Probabilistic Models for Scalable Knowledge Graph Construction

In the past decade, systems that extract information from millions of Internet documents have become commonplace. Knowledge graphs -- structured knowledge bases that describe entities, their attributes and the relationships between them -- are a powerful tool for understanding and organizing this vast amount of information. However, a significant obstacle to knowledge graph construction is the unreliability of the extracted information, due to noise and ambiguity in the underlying data or errors made by the extraction system and the complexity of reasoning about the dependencies between these noisy extractions. My dissertation addresses these challenges by exploiting the interdependencies between facts to improve the quality of the knowledge graph in a scalable framework. I introduce a new approach called knowledge graph identification (KGI), which resolves the entities, attributes and relationships in the knowledge graph by incorporating uncertain extractions from multiple sources, entity co-references, and ontological constraints. I define a probability distribution over possible knowledge graphs and infer the most probable knowledge graph using a combination of probabilistic and logical reasoning. Such probabilistic models are frequently dismissed due to scalability concerns, but my implementation of KGI maintains tractable performance on large problems through the use of hinge-loss Markov random fields, which have a convex inference objective. This allows the inference of large knowledge graphs using 4M facts and 20M ground constraints in 2 hours. To further scale the solution, I develop a distributed approach to the KGI problem which runs in parallel across multiple machines, reducing inference time by 90%. Finally, I extend my model to the streaming setting, where a knowledge graph is continuously updated by incorporating newly extracted facts. I devise a general approach for approximately updating inference in convex probabilistic models, and quantify the approximation error by defining and bounding inference regret for online models. Together, my work retains the attractive features of probabilistic models while providing the scalability necessary for large-scale knowledge graph construction. These models have been applied on a number of real-world knowledge graph projects, including the NELL project at Carnegie Mellon and the Google Knowledge Graph.

Supporting requirement elicitation and ontology testing in knowledge graph engineering

Knowledge graphs and ontologies are closely related concepts in the field of knowledge representation. In recent years, knowledge graphs have gained increasing popularity and are serving as essential components in many knowledge engineering projects that view them as crucial to their success. The conceptual foundation of the knowledge graph is provided by ontologies. Ontology modeling is an iterative engineering process that consists of steps such as the elicitation and formalization of requirements, the development, testing, refactoring, and release of the ontology. The testing of the ontology is a crucial and occasionally overlooked step of the process due to the lack of integrated tools to support it. As a result of this gap in the state-of-the-art, the testing of the ontology is completed manually, which requires a considerable amount of time and effort from the ontology engineers. The lack of tool support is noticed in the requirement elicitation process as well. In this aspect, the rise in the adoption and accessibility of knowledge graphs allows for the development and use of automated tools to assist with the elicitation of requirements from such a complementary source of data. Therefore, this doctoral research is focused on developing methods and tools that support the requirement elicitation and testing steps of an ontology engineering process. To support the testing of the ontology, we have developed XDTesting, a web application that is integrated with the GitHub platform that serves as an ontology testing manager. Concurrently, to support the elicitation and documentation of competency questions, we have defined and implemented RevOnt, a method to extract competency questions from knowledge graphs. Both methods are evaluated through their implementation and the results are promising.

Graph-based Vision-Language Model for Multi-Modal Retrieval of Fashion Products

Artificial Intelligence is reshaping the field of fashion industry in different ways. E-commerce retailers exploit their data through AI to enhance their search engines, make outfit suggestions and forecast the success of a specific fashion product. However, it is a challenging endeavour as the data they possess is huge, complex and multi-modal. The most common way to search for fashion products online is by matching keywords with phrases in the product's description which are often cluttered, inadequate and differ across collections and sellers. A customer may also browse an online store's taxonomy, although this is time-consuming and doesn't guarantee relevant items. With the advent of Deep Learning architectures, particularly Vision-Language models, ad-hoc solutions have been proposed to model both the product image and description to solve this problems. However, the suggested solutions do not exploit effectively the semantic or syntactic information of these modalities, and the unique qualities and relations of clothing items. In this work of thesis, a novel approach is proposed to address this issues, which aims to model and process images and text descriptions as graphs in order to exploit the relations inside and between each modality and employs specific techniques to extract syntactic and semantic information. The results obtained show promising performances on different tasks when compared to the present state-of-the-art deep learning architectures.

