Comments for blog.limedaley.com/post/releaseforge-for-sourceforge

Comment #2317

2008-05-04T13:39:41Z

Hi Jon

Thanks for pointing us to that handy piece of software! I'm uploading release 1.2.8 as I'm writing this. Had no login issues with the version 1.1 I installed from the repositories (apt-get install releaseforge on Ubuntu 7.10 universe)

cheers, reto

Comment #2316

2008-05-02T15:33:30Z

I spent one year at CMU working on cross-lingual information retrieval, where the query was in a different language from the documents being searched. It basically worked by translating the query but with probability weights attached to each term (how likely the word is to translate the one you typed) and carrying those weights through to the language model for each document. It was pretty similar to the vector approach but used a weighted "bag of words" instead.

Comment #2315

2008-05-02T10:43:54Z

I think you can teach an old dog new tricks. I don't know half of what you are talking about here, but I have discovered that I am better than many people at finding things through Google. I don't know why though -- I don't have a particular scheme; I just type in what seems logical to me.

Comment #2314

2008-05-02T04:54:21Z

And yes, re-reading your post more carefully I see what you are saying. I saw the "All the Boolean operators in the world won't help if the words you use are vague or you don't use enough words", to mean natural language searching, since that is the sort of thing a former employer would have said. Their idea was interesting enough - if you have a document and you want to find another one like it, just copy and paste the whole paragraph into the search engine. It then calculates vectors of different terms in the paragraph (or whole article) to decide what things you are really looking for. The typical example that they would use to explain why their search was better than keyword searching is maybe you want articles about dogs and not about cats, so you type in "dogs -cats", but then you wouldn't get the article that has pages and pages about dogs, and in one place it says, "dogs are completely different than cats" in a keyword search, but in a vector search, the cat vector would be quite small in those sorts of documents, and so could still possibly match a search where you typed: "I want to learn about dogs, but I don't care about cats" or pasting in your research paper about dogs, etc, etc.

They had one customer, and lasted for a while as a CMU startup, but the funding went away at some point, and I haven't heard of anything particularly similar. Peter has some stuff sort of related, but not for search engines, I think. The Vivisimo search engine is sort of like that, though it is more like clustered keywords. I have used it in the past (and like the name vivisimo better than clusty, though it is harder to spell) and it has always been too slow, though I suspect that it might be better for people who have a harder time typing the right keywords.

Comment #2313

2008-05-02T04:46:38Z

The only way I found it was that I remembered I had saved my Firefox bookmarks from the computer at my last job, and haven't gotten around to copying it to my current computers. I probably should have an online bookmark page instead of copying when I move between computers, though the moving happens so infrequently that I haven't looked into a better solution.

Comment #2312

2008-05-02T01:40:02Z

Heh. I wasn't trying to say that search engines needed to change, as much I was saying that people need to learn how to use search engines more effectively (I was approaching the problem as someone with a degree in English, not as a programmer) I remember reading your various posts about the funny things people type into a search engine that brings them to your site.

If it was me looking for "Release Forge" without remembering the exact title, I probably would have used combinations of key words similar to what you did. How did you find it, or re-find it, in the end?