Welcome to IOU Digital Library

This master’s thesis describes the design and implementation of an information retrieval (IR) system called SOZE, which uses growing neural networks (growing cell structures and growing neural gas) to model associative memory for word retrieval. The author explores word‑signal spaces based on n‑grams, “rubber” n‑grams, and phonetics, and reports experiments comparing the neural‑network–based retrieval with a fuzzy string–matching system used by MediaLab; results show that the growing‑networks often outperform the self‑organizing map, though fuzzy string matching still achieves near‑100% retrieval rate overall, with occasional cases where the network retrieves targets that the fuzzy matcher misses.

no reviews

245: $a: Information Retrieval using Associative Memory
245: $b:
082: $a:
100: $a: Jasper Kaizer
650: $a: Computer Science
250: $a:
440: $v:
250: $b: Universiteit van Amsterdam
250: $c: 1999
020: $a:
300: $a: 61