Bio-QA-GNN: Reasoning with Language Models and Knowledge Graphs for Interpretable Biomedical Question Answering

Nowadays the idea of injecting world or domain-specific structured knowledge into pre-trained language models (PLMs) is becoming an increasingly popular approach for solving problems such as biases, hallucinations, huge architectural sizes, and explainability lack—critical for real-world natural language processing applications in sensitive fields like bioinformatics. One recent work that has garnered much attention in Neuro-symbolic AI is QA-GNN, an end-to-end model for multiple-choice open-domain question answering (MCOQA) tasks via interpretable text-graph reasoning. Unlike previous publications, QA-GNN mutually informs PLMs and graph neural networks (GNNs) on top of relevant facts retrieved from knowledge graphs (KGs). However, taking a more holistic view, existing PLM+KG contributions mainly consider commonsense benchmarks and ignore or shallowly analyze performances on biomedical datasets. This thesis start from a propose of a deep investigation of QA-GNN for biomedicine, comparing existing or brand-new PLMs, KGs, edge-aware GNNs, preprocessing techniques, and initialization strategies. By combining the insights emerged in DISI's research, we introduce Bio-QA-GNN that include a KG. Working with this part has led to an improvement in state-of-the-art of MCOQA model on biomedical/clinical text, largely outperforming the original one (+3.63\% accuracy on MedQA). Our findings also contribute to a better understanding of the explanation degree allowed by joint text-graph reasoning architectures and their effectiveness on different medical subjects and reasoning types. Codes, models, datasets, and demos to reproduce the results are freely available at: \url{https://github.com/disi-unibo-nlp/bio-qagnn}.

Knowledge-based Chord Embeddings

This thesis develops AI methods as a contribution to computational musicology, an interdisciplinary field that studies music with computers. In systematic musicology a composition is defined as the combination of harmony, melody and rhythm. According to de La Borde, harmony alone "merits the name of composition". This thesis focuses on analysing the harmony from a computational perspective. We concentrate on symbolic music representation and address the problem of formally representing chord progressions in western music compositions. Informally, chords are sets of pitches played simultaneously, and chord progressions constitute the harmony of a composition. Our approach combines ML techniques with knowledge-based techniques. We design and implement the Modal Harmony ontology (MHO), using OWL. It formalises one of the most important theories in western music: the Modal Harmony Theory. We propose and experiment with different types of embedding methods to encode chords, inspired by NLP and adapted to the music domain, using both statistical (extensional) knowledge by relying on a huge dataset of chord annotations (ChoCo), intensional knowledge by relying on MHO and a combination of the two. The methods are evaluated on two musicologically relevant tasks: chord classification and music structure segmentation. The former is verified by comparing the results of the Odd One Out algorithm to the classification obtained with MHO. Good performances (accuracy: 0.86) are achieved. We feed a RNN for the latter, using our embeddings. Results show that the best performance (F1: 0.6) is achieved with embeddings that combine both approaches. Our method outpeforms the state of the art (F1 = 0.42) for symbolic music structure segmentation. It is worth noticing that embeddings based only on MHO almost equal the best performance (F1 = 0.58). We remark that those embeddings only require the ontology as an input as opposed to other approaches that rely on large datasets.

Du document à la donnée et retour. La fourmilière ou les Lumières

Même si le mot est ancien, la référence à la notion de document est récente dans l'Histoire, sans doute en résonance avec l'organisation de la société industrielle, sa régulation et ses valeurs. Au tournant du millénaire, le web s'est appuyé sur un renversement du circuit documentaire, jusqu'à, dans le web de données, un court-circuitage radical. S'agit-il de l'effacement d'une notion périmée au profit d'une autre ou d'un simple décalage ? Le succès du web accompagne des transformations sociales et économiques profondes. Quelles seront alors, sur nos régimes de vérité, de preuve et de transmission, les conséquences de l'éventuelle obsolescence de la notion de document ?

Data sensitivity detection in chat interactions for privacy protection

In recent years, there has been exponential growth in using virtual spaces, including dialogue systems, that handle personal information. The concept of personal privacy in the literature is discussed and controversial, whereas, in the technological field, it directly influences the degree of reliability perceived in the information system (privacy ‘as trust’). This work aims to protect the right to privacy on personal data (GDPR, 2018) and avoid the loss of sensitive content by exploring sensitive information detection (SID) task. It is grounded on the following research questions: (RQ1) What does sensitive data mean? How to define a personal sensitive information domain? (RQ2) How to create a state-of-the-art model for SID?(RQ3) How to evaluate the model? RQ1 theoretically investigates the concepts of privacy and the ontological state-of-the-art representation of personal information. The Data Privacy Vocabulary (DPV) is the taxonomic resource taken as an authoritative reference for the definition of the knowledge domain. Concerning RQ2, we investigate two approaches to classify sensitive data: the first - bottom-up - explores automatic learning methods based on transformer networks, the second - top-down - proposes logical-symbolic methods with the construction of privaframe, a knowledge graph of compositional frames representing personal data categories. Both approaches are tested. For the evaluation - RQ3 – we create SPeDaC, a sentence-level labeled resource. This can be used as a benchmark or training in the SID task, filling the gap of a shared resource in this field. If the approach based on artificial neural networks confirms the validity of the direction adopted in the most recent studies on SID, the logical-symbolic approach emerges as the preferred way for the classification of fine-grained personal data categories, thanks to the semantic-grounded tailor modeling it allows. At the same time, the results highlight the strong potential of hybrid architectures in solving automatic tasks.

