Semantic search vs Google: In Medicine May 26, 2008
Posted by Bertalan Meskó in Innovation, Medical Search, Medicine, Semantic Web, Web 3.0.trackback
I’ve been playing with Powerset for a while. It seems to be a service that can take us to the world of semantic web or web 3.0. It uses Wikipedia and Freebase as resources. The main idea is to ask questions instead of search for terms. Let’s give it a try.
If you make a search for “Who discovered penicillin” in Google, you will see this (Alexander Fleming Discovers Penicillin) and many more similar articles. Even if we know the truth is different.
If you ask the same question in Powerset, you get this:
It’s a bit more accurate, isn’t it?
Try it and let me know if you find something interesting. And don’t forget to check out my Personalized Medical Search!
Further reading:
I have mostly been unimpressed with Powersets results. Too unpredictable in quality, and limited indexing capability (and scaling will be hard since NLP is expensive). I like the UI and believe it supplements traditional links, based search.
Where Powerset will do well is if it is searching structured information. Its results when using Freebase especially are quite decent.
[...] The main idea is to ask questions instead of search for terms. Let’s give it a try.” Article Bertalan Meskó, ScienceRoll, 26 May [...]
I have been in this business for so long, since the web started in 1993, you had to use a browser called Mosaic, and I am the least impress from the lack of intelligence on the search engines. But just recently Yahoo launched their drop down search which completes sentences as to search terms, still I like the idea of asking questions, and I thought Ask Jeeves was taking the lead on that. There are so many useless search results, and I know google is implementing better algorithms to filter out forcing search results to produce meaningless queries. I think the question strategy allows to narrow results, way to go!
Hi Bertie,
I think I disagree with your statement that “The main idea [of the semantic web] is to ask questions”. I think it has nothing to do with the way queries are phrased (e.g. “discoverer of penicillin” vs. “who discovered penicillin?” vs. “penicillin discovery”) and everything to do with realizing that language is full of synonyms and that words have relationships to other words. Semantic search has the potential to change search in the same way that the use of stemming algorithms in search engines did. Stemming algorithms reduce words to their root, so “fishing”, “fished”, and “fishes” all reduct to the same word “fish”. This means that if you want to find about fishing the Hudson river, you don’t need to search on “fishing hudson river”, “fish hudson river”, etc. Rather than strict word matching, most search engines allow you some leeway in your queries and give you results that you perhaps didn’t explicitly ask for, but are clearly relevant.
A semantic search engine would be able to “understand” that “discoverer” and “inventor” mean essentially the same thing in the context of penicillin and return results not just for “discover penicillin” but also pages that match “inventor penicillin” etc. A semantic search engine might also display results about benzylpenicillin and phenoxymethylpenicillin (which are types of penicillin) or “beta-lactam antibiotics” even though the word “penicillin” is not a matched word in the document.
Oh, and that was not intended to be an emoticon in my last post, but the concluding parenthesis.
I am most eager to give PowerSet a try for medical research purposes. Thank you for your posting. I will also give your personalized medical search a try!
Thanks
Diane, I’m looking forward to your feedback.
Hilary, thank you for the long explanation. Now I think I understand it better.
I used to use answers.com