The readings for this week, while a little complex, were fairly interesting. It is nice to see metadata fields or other structured fields brought into the discussion as they seemed a blatant exemption up to this point. Weighting of the zones of a document, also, seemed like an obvious improvement on the simple models we have discussed until now. I found the description of 'machine-learned relevance' to be overly complicated - is this just a simple calibration process (i.e. build the weights based on expert opinion for a known document set then calibrate the parameters for the calculation of weights to match the expert set)?
The discussion of the vector space model was slightly perplexing, though I think I understand the basics. I hope we can clarify in slightly simpler terms in class.
No comments:
Post a Comment