Digital Enterprise Research Institute From Web 1 0
Digital Enterprise Research Institute From Web 1. 0 to Web 2. 0 and Back – How did your Grandma Use to Tag? Sheila Kinsella, Adriana Budura, Gleb Skobeltsyn, Sebastian Michel, John G. Breslin, Karl Aberer Copyright 2008 Digital Enterprise Research Institute. All rights reserved. www. deri. ie
From Web 1. 0 – Anchortext… Digital Enterprise Research Institute www. deri. ie Users link to resources of interest from their webpages (implicit annotations) n Limited to web authors n
…to Web 2. 0 - Tagging Digital Enterprise Research Institute Explicit annotations of resources n Anyone can participate n www. deri. ie
So has anything changed? Digital Enterprise Research Institute www. deri. ie
Example results (1) Digital Enterprise Research Institute www. deri. ie
Example results (2) Digital Enterprise Research Institute www. deri. ie
Example results (3) Digital Enterprise Research Institute www. deri. ie
Experimental study Digital Enterprise Research Institute n Datasets: ¨ n www. deri. ie Web: WEBSPAM-2007, 12 M pages from. uk domain ¨ Del. icio. us: 2007 Crawl containing tags for 4. 5 M URLs ¨ Overlap between datasets: 192 k URLs Goal: ? ≈
Results Digital Enterprise Research Institute www. deri. ie
Overlap between the tags Digital Enterprise Research Institute www. deri. ie Anchortags overlap with del. icio. us tags n Two measures: n ¨ deliciousness@k: how many anchortags are also among del. icio. us tags ¨ anchortextness@k: how many del. icio. us tags can be inferred from anchortext k=1 k=2 k=3 k=4 k=5 Deliciousness@k 0. 44 0. 43 0. 42 0. 39 0. 36 Anchortextness@k 0. 40 0. 37 0. 34 0. 32 0. 30
User study Digital Enterprise Research Institute n n Tags for 20 URLs: ¨ top-5 tags from del. icio. us ¨ top-5 anchortags 25 people score every tag: ¨ 0 (non-relevant), ¨ 1 (somewhat relevant), ¨ 2 (very relevant). www. deri. ie
But users don’t always agree Digital Enterprise Research Institute n www. deri. ie Extent of Agreement Freqency All 3 scores agree 35% All 3 scores almost agree 50% All other cases 15% Relevance of a tag is highly subjective
Many new social tags Digital Enterprise Research Institute www. deri. ie
Conclusions Digital Enterprise Research Institute www. deri. ie Substansial overlap between top tags derived from anchortext and del. icio. us tags n Users rate anchortags at a similar quality to userassigned tags – but this is subjective n ≈ n Del. icio. us still provides many new tags ≠
- Slides: 14