Ugh! Ipad fell of my lap and I lost my post somehow.
I need to learn to use my smart phone as a presentation remote control like Amelie is doing.
term frequency (tf) # of occurences in the doc
inverse document frequency (idf)
high= many times in a small number of documents
lower=fewer times in a document or in many documents
lowest=term occurs in almost all documents
This is not enough for good relevance.
Example of looking for "rope" Foo Fighters song on Soundcloud. Without page rank, you don't get what you expect.
Boost the popular results.
The web is a graph.
Some nodes are visited more often
- Nodes with many links
-coming from frequently visited nodes
If you end up on a page where there is no link, you can enter an address to go somewhere else.
Going to a node without using a link.
Adjaceny matrix can represent the graph of nodes, links, and teleport.
Each row represents a node and the links between.
Empty rows don't link to anything. Add teleport to all 0 rows.
I'm sure this does not make any sense without the diagrams and matrixes. Oh well.
"It's fairly simple when you think about it," she says. I think that is true, but equations with tildas in them intimidate me.
Univeral search on Soundcloud
They wanted users, songs, and sets mixed by relevance.
Their graph has not only nodes, but also node types and how they are linked.
(User A follows user B and created playlist A)
They weight links based on these relationships.
Introduction to Information Retrieval is a great book for people interested in this topic.