Thursday, March 13, 2008

Abuse of Tagging System on amazon

Amazon has a feature where a reader can associate a word (called a tag) to a book, really user-defined keywords. However, this can be abused. Consider the following book:
Theoretical Aspects of Local Search
  1. fraud (131)
  2. lies (119)
  3. fantasy (103)
  4. mythology (94)
  5. morons (69)
  6. illogical (68)
  7. unintelligent (68)
  8. irony (65)
My my! Is the book really that bad? I recently had it reviewed for my SIGACT NEWS column and it seems like a fine book. Someone is abusing the tag system. But it gets worse. Look at the following book on a similar topic.
Local Search in Combinatorial Optimization
  1. fraud (131)
  2. lies (119)
  3. breathtaking inanity (111)
  4. fantasy (103)
  5. junk science (71)
  6. problem solving techniques (1)

What are the odds that exactly 131 people think that both of these books are frauds? that 119 people think both books are full of lies? that 103 people think that both are fantasy? I have it on good authority that these books are not bad. They may be too theoretical for someone's tastes, but thats no reason to trash them 131 times. So what is going on here? My guess: someone or some group does not like the area and has a program that tags things automatically.

This also happens in the humanities, which is perhaps less surprising since its more subjective.
The Columbia Anthology of British Poetry
  1. anthologies (1)
  2. avoid at all costs (1)
  3. crazy (1)
  4. cult (1)
  5. evil (1)
  6. fraud (1)
  7. insane (1)
  8. junk science (1)
  9. poetry collection (1)
  10. snake oil (1)
  11. vocational (1)
  12. discenment (1)
One big difference- people in the humanities have not learned how to write programs that will tag something multiple times.

6 comments:

  1. The number seems to say how often a tag has been used overall, for different items. Also, the tags are for finding matching items, right? So you might find the books if you search for "fraud". As an evil world domination plan, this leaves room for improvement...

    ff

    ReplyDelete
  2. This is just because the tagging system of Amazon is not spam proof! I am sure Amazon can develop some algorithms to filter out such spam. For example :

    Maintain a list of possible keywords that repeat across books. Most possibly such words have low information content which is why they repeat so often! Such words combined with a editorial list of words can be used to filter out some spam.

    ReplyDelete
  3. The whole purpose of key words is that they span many books. Key words linked to only a couple of books are not useful.

    This hardly seems like abuse. In Amazon you can review the book and state that it is fraudulent.

    In fact I checked these books and those key words are not associated with them. Perhaps "fraud" is not meant to apply to the book, but rather to the article about the book.

    ReplyDelete
  4. Indeed, the tags seem to have been cleaned up. Now, "Local Search in Combinatorial Optimization" is tagged with amazon oddities (18). Interesting. :)

    ReplyDelete
  5. if you think thats bad, take a look at any of the books available for the kindle platform. theyre ALL tagged with things like "kindle swindle" and "drm scam"... ok maybe not all, but ive yet to find one that wasnt.

    it really boggles my mind that people put so much energy into these tag-bombs.

    ReplyDelete