ePrecis - next generation search

ePrecis - next generation search

Summary: One of the disruptive start-ups contenders emailed to me was a new search engine called ePrécis, described as "An English language processing API". According to Ward Johnson, the person who emailed me about this, ePrécis was developed by a group of linguists at the University of Minnesota.

TOPICS: Browser

One of the disruptive start-ups contenders emailed to me was a new search engine called ePrécis, described as "An English language processing API". According to Ward Johnson, the person who emailed me about this, ePrécis was developed by a group of linguists at the University of Minnesota. They worked for several years to categorize each word in the dictionary based on its linguistic context, using a 3-dimensional matrix to assign the values - effectively creating their own dictionary. An API was subsequently written to evaluate each concept and rank them accordingly.

In a paper written by James Matthewson (ex-Editor of Computer User magazine and currently on the linguistic staff at IBM-Rochester), ePrécis is compared to the Semantic Web's effort to process natural language - and in particular OWL (Ontology Language Overview). Matthewson wrote that ePrécis "can do today what OWL might only do after much rapid refinement and adoption. And unlike the Semantic Web, ePrécis requires no work by the webmasters of existing Web sites. It does all the work to gather the data, analyze it, and produce the output."

Here is Matthewson's semi-technical explanation of what exactly ePrécis is:

"ePrécis is not a program per se, but a C++ language application programmer interface (API) that can be embedded in any number of applications to return relevant outputs given a wide variety of natural language inputs. In addition to plugging into Web browsers or search engines, it could plug into word processing programs to automatically provide abstracts, executive summaries, back-of-the book indexes, and writing or translation support."

Given that description, ePrécis doesn't sound like it's necessarily a competitor to Google. Indeed, it could very well plug into Google. Later in the paper, Matthewson says that "it can seamlessly fit within any program that holds, searches, analyses, or presents information in the English language. As such, it is a chameleon technology, changing itself to fit the market."

We can compare ePrécis with Google, at www.eprecis.com, where it is deployed in a search engine format. At Ward's suggestion, I compared the results by running identical searches. First off, I tried searching for my favorite term - you guessed it, Web 2.0. Now this isn't a term that you'll find in a dictionary, so I wondered how ePrécis would perform. Well it brought back my ZDNet blog as the number 1 result and another article of mine at number 2! Very nice :-) Google's "web 2.0" search had the official Web 2.0 Conference page at number 1 and an article by some guy I've never heard of called Tim O'Reilly at number 2. So it's obvious which search engine wins on the "web 2.0" search - ePrécis ;-)

But seriously, the two sets of results were different - but nothing really suggested one or the other was better. I next tried a more common search: "daily show". ePrecis came back with the Wikipedia page for Jon Stewart's tv show at number 1, followed by an imdb.com link. Google had the official Comedy Central website at numbers 1 and 2, followed by imdb.com at number 3. Again I didn't really notice that much of a difference, other than Google's result set had the more 'official' sites ranked higher. If anything, the unofficial results that ePrecis placed higher were more interesting - e.g. the Wikipedia result for Daily Show.

Then I noticed that Ward had also contacted Amy Wohl and she had a comment on her blog: "ePrecis works best (this is a pattern matching thing) on longer words and technical queries." 

OK, so I tried a longer query: "what happened to Ted Nelson Xanadu?" Google had NIL results. ePrecis had pages of fantastic results! Just what I was looking for, in fact. I decided to try a search with more words, but not in a sentence structure (maybe that was confusing Google). I tried this: "semantic web berners-lee". Google came up with two W3C articles at the top, followed by a much-cited Scientific American article. ePrecis came back with more interesting results, links I'd click on to find out more. And that's what I want in a search engine - surprise me, show me things I haven't seen before but are relevant and interesting.

So my quick comparisons weren't entirely conclusive, but I saw enough promise in ePrecise to think that this could be potentially disruptive if it gained popularity. Perhaps a more interesting scenario is if ePrecis is paired with a Google search engine competitor, a Yahoo or an MSN, and see what happens then. What do you think? Give ePrecis a try and let me know in the Talkback section (see below).

Topic: Browser

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • PreviewSeek

    I'm personally amazed with PreviewSeek which self-proclaims to be "the world's most advanced search engine."

    Using the same query "What happened to Ted Nelson Xanadu?" (no quotes) the first result was a 1995 feature from Wired magazine specifically talking about what happened to Ted Nelson and Xanadu which neither Google or ePrecis had listed. After that were some of the same results, but more focused on people directly talking about "what happened".

    As a cool side feature it shows a thumbnail of the website (although you can do this for Google with a Firefox plugin) and a tree on the left-hand navigation of the page helping you cruise through various related items rather than simply scrolling through a long list.

    Check it out - I was impressed and have started using it rather than Google.
  • sPrecis and Restraint of Trade

    I just tried: "paraconsistent," "dialethis," and "dialethia" and got back "Error 403: Forbidden". I have SBC DSL Yahoo as ISP.
  • what the *&^% about the semantic web?

    I am not sure what you think eprecis is doing and how that relates to the semantic we. I am a researcher in the semweb and NLP is defintely not what the semweb does (at least the way W3C advocates it)

    This seems more a statistical form of deriving term-term relations.