I have many articles in a database (with title,text), I'm looking for an algorithm to find the X most similar articles, something like Stack Overflow's "Related Questions" when you ask a question.
I tried googling for this but only found pages about other "similar text" issues, something like comparing every article with all the others and storing a similarity somewhere. SO does this in "real time" on text that I just typed.
How?
You can use SQL Server Full-text index to get the smart comparison, I believe that SO is using an ajax call, that does a query to return the similar questions.
What technologies are you using?
I suggest to index your articles using Apache Lucene, a high-performance, full-featured text search engine library written entirely in Java. It is a technology suitable for nearly any application that requires full-text search, especially cross-platform. Once indexed, you could easily find related articles.
SO does the comparison only on the title, not on the body text of the question, so only on rather short strings.
You can use their algorithm (no idea what it looks like) on the article title and the keywords. If you have more cpu time to burn, also on the abstracts of your articles.
Seconding the Lucene suggestion for full-text, but note that java is not a requirement; a .NET port is available. Also see the main Lucene page for links to other projects, including Lucy, a C port.
Maybe what your looking for is something that does paraphrasing. I only have cursory knowledge of this, but paraphrasing is a natural language processing concept to determine if two passages of text actually mean the same thing - although the may use entirely different words.
Unfortunately I don't know of any tools that allow you to do this (although I'd be interested in finding one)
If you are looking for words that wound alike, you could convert to soundex and the the soundex words to match ... worked for me
It depends upon your definition of similiar.
The edit-distance algorithm is the standard algorithm for (latin language) dictionary suggestions, and can work on whole texts. Two texts are similiar if they have basically the same words (eh letters) in the same order. So the following two book reviews would be fairly similiar:
1) "This is a great book"
2) "These are not great books"
(The number of letters to remove, insert, delete or alter to turn (2) into (1) is termed the 'edit distance'.)
To implement this you would want to visit every review programmatically. This is perhaps not as costly as it sounds, and if it is too costly you could do the comparisions as a background task and store the n-most-similiar in a database field itself.
Another approach is to understand something of the structure of (latin) languages. If you strip short (non-capitialised or quoted) words, and assign weights to words (or prefixes) that are common or unique, you can do a Bayesianesque comparision. The two following book reviews might be simiplied and found to be similiar:
3) "The french revolution was blah blah War and Peace blah blah France." -> France/French(2) Revolution(1) War(1) Peace(1) (note that a dictionary has been used to combine France and French)
4) "This book is blah blah a revolution in french cuisine." -> France(1) Revolution(1)
To implement this you would want to identify the 'keywords' in a review when it was created/updated, and to find similiar reviews use these keywords in the where-clause of a query (ideally 'full text' searching if the database supports it), with perhaps a post-processing of the results-set for scoring the candidates found.
Books also have categories - are thrillers set in France similiar to historical studies of France, and so on? Meta-data beyond title and text might be useful for keeping results relevant.
One common algorithm used is the Self-Organizing Map. It is a type of neural network that will automatically categorize your articles. Then you can simply find the location that a current article is in the map and all articles near it are related. The important part of the algorithm is how you would vector quantize your input. There are several ways to do with with text. You can hash your document/title, you can count words and use that as an n dimensional vector, etc. Hope that helps, although I may have opened up a Pandora's box for you of an endless journey in AI.
Edit distance isn't a likely candidate, as it would be spelling/word-order dependent, and much more computationally expensive than Will is leading you to believe, considering the size and number of the documents you'd actually be interested in searching.
Something like Lucene is the way to go. You index all your documents, and then when you want to find documents similar to a given document, you turn your given document into a query, and search the index. Internally Lucene will be using tf-idf and an inverted index to make the whole process take an amount of time proportional to the number of documents that could possibly match, not the total number of documents in the collection.