Connecting works of art within the semantic web of symbolic meanings

My doctoral research is about the modelling of symbolism in the cultural heritage domain, and on connecting artworks based on their symbolism through knowledge extraction and representation techniques. In particular, I participated in the design of two ontologies: one models the relationships between a symbol, its symbolic meaning, and the cultural context in which the symbol symbolizes the symbolic meaning; the second models artistic interpretations of a cultural heritage object from an iconographic and iconological (thus also symbolic) perspective. I also converted several sources of unstructured data, a dictionary of symbols and an encyclopaedia of symbolism, and semi-structured data, DBpedia and WordNet, to create HyperReal, the first knowledge graph dedicated to conventional cultural symbolism. By making use of HyperReal's content, I showed how linked open data about cultural symbolism could be utilized to initiate a series of quantitative studies that analyse (i) similarities between cultural contexts based on their symbologies, (ii) broad symbolic associations, (iii) specific case studies of symbolism such as the relationship between symbols, their colours, and their symbolic meanings. Moreover, I developed a system that can infer symbolic, cultural context-dependent interpretations from artworks according to what they depict, envisioning potential use cases for museum curation. I have then re-engineered the iconographic and iconological statements of Wikidata, a widely used general-domain knowledge base, creating ICONdata: an iconographic and iconological knowledge graph. ICONdata was then enriched with automatic symbolic interpretations. Subsequently, I demonstrated the significance of enhancing artwork information through alignment with linked open data related to symbolism, resulting in the discovery of novel connections between artworks. Finally, I contributed to the creation of a software application. This application leverages established connections, allowing users to investigate the symbolic expression of a concept across different cultural contexts through the generation of a three-dimensional exhibition of artefacts symbolising the chosen concept.

One archive, many readings. Personal archives as complex networks in the semantic web

Personal archives are the archives created by individuals for their own purposes. Among these are the library and documentary collections of writers and scholars. It is only recently that archival literature has begun to focus on this category of archives, emphasising how their heterogeneous nature necessitates the conciliation of different approaches to archival description, and calling for a broader understanding of the principle of provenance, recognising that multiple creators, including subsequent researchers, can contribute to shaping personal archives over time by adding new layers of contexts. Despite these advances in the theoretical debate, current architectures for archival representation remain behind. Finding aids privilege a single point of view and do not allow subsequent users to embed their own, potentially conflicting, readings. Using semantic web technologies this study aims to define a conceptual model for writers' archives based on existing and widely adopted models in the cultural heritage and humanities domains. The model developed can be used to represent different types of documents at various levels of analysis, as well as record content and components. It also enables the representation of complex relationships and the incorporation of additional layers of interpretation into the finding aid, transforming it from a static search tool into a dynamic research platform.  The personal archive and library of Giuseppe Raimondi serves as a case study for the creation of an archival knowledge base using the proposed conceptual model. By querying the knowledge graph through SPARQL, the effectiveness of the model is evaluated. The results demonstrate that the model addresses the primary representation challenges identified in archival literature, from both a technological and methodological standpoint. The ultimate goal is to bring the output par excellence of archival science, i.e. the finding aid, more in line with the latest developments in archival thinking.

Network analysis and machine learning assist drug repurposing and safety assessment in neurological diseases

In recent decades, two prominent trends have influenced the data modeling field, namely network analysis and machine learning. This thesis explores the practical applications of these techniques within the domain of drug research, unveiling their multifaceted potential for advancing our comprehension of complex biological systems. The research undertaken during this PhD program is situated at the intersection of network theory, computational methods, and drug research. Across six projects presented herein, there is a gradual increase in model complexity. These projects traverse a diverse range of topics, with a specific emphasis on drug repurposing and safety in the context of neurological diseases. The aim of these projects is to leverage existing biomedical knowledge to develop innovative approaches that bolster drug research. The investigations have produced practical solutions, not only providing insights into the intricacies of biological systems, but also allowing the creation of valuable tools for their analysis. In short, the achievements are: • A novel computational algorithm to identify adverse events specific to fixed-dose drug combinations. • A web application that tracks the clinical drug research response to SARS-CoV-2. • A Python package for differential gene expression analysis and the identification of key regulatory "switch genes". • The identification of pivotal events causing drug-induced impulse control disorders linked to specific medications. • An automated pipeline for discovering potential drug repurposing opportunities. • The creation of a comprehensive knowledge graph and development of a graph machine learning model for predictions. Collectively, these projects illustrate diverse applications of data science and network-based methodologies, highlighting the profound impact they can have in supporting drug research activities.

