Tools

LogNet (Local and Global Triple Embedding Network): LoGNet (Local and Global Triple Embedding Network) is a learning framework for triple-centric tasks in knowledge graphs (KGs). LoGNet is based on graph neural networks (GNNs) and combines local and global triple embedding information. Local triple embeddings are learned by treating triples as sequences. Global triple embeddings are learned by operating on the feature triple line graph L(G) of a knowledge graph G. The code is available at: https://github.com/giuseppepirro/lognet

Triple2Vec (Computing Triple Embeddings), with V. Fionda (U. of Calabria). This tool implements an approach to compute triple embeddings from knowledge graphs. The code is available at: https://github.com/giuseppepirro/triple2vec

SADDEN (Safeness-based Community Deception), with V. Fionda (U. of Calabria). This tool implements safeness-based community deception. The code is available at: https://github.com/giuseppepirro/com-deception

RECAP (Relatedness explanations). Knowledge graphs (KGs) are a key ingredient to complement search results, discover entities and their relations and support several knowledge discovery tasks. We face the problem of building relatedness explanations, that is, graphs that can explain how a pair of entities is related in a KG. Explanations can be used in a variety of tasks; from exploratory search to query answering. The code is available at: https://github.com/giuseppepirro/RECAP-REWORD

swget (Semantic wget), with V. Fionda (FUB), C. Gutierrez (UChile). The Web of Data can be modelled as a semantic graph with its nodes being RDF data sources and its edges RDF predicates. We defined a graph navigational language for the Web of Data called NautiLOD. NautiLOD offers the following features: (i) it provides a way to recursively navigate and retrieve data sources from the Web of Data; (ii) it leverages regular expressions to declaratively specify fragments of the Web of Data; (iii) regular expressions can be intertwined with triggers on the form of ASK-SPARQL queries to control and orient the navigation; (iv) it enables to command actions during the navigation such as, send me by email a specific piece of information if encountered during the navigation. The swget tool is an implementation of NautiLOD. The system is available at http://swget.wordpress.com. The project has been partially founded (August 2001- April 2013) by Marie Curie action IRSES- Net2: A Network for Enabling Networked Knowledge.

MaGE (Map Generator), with V. Fionda (FUB), C. Gutierrez (UChile). This project investigates the applicability of cartographic principles to the universal space of information represented by the Web. We provide a formalization of the notions of region and map for this space, suitable to be interpreted and manipulated in a semi-automated way. We investigate the problem of extracting regions of interest from the Web, and introduce algorithms to efficiently generate maps from these regions. To show the feasibility of the proposal we developed the MaGE a tool that implements the map framework. The system is available at http://mapsforweb.wordpress.com. The project has been partially founded (August 2001- April 2013) by Marie Curie action IRSES- Net2: A Network for Enabling Networked Knowledge.

GenTLE, with M. Consens (UoT), V. Fionda (FUB). GenTLE is a graph traversal language that targets Web data and offers a novel and attractive expressiveness/complexity trade-off. GenTLE expressions are evaluated on discoverable graphs. The language retains the low polynomial time (data and query) combined complexity of Nested Regular Expression languages while significantly extending their expressibility with support for path conjunction, path negation, and the output of explanation sub-graphs as answers. Implementation in progress.

GuLP (Graph Query Language with Preferences), with V. Fionda (FUB). Despite the large variety of graph query languages, little has been done in terms of functionalities that enable users to declaratively express preferences. Having preferences in a graph query language enables to drive the evaluation to consider some paths before than others thus introducing an ordering on the way results are collected and presented. This functionality is useful when one is interested in certain kinds of results more than in others and is crucial when dealing with large graphs. GuLP enables to express preferences in terms of nodes/edge attributes and complex paths. Besides, GuLP expressions can be evaluated in polynomial combined complexity. The system is available at http://graphgulp.wordpress.com