CSCE 470 Lecture 4

« previous | Monday, September 2, 2013 | next »

Empirical Investigation of effect of ranking

Videotaping

Interviewing Thinking aloud Eye tracking

Rank is important: there are a few "click monkeys"

Calculating Rank

• If (keyword on Wikipedia), get bonus points
• Number of keyword counts (in title? in URL?)
• Popularity (clicks, dwell time, tweet mentions, Facebook likes)
• PageRank structure
• Aesthetics
• Load Time
• hosting malware or email harvesting crap.
• Machine learning rankers

Vector Space Retrieval

One of the (four) most important things in the course.

(No really, he will call you randomly 5 years later and ask.)

also known as:

• Bag of Words Model
• Vector Space Model

$n$-dimensional space

Each document is a vector in this space.

Queries exist in same space, and ranking is just distance between document and query.

Failed to parse (MathML with SVG or PNG fallback (recommended for modern browsers and accessibility tools): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle q}	D1	D2
Johnny Football	Johnny played football today	Johnny is a bad example

• Which document is closer to the query
• What is the information need? (news, autographed merchandise, purchase something, etc.)