Overview
Topic status: We're looking for students to study this topic.
Humans frequently combine concepts to produce new emergent properties. There are many theoretical models in psychology to explain how human create this new meaning, one of them being the use of analogies. For example, one can infer that a 'beach bike' will have 'thick tyres' knowing that a beach car has 'thick tyres' and that both car and bike are transportation means. Verbs have been used in computational linguistics as supports to characterize the type of relation between the words composing the compound. This project will seek to take this hypothesis further by using verbs to process analogies in a semantic space and automatically infer the emerging properties of a compound. When a new compound occurs, the first step is to retrieve an existing analogical compound, then to extract the inherited properties.
Suits: IT student with a background in programming.
Hypothesis/Aims: The hypothesis is that semantic associations based on verb usage provide a good measure of analogy and emerging properties are inherited from analogical existing combinations. The aim is to build a system that can computationally simulate this process.
Approaches:
- Design and implementation of a semantic space
- Theoretical developments on the use of the space for analogies and on property inheritance
- Empirical evaluation using human generated emergence
References: L. Sitbon, K. Kitto, B. Ramm, P. Bruza, The concept combination challenge emerges: computational methods meet human cognition, submitted to ACL-2010: 48th annual meeting for the Association for Computational Linguistics (contact Prof. Bruza for a copy)
- Study level
- Honours
- Supervisors
- QUT
- Organisational unit
Science and Engineering Faculty
- Research area
- Contact
- Please contact the supervisor.
Dr Laurianne Sitbon
Professor Peter Bruza
