MIS 373 Lecture Notes - Lecture 9: Geotagging, Bing Maps Platform, Orbitz
26 Dec 2017
School
Department
Course
Professor
Document Summary
Document similarity & applications (i) resonance analysis (ii) crowdsourced recommendation systems. Start with a numerical representation of each document. Raw or scaled frequency of each term. Tf-idf of each term in a document. Two choices to calculate similarity between the documents. (or other) matrix of documents & terms. Counting frequency of mentions or sentiment analysis necessary but not sufficient. Web sites (e. g. , hotels. com, orbitz. com, tripadvisor. com) show by one criterion. Calculate cosine similarity between query & mentions. Perform sentiment analysis to find most positive reviews. Vicinity: e. g. , microsoft virtual earth interactive sdk (software development kit) Proximity: near a beach , near downtown , near public transportation with user geotagging & automatic classification of satellite images of areas near each hotel (e. g. , geonames. org) Additional information become extra dimensions for cosine similarity analysis. Overall results: far superior to those provided by any of the major hotel search sites. Did not discuss brands (e. g. , bmw) that are farthest from .
