ISINC
ISINC

ISINC
ISINC
 : 
Inference Systems for Inconsistent Information: logical foundations
Inference Systems for Inconsistent Information: logical foundations

A Project coordinated by IIIA.

Web page:

Principal investigator:

Collaborating organisations:

Universitat de Lleida

Universitat de Lleida

Funding entity:

Ministerio de Ciencia e Innovación
Ministerio de Ciencia e Innovación

Funding call:

Funding call URL:

Project #:

PID2019-111544GB-C21
PID2019-111544GB-C21

Funding amount:

93.533,00€
93.533,00€

Duration:

2020-06-01
2020-06-01
2023-05-31
2023-05-31

Extension date:

In this project the main goal is to advance the state-of-the-art in inconsistency-tolerant inference models in different scenarios: MaxSAT techniques in classical and many-valued logics, non classical graded logics, argumentation frameworks, both in theoretic and practical aspects, and in their application to the analysis of discussions in social networks. The main difficulty we want to tackle is the existence of inconsistency in knowledge bases, a common property in knowledge bases that come from real applications, specially when the information is obtained as the aggregation of information coming from different sources.

On the one hand, we plan to use extensions of different non-classical logics, mainly based on fuzzy logics and modal fuzzy logics, for being able to extract useful information from such kind of knowledge bases, and being able to manage both uncertain and inconsistent information. We will define models over these expanded logics, as well as inference algorithms that can be used to extract useful information under these new logics. The algorithms studied will be either ad-hoc or based on SAT/MaxSAT reductions.

On the other hand, we also plan to consider an approach for working with inconsistent information based on extensions of argumentation models, incorporating again both uncertain and inconsistent information. We will define argumentation models and algorithms for them, trying to identify special cases that can be solver with efficient algorithms. We also plan to study approximate inference algorithms for these problems, based on machine learning methods.

As an application domain, we plan to test our models and algorithms on different problems related to the analysis of discussions and comment threads in different social networks, where inconsistency is a very common property in these scenarios, but we may also encounter uncertain information, as not all the pieces of information we find in them are always believed to have the same strength.

The first subproject will mostly focus on the theoretical aspects of the project and its main activity will be the definition of new logical formalisms and inference systems to deal with inconsistencies in different scenarios. Some of the theoretical problems we propose are motivated by our experience in the development of proof procedures and the resolution of challenging combinatorial optimization problems.

In this project the main goal is to advance the state-of-the-art in inconsistency-tolerant inference models in different scenarios: MaxSAT techniques in classical and many-valued logics, non classical graded logics, argumentation frameworks, both in theoretic and practical aspects, and in their application to the analysis of discussions in social networks. The main difficulty we want to tackle is the existence of inconsistency in knowledge bases, a common property in knowledge bases that come from real applications, specially when the information is obtained as the aggregation of information coming from different sources.

On the one hand, we plan to use extensions of different non-classical logics, mainly based on fuzzy logics and modal fuzzy logics, for being able to extract useful information from such kind of knowledge bases, and being able to manage both uncertain and inconsistent information. We will define models over these expanded logics, as well as inference algorithms that can be used to extract useful information under these new logics. The algorithms studied will be either ad-hoc or based on SAT/MaxSAT reductions.

On the other hand, we also plan to consider an approach for working with inconsistent information based on extensions of argumentation models, incorporating again both uncertain and inconsistent information. We will define argumentation models and algorithms for them, trying to identify special cases that can be solver with efficient algorithms. We also plan to study approximate inference algorithms for these problems, based on machine learning methods.

As an application domain, we plan to test our models and algorithms on different problems related to the analysis of discussions and comment threads in different social networks, where inconsistency is a very common property in these scenarios, but we may also encounter uncertain information, as not all the pieces of information we find in them are always believed to have the same strength.

The first subproject will mostly focus on the theoretical aspects of the project and its main activity will be the definition of new logical formalisms and inference systems to deal with inconsistencies in different scenarios. Some of the theoretical problems we propose are motivated by our experience in the development of proof procedures and the resolution of challenging combinatorial optimization problems.

No publications uploaded yet
Amanda Vidal
Contract Researcher
Phone Ext. 252
Eva Armengol
Tenured Scientist
Phone Ext. 226
Felip Manyà
Tenured Scientist
Phone Ext. 248
Francesc Esteva
Adjunct Professor Ad Honorem
Phone Ext. 219
Lluís Godo
Research Professor
Phone Ext. 217
Pedro Meseguer
Scientific Researcher
Phone Ext. 237
Pilar Dellunde
Adjunct Scientist
Phone Ext. 239
Tommaso Flaminio
Tenured Scientist
Phone Ext. 233