I am currently a postdoctoral research fellow, member of the CORE: Conceptual and Cognitive Modelling Research Group at the Free University of Bozen-Bolzano.
I have a Master’s Degree in Philosophy and a Ph.D. in Computer Science. My research concerns primarily Artificial Intelligence, with a particular focus on the combination of Knowledge Representation and Machine Learning techniques. Other research interests include Logic and Philosophy of Mind.
When building a new application we are more and more confronted with the need of reusing and inte... more When building a new application we are more and more confronted with the need of reusing and integrating pre-existing knowledge, e.g., ontologies, schemas, data of any kind, from multiple sources. Nevertheless, it is a fact that this prior knowledge is virtually impossible to reuse as-is. This difficulty is the cause of high costs, with the further drawback that the resulting application will again be hardly reusable. It is a negative loop which consistently reinforces itself. iTelos is a general purpose methodology aiming at minimizing as much as possible the effects of this loop. iTelos is based on the intuition that the data level and the schema level of an application should be developed independently, thus allowing for maximum flexibility in the reuse of the prior knowledge, but under the overall guidance of the needs to be satisfied, formalized as competence queries. This intuition is implemented by codifying all the requirements, including those concerning reuse, as part of a...
It is a fact that, when developing a new application, it is virtually impossible to reuse, as-is,... more It is a fact that, when developing a new application, it is virtually impossible to reuse, as-is, existing datasets. This difficulty is the cause of additional costs, with the further drawback that the resulting application will again be hardly reusable. It is a negative loop which consistently reinforces itself and for which there seems to be no way out. iTelos is a general purpose methodology designed to break this loop. Its main goal is to generate reusable Knowledge Graphs (KGs), built reusing, as much as possible, already existing data. The key assumption is that the design of a KG should be done middle-out meaning by this that the design should take into consideration, in all phases of the development: (i) the purpose to be served, that we formalize as a set of competency queries, (ii) a set of pre-existing datasets, possibly extracted from existing KGs, and (iii) a set of pre-existing reference schemas, whose goal is to facilitate sharability. We call these reference schemas,...
One of the major barriers to the training of statistical models on knowledge representations is t... more One of the major barriers to the training of statistical models on knowledge representations is the difficulty that scientists have in finding the best input data to be used for addressing their prediction goal. In addition to this, a key challenge is to determine how to manipulate these relational data, which are often in the form of particular triples (i.e., subject, predicate, object), to enable the learning process. This paper describes the LiveSchema initiative, namely a gateway that leverages the gold-mine of relational data collected by many existing ontology catalogs. By implementing a continuously updating aggregation facility, LiveSchema aims at providing a family of services that can be used to easily access, accurately analyze, transform and re-use data in a machine learning scenario.
The representation of concepts is a lively research activity in several artificial intelligence (... more The representation of concepts is a lively research activity in several artificial intelligence (AI) areas, such as knowledge representation, machine learning, and natural language processing. So far, many solutions have been proposed adopting different assumptions about the nature of concepts. Each of these solutions has been developed for capturing some specific features and for supporting some specific (artificial) cognitive operations. This paper provides a teleological explanation of the most widely shared approaches in AI to the representation of concepts. The paper aims at providing four main contributions: i) an overview of the mainstream philosophical theories of concepts; ii) a categorization of a wide range of AI solutions inspired by such theories of concepts; iii) the proposal of a method for a comprehensive explanation of the current approaches to concepts in AI based on a teleosemantic perspective; iv) an illustration of how the proposed explanation could constitute a...
In this short paper we introduce two main elements of a general methodology, called iTelos, for t... more In this short paper we introduce two main elements of a general methodology, called iTelos, for the management of knowledge diversity. The first is Knowledge Lotuses, a general purpose tool for the representation of knowledge diversity, while the second is a set of metrics which allow to quantify it, as it occurs within and across knowledge resources.
