The research theme which is at the core the research of the Knowdive Group is how to provide a Computational theory of meaning, and to use it as the main means for understanding (i)the diversity of people, as it occurs in their representations of the world (what we call the personal context) and (ii)the implications that the diversity of people has on social relations, as they appear in anyone’s everyday life.
By computational theory of meaning here we mean (the first elements of) an explanation of how it is possible to replicate on a machine the causal relation which exists between what is perceived by a human or a machine, e.g., via sensors or vision, what is represented as knowledge expressed in certain language, the reasoning which is performed on this knowledge and the consequent action. As actions, we limit ourselves only to human-machine interactions and human-human interactions mediated by the machine.
The research of the group is organized along four main research lines, as follows:
- Language:the main focus here is to provide diversity aware computational models of natural language. This work is organized in two main areas:
- The Universal Knowledge Core (UKC): The UKC is a linguistic resource, based on the organizational principles of the Princeton Wordnet, which aggregates hundreds of world lexicons. One of the main research focuses is to understand the diversity of languages;
- Scroll: Scroll is a multilingual natural language processing engine tuned to what we call the Language of Data, namely the rather simplified language (e.g. noun or simple verbal phrases) used to represent structured or semi-structured data. One of the main research issues is the extent to which it is possible to develop a single integrated engine which covers multiple languages.
- Knowledge: The main goal here is to understand how it is possible to develop knowledge representation formalisms, what we call teleologies, which, still maintaining a unitary view of the schema and the knowledge underlying the representation of data, can adapt and evolve as a function of the ever changing input itself.
- Perception and action:The main reference scenario here is , e.g., a phone or a camera which lives in symbiosis with a reference user and which keeps getting input data for its sensors. A main research ouput is iLog, i.e. an integrated toolset for data collection, reasoning and machine learning which allows the machine to learn the same facts as those learned by its reference user.
- End-to-end applications: We apply the technologies described above in end-to-end applications which, on one side ,allow us to finalize and tune the underlying technologies while, on the other side, are aimed at solving real world problems.
- StarLinker – multiligual adaptive data integration: The problem statement here can be described as that of generating a knowledge graph which allows to integrate multilingual datasets which keep evolving and being enriched by new datasets. Examples of application scenarios are: the University data, mobility and health.
- Witmee – the friend of a life: The problem statement here can be described as thatof continuously getting input data from the sensors, integrating what islearned inside a knowledge graph representing what is already known, and using what is known to support the user in her everyday life and also in her social interactions. Examples of application scenarios are the University student life and mobility services.
Last updated: July 26, 2018
Fausto Giunchiglia - 2018
We all live in our own world which is different from that of everybody else. Perception creates it, language allows to share its description with the other people, thus causing its objectivation, knowledge is what we come to learn about it by observing what repeats itself through change, as described by language. This process has been in place for centuries, most likely since the dawn of the human kind...