Talk:Recommendation system

Pat's review of this article, as well as Collaborative filtering

These two articles are as they stood on 8/13 constitute a fine beginning, and I learned a lot from reading them. Some comments, questions, and ideas for additional development are detailed in the following subsections:Pat Palmer 23:01, 23 August 2010 (UTC)

Article cohesion

If I understand correctly, Collaborative filtering is a sub-topic of the more general topic "Recommendation system". I'm not quite sure why, other than length, Collaborative filtering is a separate article. The other equivalent subsections might, for consistency's sake, also become separate articles (which someone could fill in later, if not immediately). Or it all could be combined into one big article.Pat Palmer 21:47, 23 August 2010 (UTC)

Introductions

As separate articles, IMO, the introductions to both articles should clarify how the two relate. Both articles start with a very brief definition, and then launch into the history. In terms of writing style, the introductions might benefit from a statement about why this topic is important, how ubiquitous RS's are, how lay people may not even recognize that they are interacting with a sophisticated reasoning engine, or something that shows the importance of the topic. Perhaps also, a statement of intent for the coming article--will it describe how RS's are used, how they are implemented, or both? How long have RS's been in use? All this kind of stuff could somehow at least be mentioned in 2-3 paragraphs at most to set the tone for the coming article. Unfortunately, without further revisions, IMO the Wikipedia version of this topic ("Recommender systems") has a far better introduction at the moment.Pat Palmer 22:20, 23 August 2010 (UTC)

Graphics

As it stands, the article seems to assume that the reader already has a pretty good idea of what a recommendation system is. I think it would be useful, near the top, to explain it for, say, a person who is not a computer expert but who, likely, does shop in Amazon or rent movies from Netflix. Perhaps some partial screen shots from these very widely used sites (Amazon, Netflix) could be located near the top of the article(s), with captions that explain what is happening. IMO, the use of graphics and illustrations makes an article spring to life, and the harder the material, the more need for these aids. Just for example, Amazon doesn't just have one strategy going on; it is recommending on many different levels, showing what other people "who bought this" also bought, showing people's review and ratings of specific products and invited people to review products, and showing similar items from searches at the bottom of the window. All these three could be illustrated with a partial screen shot to great effect.Pat Palmer 21:58, 23 August 2010 (UTC)

Data gathering, privacy

IMO, both articles would benefit for adding information about the data gathering process. By focusing primarily on algorithms, we may miss an important aspect of RS and CF, that of how data is gathered while people shop or use websites often without the knowledge or consent of users. Perhaps not many people would object to a site keeping a history of their actions for use in recommendations for that site, but of more widespread concern is the fact that many companies sell this personal usage information to third parties. There have been, and remain, serious privacy concerns and debates about some of the data gathering tactics used by companies that implement RS's, and I think the article would gain balance by adding a discussion of these aspects of the technology.Pat Palmer 22:02, 23 August 2010 (UTC)

I just happened across a useful article on this matter, at http://www.nytimes.com/2010/08/30/technology/30adstalk.html Pat Palmer 18:54, 30 August 2010 (UTC)

References

I haven't diagnosed the reason, but there is something odd about the numbering sequence for the references at the bottom of both articles. Also, there are numerous claims in the articles that, IMO, need to have references on them but do not. For example, the IF sections of "History", near the beginning of RS, states that Lotus Notes had built-in "collaborative filtering mechanisms". Really, truly? I've worked extensively w/Lotus Notes, and I never ran across anything that I thought was that. So, this kind of statement really does need to have a reference right there so that I can check out what it was about Lotus Notes that I don't know. Another example: Pattie Maes' contributions should, IMO, be documented with a reference. Anything that is a very specific claim (for which accuracy might be open to dispute) would benefit from having a reference on it. I am also just dying to understand how Mosaic, the first internet browser, comes to be classified as a collaborative filtering system. Please elaborate!Pat Palmer 22:17, 23 August 2010 (UTC)

Links to other topics, and use of tabs

It would be nice to see links to other topics (even if the articles for them don't exist yet). Also, it would be nice to see the External Links tab and References tabs fleshed out.Pat Palmer 22:06, 23 August 2010 (UTC)