3 ms·
I was really excited by this writeup, so I tried it. Four test queries returned nothing that seemed useful or even relevant: fluid dynamics + electromagnetism
by leephillips 13y ago
I was really excited by this writeup, so I tried it. Four test queries returned nothing that seemed useful or even relevant:
fluid dynamics + electromagnetism : expected magnetrohydrodynamics, got Maxwell’s equations and classical mechanics (not useful);
verse + 5 - rhyme : expected blank iambic pentameter, Shakespeare, etc.: got nonsense;
writer + American + Russian + Great - Nobel Prize : expected Nabokov, got Meirkhaim Gavrielov + 1 nonsense result;
plant + illegal - addictive : expected cannabis, chronic, etc; got “Plants” (thanks) and “Nuclear Weapon” (?!? ) and some Hungarian village.
EDIT: I thought maybe I wasn't being sufficiently imaginative, so I tried "Nixon + Clinton - JFK" and got nothing that looked interesting. Then I noticed that the "Nixon" part of my query was "disambiguated" to something like "non_film", and the word "Nixon" was just stripped out. I think this thing is just broken.
- juxtaposicion 13y agoIt's not perfect; the real limiting factor is the volume of text. For the research paper behind word2vec, to get accurate associations for common words (King, man, woman, etc.) required a news sources training text of order a billion words. This is roughly the size of all of Wikipedia, but the text in Wikipedia has many more rare words, which spreads the number of training examples per word rather thin. To address that, I'm thinking of expanding the project to use Common Crawl data (http://commoncrawl.org/ http://commoncrawl.org/) to dramatically expand the available supply of data.
- tveita 13y agoI had the same experience. I haven't really gotten any convincing results except for the example queries verbatim. Some queries I tried: medicine - science + superstition aquarium - wet + dry mario - nintendo + sega luigi - green + red pocahontas - native americans + aliens