Skip to main content
Go to the home page of the European Commission (opens in new window)
English en
CORDIS - EU research results
CORDIS

Tensors and Neural Networks for Computational Creativity

Project description

Developing computational models of linguistic creativity

Creativity in language is everywhere, not just in literature but also in everyday language use. A computational model of linguistic creativity can be useful for two reasons: it could provide insight into the processes of human creative expression, and it could enable natural language generation systems to produce realistic discourse. Funded by the European Research Council, the TENACITY project aims to develop autonomous models of linguistic creativity through a novel approach that combines tensors and neural networks. Standard neural network techniques are already performing well in some natural language-processing tasks, but they simply imitate human language production and lack creativity. TENACITY aspires to overcome this limitation by developing algorithms capable of grasping meaning from text and expressing it creatively.

Objective

Creativity in language is ubiquitous. It is abundantly present in work with an explicit creative intention - such as literary novels or poems - but weighty doses of creativity also pervade everyday language use. We believe that a computational model of creativity that focuses on language will shed light on the enigmatic processes and interactions that come into play when we humans express ourselves in creative ways. Moreover, natural language generation systems - in order to produce realistic utterances - need to be endowed with a certain capacity for creativity. The main goal of this research project is to develop unsupervised models of language that exhibit creativity. In order to do so, we propose an integrated approach that combines a number of important and innovative techniques. First of all, we rely on constructs from linear algebra called tensors in order to express language content according to different parameters. Using tensors, we are able to induce latent semantics from multi-way co-occurrences of textual content, which can subsequently be used for the generation of creative expressions. Secondly, we rely on advanced machine learning techniques, notably neural networks. Neural network techniques have recently shown impressive performance in a number of natural language processing tasks. Yet, these techniques are mainly mimicking human language production, and thus are showing little creativity in language generation; by adapting neural network approaches in various ways, as well as integrating them with our tensor-based approach, we expect to develop algorithms that are able to grasp the meaning of textual content in a more profound and elaborate way, and at the same time are able to express it with creative intent. The project has the potential for groundbreaking results, not only because it would deepen our understanding of creativity, but also because of practical applications within the field of natural language processing.

Fields of science (EuroSciVoc)

Keywords

Programme(s)

Topic(s)

Funding Scheme

HORIZON-ERC - HORIZON ERC Grants

See all projects funded under this funding scheme

Call for proposal

(opens in new window) ERC-2022-COG

See all projects funded under this call

Host institution

KATHOLIEKE UNIVERSITEIT LEUVEN
Net EU contribution
€ 1 988 500,00
Address
OUDE MARKT 13
3000 Leuven
Belgium
Region
Vlaams Gewest Prov. Vlaams-Brabant Arr. Leuven
Activity type
Higher or Secondary Education Establishments
Links
Total cost
€ 1 988 500,00

Beneficiaries (1)