MARGOT
A web server for argumentation mining designed for the end-user.
ARGUMENTATION MINING
Argumentation mining has recently become a hot topic in AI, addressing the highly challenging task of
automatically extracting structured arguments from unstructured textual corpora.
DEVELOPMENTAL VISUAL AGENTS
Although computer vision has recently seen a period of great development, Performing visual scene understanding in unrestricted videos
still remains an open challenge. In the Artificial Intelligence Group at the University of Siena I worked at the project of Developmental Visual Agents (DVA),
that are visual systems implementing a never-ending learning process from any kind of videos, continuously interacting with humans to ask for supervisions.
The interaction protocol implemented by these agents is named "Learning to see like children".
MLOCSR
The recognition of the structure of a chemical molecule directly from an image is an extremely challenging task.
Basically, it can be seen as a special OCR for molecular graphs rather than for characters. We developed a system (available as a webserver)
which combines pattern recognition techniques (for point/line/text extraction) with Markov Logic for the assembly of the final chemical graph.
METALDETECTOR
Metals play a crucial role in the life of the cell, as they can provide catalytic, regulatory or structural roles
critical to protein function, participate in a variety of biological processes, and are implicated in many diseases,such as Parkinson and Alzheimer.
Correctly identifying metalloproteins and residues bonded to metal ions is therefore a crucial task in bioinformatics.
MetalDetector is a web server which can predict metal binding sites directly from protein sequence.
GS-MLNs
Grounding-Specific Markov Logic Networks (GS-MLNs) are an extension of Markov Logic Networks which allows to embed discriminative
classifiers such as neural networks within the framework of Markov logic, therefore creating a hybrid model which can handle vectors of
continuous features.
TETs
Type Extension Trees (TETs) are a powerful representation language for "count-of-count" features characterizing the combinatorial structure of neighborhoods of entities in relational domains.
eSFBDS
Efficient single frontier bidirectional search is a very efficient heuristic search algorithm, which performs
a bidirectional search on a double node search tree, where each double node includes two states, one for the forward search
direction, and the other for the backward direction.
An implementation of the eSBS algorithm can be downloaded for the 15 puzzle, but the code can be easily adapted to other puzzles
and search problems (see our SoCS 2012 paper). For any further question, or for an implementation
of eSBS on other domains, feel free to contact me by e-mail.