Web Site Searching and the User Experience
BayCHI, January 1999
Avi Rappoport, Search Tools Consulting

Search Quality

Information Retrieval Theory

• Recall vs. precision

  • Recall = all the relevant pages

  • Precision = only the relevant pages



• Extended character sets and multiple languages

  • Multiple extended ASCIIs - Unix vs. Windows vs. Mac vs. Unicode

  • Non-Roman Character Sets

  • Multiple languages - English "mole" vs. Spanish "mole"

• Stopwords

  • a, an and the are often removed from the index

  • Common words may be vital to some queries, e.g. song titles, "Because of You"

• Stemming

  • Indexing and searching on root forms rather than word forms, so searching for "paid" will retrieve "pay" (can be problematic!)

• Fuzzy matching

  • Using linguistic analysis to match words which are similar in many ways: spelling, sound, etc.

    • ZDNet (Thunderstone Texis)

Table of Contents | Previous Page | Next Page