Frontiers in Artificial Intelligence and Applications, 2020
The aim of transfer learning is to reuse learnt knowledge across different contexts. In the parti... more The aim of transfer learning is to reuse learnt knowledge across different contexts. In the particular case of cross-domain transfer (also known as domain adaptation), reuse happens across different but related knowledge domains. While there have been promising first results in combining learning with symbolic knowledge to improve cross-domain transfer results, the singular ability of ontologies for providing classificatory knowledge has not been fully exploited so far by the machine learning community. We show that ontologies, if properly designed, are able to support transfer learning by improving generalization and discrimination across classes. We propose an architecture based on direct attribute prediction for combining ontologies with a transfer learning framework, as well as an ontology-based solution for cross-domain generalization based on the integration of top-level and domain ontologies. We validate the solution on an experiment over an image classification task, demonst...
Proceedings of the Seventeenth International Conference on Principles of Knowledge Representation and Reasoning, 2020
Semantic Heterogeneity is the problem that arises when multiple resources present differences in ... more Semantic Heterogeneity is the problem that arises when multiple resources present differences in how they represent the same real-world phenomenon. In KR, an early approach was the development of ontologies and, later on, when ontologies showed at the knowledge level the same semantic heterogeneity that they were meant to fix at the data level, to compute mappings among them. In this paper we acknowledge the impossibility of avoiding semantic heterogeneity, this being a consequence of the more general phenomenon of the diversity of the world and of the world descriptions. In this perspective, the heterogeneity of ontologies is a feature (and not a bug to be fixed by aligning them) which gives the possibility to use the most suitable ontology in any given application context. The main contributions of this paper are: (i) a novel articulation of the problem of semantic heterogeneity, as it appears at the knowledge level, as contextuality, (ii) its qualitative and quantitative formaliz...
PRICAI 2019: Trends in Artificial Intelligence, 2019
The paper focuses on two pivotal cognitive functions of both natural and AI agents, namely classi... more The paper focuses on two pivotal cognitive functions of both natural and AI agents, namely classification and identification. Inspired from the theory of teleosemantics, itself based on neuroscientific results, we show that these two functions are complementary and rely on distinct forms of knowledge representation. We provide a new perspective on well-known AI techniques by categorising them as either classificational or identificational. Our proposed Teleo-KR architecture provides a high-level framework for combining the two functions within a single AI system. As validation and demonstration on a concrete application, we provide experiments on the large-scale reuse of classificational (ontological) knowledge for the purposes of learning-based schema identification.
We start from the observation that the notion of concept, as it is used in perception, is distinc... more We start from the observation that the notion of concept, as it is used in perception, is distinct and different from the notion of concept, as it is used in knowledge representation. In earlier work we called the first notion, substance concept and the second, classification concept. In this paper we integrate these two notions into a general theory of concepts that organizes them into a hierarchy of increasing abstraction from what is perceived. Thus, at the first level, we have objects (which roughly correspond to substance concepts), which represent what is perceived (e.g., a car); at the second level we have actions, which represent how objects change in time (e.g., move); while, at the third level, we have functions (which roughly correspond to classification concepts), which represent the expected behavior of objects as it is manifested in terms of “an object performing a certain set of actions” (e.g., a vehicle). The main outcome is the notion of Teleology, where teleologies provide the basis for a solution to the problem of the integration of perception and reasoning and, more in general, to the problem of managing the diversity of knowledge.
The generation of high quality and re-usable ontologies depends on effective methodologies aimed ... more The generation of high quality and re-usable ontologies depends on effective methodologies aimed at supporting the crucial process of identifying the ontology requirements, in terms of the number of potential end-users and end-users’ informational needs. It is widely recognized that the exploitation of competency questions (CQs) plays an important role in this requirement definition phase. In this paper, we aim at introducing a new general approach to exploit (web) search trends, and the huge amount of searches that people make every-day with web search engines, as a pivotal complementary source of information for the identification of informal needs of large numbers of end-users. To achieve this goal we use the “autosuggest” results provided by search engines like Bing and Google as a goldmine of data and insights. We select a set of keywords to identify the ontology terminology, and we collect and analyze a huge amount of web search queries (WSQs) related to the selected set of ke...
