An Empirical Comparison of Token Enocoding Strategies in the Generation of Vector Representations of Structured Data.

Tools

Vincent, J. and Mintram, R., 2003. An Empirical Comparison of Token Enocoding Strategies in the Generation of Vector Representations of Structured Data. In: Artificial Intelligence and Applications (AIA 2004), 16 – 18 February 2004, Innsbruck, Austria.

Full text not available from this repository.

Abstract

A consequence of the connectionist approach to artificial intelligence is the requirement for structured data to be encoded into fixed width vector representations (VREPS). This paper provides an empirical comparison of six different strategies for encoding the tokens that appear within tree representations of this structured data. A new two element real-valued token encoding is presented and empirical results show that it produces more compact vec tors than previously possible with conventional encodings. This assessment is conducted within the General Encoder / Decoder (GED) framework and makes use of the VREP recovery profile (VRP) graphical representation to enable quantitative and qualitative judgements to be made.

Item Type:Conference or Workshop Item (Paper)
ISSN:1027-2666
Additional Information:Features inherent within data may also include structural contributions. The process of feature extraction should summarize both the principle features inherent within data corpa and the potential structural features that exist between subsegements of the data. As part of this process a local data representation of the atomic data componenets is required and this research exposes the significance of this encoding in relation to the effectiveness of the subsequent feature extraction. Typical encoding strategies might suppose that orthogonal encodings will enable effective feature deliniation but this researchs that this is not necessarily true and that a simple encoding strategies are generally superior.
Subjects:UNSPECIFIED
Group:School of Design, Engineering & Computing > Software Systems Research Centre
ID Code:1073
Deposited By:INVALID USER
Deposited On:20 Feb 2007
Last Modified:07 Mar 2013 14:36
Repository Staff Only -
BU Staff Only -
Help Guide - Editing Your Items in BURO