Thirteen Million Wikipedia Paragraphs at your Fingertips
by Gabriel Weinberg, February 2010
13,374,899 full Wikipedia paragraphs now appear in our Zero-click Info boxes when using our search engine.
Recent real examples:
- irs 8889 (compare w/ Google, Yahoo & Bing)
- mercedes distronic (compare w/ Google, Yahoo & Bing)
- catholic church yarm (compare w/ Google, Yahoo & Bing)
- python switch (compare w/ Google, Yahoo & Bing)
Previously our abstracts were limited to initial sentences from source articles. For example, see our Apache Solr results page.
With this new change, our Zero-click Info boxes should appear on many more result pages. You no longer have to match a topic name nearly exactly. If there is a Wikipedia paragraph relevant to your search, we will show it to you in its entirety.
This last point is subtle, but crucial: in its entirety. Showing you the full paragraph (and where it came from) gives the snippet context and makes it useful in a self-contained way.
So much so, that if you're searching for something in Wikipedia, you may not even have to go there anymore. And if you do, the link we give you is deep-linked into the right section.
We've taken great care to keep these paragraphs extremely relevant to your searches. We're doing some NLP and using available contextual information.
We have many improvement ideas already though. We're releasing now because we think it improves search results as is.
Over the next year, we hope to expand our data sources and adapt our algorithms to be even more sophisticated in order to show you even more Zero-click Info, more of the time.
As always, we'd appreciate your ideas and comments!