Conceptual models have a significant role in the
representation of knowledge and in promoting a c... more Conceptual models have a significant role in the representation of knowledge and in promoting a common understanding of reality. These concrete information artifacts are expressed by means of specific modeling languages and can be used to organize knowledge around entities in databases. So far, a lot of effort has been made in creating powerful modeling languages for providing well-structured conceptual models. However, many challenges are still there. Modeling choices are inspired by implicit ontological assumptions and the way modeling languages are used to describe reality is completely arbitrary. The structural heterogeneity of modeled information causes issues when data need to be combined and shared in a unified view. In order to decrease such heterogeneity and promote a common understanding of the modeled reality, it is essential to make the modeling assumptions and the ontological distinctions concerning the basic constructs of the modeling language explicit. The computational impact of the ontological foundation of modeling languages has not been widely explored yet and there have been very few attempts at creating ontologically well-founded modeling languages. Therefore, this paper, which refers to an ongoing research project, aims at describing the basic steps to be followed for an ontological foundation for ER modeling language.
In this paper, a new knowledge representation formalism, called the entity model, is introduced. ... more In this paper, a new knowledge representation formalism, called the entity model, is introduced. This model can be used to address knowledge diversity by making the modeling assumptions of different knowledge representations explicit and by rooting them in a world representation. The entity model can be used to: 1) detect the possible ways in which the diversity appears in ER models and therefore improving their representational adequacy; 2) make the modeling assumptions behind different ER models explicit; 3) combine the different ER models in a unified view, thus enabling data integration.
When building a new application we are more and more confronted with the need of reusing and inte... more When building a new application we are more and more confronted with the need of reusing and integrating pre-existing knowledge, e.g., ontologies, schemas, data of any kind, from multiple sources. Nevertheless, it is a fact that this prior knowledge is virtually impossible to reuse as-is. This difficulty is the cause of high costs, with the further drawback that the resulting application will again be hardly reusable. It is a negative loop which consistently reinforces itself. iTelos is a general purpose methodology aiming at minimizing as much as possible the effects of this loop. iTelos is based on the intuition that the data level and the schema level of an application should be developed independently, thus allowing for maximum flexibility in the reuse of the prior knowledge, but under the overall guidance of the needs to be satisfied, formalized as competence queries. This intuition is implemented by codifying all the requirements, including those concerning reuse, as part of a...
It is a fact that, when developing a new application, it is virtually impossible to reuse, as-is,... more It is a fact that, when developing a new application, it is virtually impossible to reuse, as-is, existing datasets. This difficulty is the cause of additional costs, with the further drawback that the resulting application will again be hardly reusable. It is a negative loop which consistently reinforces itself and for which there seems to be no way out. iTelos is a general purpose methodology designed to break this loop. Its main goal is to generate reusable Knowledge Graphs (KGs), built reusing, as much as possible, already existing data. The key assumption is that the design of a KG should be done middle-out meaning by this that the design should take into consideration, in all phases of the development: (i) the purpose to be served, that we formalize as a set of competency queries, (ii) a set of pre-existing datasets, possibly extracted from existing KGs, and (iii) a set of pre-existing reference schemas, whose goal is to facilitate sharability. We call these reference schemas,...
One of the major barriers to the training of statistical models on knowledge representations is t... more One of the major barriers to the training of statistical models on knowledge representations is the difficulty that scientists have in finding the best input data to be used for addressing their prediction goal. In addition to this, a key challenge is to determine how to manipulate these relational data, which are often in the form of particular triples (i.e., subject, predicate, object), to enable the learning process. This paper describes the LiveSchema initiative, namely a gateway that leverages the gold-mine of relational data collected by many existing ontology catalogs. By implementing a continuously updating aggregation facility, LiveSchema aims at providing a family of services that can be used to easily access, accurately analyze, transform and re-use data in a machine learning scenario.
The representation of concepts is a lively research activity in several artificial intelligence (... more The representation of concepts is a lively research activity in several artificial intelligence (AI) areas, such as knowledge representation, machine learning, and natural language processing. So far, many solutions have been proposed adopting different assumptions about the nature of concepts. Each of these solutions has been developed for capturing some specific features and for supporting some specific (artificial) cognitive operations. This paper provides a teleological explanation of the most widely shared approaches in AI to the representation of concepts. The paper aims at providing four main contributions: i) an overview of the mainstream philosophical theories of concepts; ii) a categorization of a wide range of AI solutions inspired by such theories of concepts; iii) the proposal of a method for a comprehensive explanation of the current approaches to concepts in AI based on a teleosemantic perspective; iv) an illustration of how the proposed explanation could constitute a...
