Problems with “traditional” IR Systems on the Web

• According to Lawrence & Giles (Nature ‘99) search engines tend to have:

  • Decreasing recall - decreasing coverage due to the rapid growth and dynamic structure of the Web

    • Recall = Retrieved & Relevant / Relevant
  • Low precision - huge lists of hits with inaccurate ranking

    • Precision = Retrieved & Relevant / Retrieved
  • Stale links and duplicated pages - due to their static nature it takes long time to refresh indexes

Previous slide

Next slide

Back to first slide

View graphic version