Rappresentazione semantica delle informazioni in ambito storico e giuridico

Rappresentazione della conoscenza in banca di dati testuali non strutturati in lingua Italiana.

Content aggregation on knowledge bases using graph clustering

Recently, research projects such as PADLR and SWAP have developed tools like Edutella or Bibster, which are targeted at establishing peer-to-peer knowledge management (P2PKM) systems. In such a system, it is necessary to obtain provide brief semantic descriptions of peers, so that routing algorithms or matchmaking processes can make decisions about which communities peers should belong to, or to which peers a given query should be forwarded. This paper proposes the use of graph clustering techniques on knowledge bases for that purpose. Using this clustering, we can show that our strategy requires up to 58% fewer queries than the baselines to yield full recall in a bibliographic P2PKM scenario.

Linked knowledge sources for topic classification of microposts:a semantic graph-based approach

Short text messages a.k.a Microposts (e.g. Tweets) have proven to be an effective channel for revealing information about trends and events, ranging from those related to Disaster (e.g. hurricane Sandy) to those related to Violence (e.g. Egyptian revolution). Being informed about such events as they occur could be extremely important to authorities and emergency professionals by allowing such parties to immediately respond. In this work we study the problem of topic classification (TC) of Microposts, which aims to automatically classify short messages based on the subject(s) discussed in them. The accurate TC of Microposts however is a challenging task since the limited number of tokens in a post often implies a lack of sufficient contextual information. In order to provide contextual information to Microposts, we present and evaluate several graph structures surrounding concepts present in linked knowledge sources (KSs). Traditional TC techniques enrich the content of Microposts with features extracted only from the Microposts content. In contrast our approach relies on the generation of different weighted semantic meta-graphs extracted from linked KSs. We introduce a new semantic graph, called category meta-graph. This novel meta-graph provides a more fine grained categorisation of concepts providing a set of novel semantic features. Our findings show that such category meta-graph features effectively improve the performance of a topic classifier of Microposts. Furthermore our goal is also to understand which semantic feature contributes to the performance of a topic classifier. For this reason we propose an approach for automatic estimation of accuracy loss of a topic classifier on new, unseen Microposts. We introduce and evaluate novel topic similarity measures, which capture the similarity between the KS documents and Microposts at a conceptual level, considering the enriched representation of these documents. Extensive evaluation in the context of Emergency Response (ER) and Violence Detection (VD) revealed that our approach outperforms previous approaches using single KS without linked data and Twitter data only up to 31.4% in terms of F1 measure. Our main findings indicate that the new category graph contains useful information for TC and achieves comparable results to previously used semantic graphs. Furthermore our results also indicate that the accuracy of a topic classifier can be accurately predicted using the enhanced text representation, outperforming previous approaches considering content-based similarity measures. © 2014 Elsevier B.V. All rights reserved.

Cultural Knowledge for Named Entity Disambiguation: A Graph-Based Semantic Relatedness Approach

One of the ultimate aims of Natural Language Processing is to automate the analysis of the meaning of text. A fundamental step in that direction consists in enabling effective ways to automatically link textual references to their referents, that is, real world objects. The work presented in this paper addresses the problem of attributing a sense to proper names in a given text, i.e., automatically associating words representing Named Entities with their referents. The method for Named Entity Disambiguation proposed here is based on the concept of semantic relatedness, which in this work is obtained via a graph-based model over Wikipedia. We show that, without building the traditional bag of words representation of the text, but instead only considering named entities within the text, the proposed method achieves results competitive with the state-of-the-art on two different datasets.

Edge-aware Graph Attention Networks: Joint Reasoning On Text and Knowledge Graphs for Biomedical Question Answering

Much of the real-world dataset, including textual data, can be represented using graph structures. The use of graphs to represent textual data has many advantages, mainly related to maintaining a more significant amount of information, such as the relationships between words and their types. In recent years, many neural network architectures have been proposed to deal with tasks on graphs. Many of them consider only node features, ignoring or not giving the proper relevance to relationships between them. However, in many node classification tasks, they play a fundamental role. This thesis aims to analyze the main GNNs, evaluate their advantages and disadvantages, propose an innovative solution considered as an extension of GAT, and apply them to a case study in the biomedical field. We propose the reference GNNs, implemented with methodologies later analyzed, and then applied to a question answering system in the biomedical field as a replacement for the pre-existing GNN. We attempt to obtain better results by using models that can accept as input both node and edge features. As shown later, our proposed models can beat the original solution and define the state-of-the-art for the task under analysis.