In this short paper we introduce two main elements of a general methodology, called iTelos, for t... more In this short paper we introduce two main elements of a general methodology, called iTelos, for the management of knowledge diversity. The first is Knowledge Lotuses, a general purpose tool for the representation of knowledge diversity, while the second is a set of metrics which allow to quantify it, as it occurs within and across knowledge resources.
Frontiers in Artificial Intelligence and Applications, 2020
The aim of transfer learning is to reuse learnt knowledge across different contexts. In the parti... more The aim of transfer learning is to reuse learnt knowledge across different contexts. In the particular case of cross-domain transfer (also known as domain adaptation), reuse happens across different but related knowledge domains. While there have been promising first results in combining learning with symbolic knowledge to improve cross-domain transfer results, the singular ability of ontologies for providing classificatory knowledge has not been fully exploited so far by the machine learning community. We show that ontologies, if properly designed, are able to support transfer learning by improving generalization and discrimination across classes. We propose an architecture based on direct attribute prediction for combining ontologies with a transfer learning framework, as well as an ontology-based solution for cross-domain generalization based on the integration of top-level and domain ontologies. We validate the solution on an experiment over an image classification task, demonst...
Proceedings of the Seventeenth International Conference on Principles of Knowledge Representation and Reasoning, 2020
Semantic Heterogeneity is the problem that arises when multiple resources present differences in ... more Semantic Heterogeneity is the problem that arises when multiple resources present differences in how they represent the same real-world phenomenon. In KR, an early approach was the development of ontologies and, later on, when ontologies showed at the knowledge level the same semantic heterogeneity that they were meant to fix at the data level, to compute mappings among them. In this paper we acknowledge the impossibility of avoiding semantic heterogeneity, this being a consequence of the more general phenomenon of the diversity of the world and of the world descriptions. In this perspective, the heterogeneity of ontologies is a feature (and not a bug to be fixed by aligning them) which gives the possibility to use the most suitable ontology in any given application context. The main contributions of this paper are: (i) a novel articulation of the problem of semantic heterogeneity, as it appears at the knowledge level, as contextuality, (ii) its qualitative and quantitative formaliz...
PRICAI 2019: Trends in Artificial Intelligence, 2019
The paper focuses on two pivotal cognitive functions of both natural and AI agents, namely classi... more The paper focuses on two pivotal cognitive functions of both natural and AI agents, namely classification and identification. Inspired from the theory of teleosemantics, itself based on neuroscientific results, we show that these two functions are complementary and rely on distinct forms of knowledge representation. We provide a new perspective on well-known AI techniques by categorising them as either classificational or identificational. Our proposed Teleo-KR architecture provides a high-level framework for combining the two functions within a single AI system. As validation and demonstration on a concrete application, we provide experiments on the large-scale reuse of classificational (ontological) knowledge for the purposes of learning-based schema identification.
We start from the observation that the notion of concept, as it is used in perception, is distinc... more We start from the observation that the notion of concept, as it is used in perception, is distinct and different from the notion of concept, as it is used in knowledge representation. In earlier work we called the first notion, substance concept and the second, classification concept. In this paper we integrate these two notions into a general theory of concepts that organizes them into a hierarchy of increasing abstraction from what is perceived. Thus, at the first level, we have objects (which roughly correspond to substance concepts), which represent what is perceived (e.g., a car); at the second level we have actions, which represent how objects change in time (e.g., move); while, at the third level, we have functions (which roughly correspond to classification concepts), which represent the expected behavior of objects as it is manifested in terms of “an object performing a certain set of actions” (e.g., a vehicle). The main outcome is the notion of Teleology, where teleologies provide the basis for a solution to the problem of the integration of perception and reasoning and, more in general, to the problem of managing the diversity of knowledge.
