is296a-2 Social Information Spaces
Warren Sack

Assignment 1
Vivien Petras
vivienp@sims.berkeley.edu

Amazon.com Recommender System

1. Nonfiction Books

Starting Point: Information Retrieval (Modern Information Retrieval by Baeza-Yates and Ribeiro-Neto)

Subsets of books found: Information Retrieval, Linguistics, Computational Linguistics, NLP and Speech Recognition, Text / Image Compression, Artifical Intelligence

Outliers: C# Programming, Wireless Applications, Bioinformatics

Does this structure describe the academic field?
It is an interesting structure, but not one that I would have come up with. Some books in the field of IR that I think are also relevant, were not recommended (partly because they are not available or old). I consider C# Programming, Wireless Applications and Bioinformatics as outliers because they were "recommended" by only one book of the other fields and are not closely related to IR or the connected fields.

What I miss in this collection are books from: Web Search, Library / Catalog Searching, Data and Text Mining, Metadata Languages - I assume they are not "popular" right now, out of print, or are not bought together.

Patterns: Almost all books in the subsets relate to each other in circles rather than in a tree hierarchy. IR, Linguistics and NLP are closely related in this recommender system. The most recommended book is Foundations of Computational Linguistics: Man-Machine Communication in Natural Language (right in the middle between IR, Linguistics and Computational Linguistics).

Map: click picture to see enlarged structure; green lines are books than can be found in each others recommendation lists, blue lines are books that recommend the other book but are not recommended by it, red lines depict relations to outliers (see also Visio drawing)

 

2. Fiction Books

Map: click picture to see enlarged structure; green lines are books than can be found in each others recommendation lists, blue lines are books that recommend the other book but are not recommended by it, red lines depict relations to outliers (see also Visio drawing)


© Vivien Petras, vivienp@sims.berkeley.edu

Last updated: 09-05-2001

All cover images courtesy www.amazon.com