/view.php?id=40331
"Fondamenti di Informatica T1", Prof. Paola Mello e Federico Chesani, material on the website: https://iol.unibo.it/course/view.php?id=41087
Lectures Material
Natural Language Processing and Learning (Bologna, 12/04/2019, as part of the Intelligent Systems M course of Prof. Michela Milano)
Thesis and project activities topics
I also supervise bachelor thesis, master thesis and project activities.
Possible topics are listed here, others may be available in the AI course website: http://lia.deis.unibo.it/Courses/AI/fundamentalsAI2018-19/Progetti.html
Topics:
- Neural networks applied to games: application of neural networks and deep networks for the solution of board games and (small/simple) video games.
- Solving/find the complexity of a board game: make use of AI technique to compute the complexity of a board game or to solve it (to a certain degree)
- Natural Language Processing: projects and thesis on argumentation mining (finding the discussion inside a text), sentiment analysis (finding the "sentiment" in a text), fake news identification (from text), cyberbullying identification (from text)
- Deep networks for CSPs: application and analysis of deep networks for the solution of constrained problems.
Specific projects:
- Creation of a Keyforge AI player: design and implement an AI to play the game Keyforge card game. A possible implementation of the game can be found here: https://github.com/keyteki/keyteki.
- Creation of predictor for Keyforge match outcome: create an AI module capable of predicting the outcome of a Keyforge game. An existing dataset of games can be used for evaluation (and for training in case of machine learning approaches).
- Creation of a Keyforge deck evaluation: create a machine-learning based system capable of automatically score a deck of Keyforge according to popular metrics (for example, SAS and AERC, but other suggestions are welcomed)
Past activities
Some of the activities I have supervised are the following:
- Mihail Bida, Francesco Pandol, "Framework per giocare ad Hanabi e implementazione di diverse strategie di gioco", 2020. Code repository.
- Andrea Conti, project activity, "Deep Q-Learning applicato a Super Mario Bros", 2019. Code repository.
- Francesco Giovanelli, master thesis, "Model Agnostic solution of CSPs with Deep Learning", 2019. Code repository.
- Giacomo Pinardi, bachelor thesis, "Apprendimento supervisionato di un gioco da tavolo asimmetrico tramite reti neurali: un caso di studio su Tablut", 2019
- Alessio Leurini, bachelor thesis, "Cross-Domain Sentiment Analysis", 2019
- Alessandro Ravaglia, bachelor thesis, "Studio delle tecniche di reinforcement learning e delle loro applicazioni ai giochi da tavolo", 2019
- Guardati Simone, project activity, "Applicazione di tecniche di Machine Learning al gioco Mastermind", 2019
- Andrea Piretti, bachelor thesis, "Sviluppo di un'architettura software distribuita con supporto a giocatori artificiali: il caso di studio del gioco da tavolo Tablut", 2018
- Nicola Alessi, project activity, 2018
- Grilli Matteo, bachelor thesis, "Reti neurali profonde applicate a giochi di carte digitali: un caso di studio su Hearthstone semplificato", 2017
Publications and Research Material
Publications and Research MaterialNLP
2020, COLING: Cross-lingual Annotation Projection in Legal Texts
2020, IEEE Transactions on Neural Networks and Learning Systems: Attention in Natural Language Processing
2020, Frontiers on Big Data: Neural-Symbolic Argumentation Mining: an Argument in Favor of Deep Learning and Reasoning
2018, 5th Workshop on Argument Mining: Argumentative Link Prediction using Residual Networks and Multi-Objective Learning
Deep Networks and CSPs
2018, CPAIOR: Model agnostic solution of CSPs via Deep Learning: a preliminary study
Games
2019, technical report: An Upper Bound on the Complexity of Tablut
2018, IEEE Transactions on Games: Can Deep Networks Learn to Play by the Rules? A Case Study on Nine Men’s Morris
Education
2017, AI*IA: A Game-Based Competition as Instrument for Teaching Artificial Intelligence
IoT
2019, Sensors: HABITAT: An IoT solution for independent elderly