The generation of high quality and re-usable ontologies depends on effective methodologies aimed ... more The generation of high quality and re-usable ontologies depends on effective methodologies aimed at supporting the crucial process of identifying the ontology requirements, in terms of the number of potential end-users and end-users’ informational needs. It is widely recognized that the exploitation of competency questions (CQs) plays an important role in this requirement definition phase. In this paper, we aim at introducing a new general approach to exploit (web) search trends, and the huge amount of searches that people make every-day with web search engines, as a pivotal complementary source of information for the identification of informal needs of large numbers of end-users. To achieve this goal we use the “autosuggest” results provided by search engines like Bing and Google as a goldmine of data and insights. We select a set of keywords to identify the ontology terminology, and we collect and analyze a huge amount of web search queries (WSQs) related to the selected set of ke...
Conceptual models have a significant role in the
representation of knowledge and in promoting a c... more Conceptual models have a significant role in the representation of knowledge and in promoting a common understanding of reality. These concrete information artifacts are expressed by means of specific modeling languages and can be used to organize knowledge around entities in databases. So far, a lot of effort has been made in creating powerful modeling languages for providing well-structured conceptual models. However, many challenges are still there. Modeling choices are inspired by implicit ontological assumptions and the way modeling languages are used to describe reality is completely arbitrary. The structural heterogeneity of modeled information causes issues when data need to be combined and shared in a unified view. In order to decrease such heterogeneity and promote a common understanding of the modeled reality, it is essential to make the modeling assumptions and the ontological distinctions concerning the basic constructs of the modeling language explicit. The computational impact of the ontological foundation of modeling languages has not been widely explored yet and there have been very few attempts at creating ontologically well-founded modeling languages. Therefore, this paper, which refers to an ongoing research project, aims at describing the basic steps to be followed for an ontological foundation for ER modeling language.
In this paper, a new knowledge representation formalism, called the entity model, is introduced. ... more In this paper, a new knowledge representation formalism, called the entity model, is introduced. This model can be used to address knowledge diversity by making the modeling assumptions of different knowledge representations explicit and by rooting them in a world representation. The entity model can be used to: 1) detect the possible ways in which the diversity appears in ER models and therefore improving their representational adequacy; 2) make the modeling assumptions behind different ER models explicit; 3) combine the different ER models in a unified view, thus enabling data integration.
Uploads
Papers by Mattia Fumagalli
representation of knowledge and in promoting a common
understanding of reality. These concrete information artifacts are
expressed by means of specific modeling languages and can be
used to organize knowledge around entities in databases. So far,
a lot of effort has been made in creating powerful modeling
languages for providing well-structured conceptual models.
However, many challenges are still there. Modeling choices are
inspired by implicit ontological assumptions and the way
modeling languages are used to describe reality is completely
arbitrary. The structural heterogeneity of modeled information
causes issues when data need to be combined and shared in a
unified view. In order to decrease such heterogeneity and
promote a common understanding of the modeled reality, it is
essential to make the modeling assumptions and the ontological
distinctions concerning the basic constructs of the modeling
language explicit. The computational impact of the ontological
foundation of modeling languages has not been widely explored
yet and there have been very few attempts at creating
ontologically well-founded modeling languages. Therefore, this
paper, which refers to an ongoing research project, aims at
describing the basic steps to be followed for an ontological
foundation for ER modeling language.
representation of knowledge and in promoting a common
understanding of reality. These concrete information artifacts are
expressed by means of specific modeling languages and can be
used to organize knowledge around entities in databases. So far,
a lot of effort has been made in creating powerful modeling
languages for providing well-structured conceptual models.
However, many challenges are still there. Modeling choices are
inspired by implicit ontological assumptions and the way
modeling languages are used to describe reality is completely
arbitrary. The structural heterogeneity of modeled information
causes issues when data need to be combined and shared in a
unified view. In order to decrease such heterogeneity and
promote a common understanding of the modeled reality, it is
essential to make the modeling assumptions and the ontological
distinctions concerning the basic constructs of the modeling
language explicit. The computational impact of the ontological
foundation of modeling languages has not been widely explored
yet and there have been very few attempts at creating
ontologically well-founded modeling languages. Therefore, this
paper, which refers to an ongoing research project, aims at
describing the basic steps to be followed for an ontological
foundation for ER modeling language.