Sunday, November 19, 2006
Distinguish the homonyms on Google
Researchers of the University of Tokyo have developed a software which will be able to select results from a Google query and distinguish unique names and identities from the results.The NewScientistTech gives the example of Michael Jackson: Michael Jackson is, of course, a big expert in beers, but he's also a old american singer in eighties. Those two persons are on the first page of a Google query and this is the problme: which web site talks about who ?
When a user searches for a name on web search engine, the program will look at the first 100 results returned, and examines common words in the search summary to see if the results will relate to different people of the same name. The program will also give users an estimation on how many different identities for the same name have been returned.
The firsts tests reveal that the software is between 70% and 95%. Co-developer of the program, Danushka Bollegala, reckons that the tool will really help webwatchers to refine their searches: "The keywords extracted by the algorithm can be used to suggest better queries to the user".
When writing those lines, Bollegala has not been bought by Google yet...
Source: NewScientistTech, thursday.
Labels: google
Saturday, October 28, 2006
A Google Calendar clock
Here is the Ambien Clock, a little desktop clock that will connect to your Google Calendar account. Your appointements then appear in the corresponding time area and also the background colour changes according to your planning: blue - nothing foreseen, yellow - move your a***, orange - too late.The clock connects via the Ambient Information Network, a technology developed by the company Ambient which also developed others similar gadgets. And since it works on AAA, this device is really wireless...
... but maybe we would have preferred Wifi. Anyway, it's just a prototype in beta test: you can go to the website and help the engineers to decide the best look.
Labels: google
Wednesday, October 04, 2006
The google translation machine
google
Peter Norvig, the director of research at Google, presented last week the company's newest innovations and the future of data analysis.He first spoke about the Statistical Machine Translation, a translation software currently under development. The goal of this program is to improve the translation's accuracy (we guessed that), and to make it more human. Indeed, the current Google Translate is not perfect at all.
A translation software is really a good example of how the ability to utilize large amounts of data is helping to expand the resources available to everyone.
To summarize, it means that with all the google's services...
- classical search on Internet (google search);
- search on blogs (blogger or google blogsearch);
- search in the press (google news);
- search in the books (google books);
- search in academic articles (google scholar);
Norvig also spoke about others projects on the same principle, as the improvment of google sets, that provides 5 keywords related to one keyword, or a new application that can show the trend graphs that track the volume of different searches at different times of the year (a little bit like google trends, finally...)
Source: The Daily Californian, last week.
For those who do not know:
- google books, google scholar ... at least two services that are 99,9% non-spam certified, i.e. without data that fools the reader. A perfect content for this subject.
- two new options recently appear in google translate: Arabic-English and English-Arabic. They use the prototype of this Statistical Machine Translation. Here are test results.
Labels: google
Friday, July 07, 2006
My poor Search Engine Marketing...
google 
Companies like BCC, Heytens or Nextiraone required their services to improve their position on the search engines. An unenthusiastic result when we know that Google removed these Belgian websites from their index, after an attempt of dupery. Duping a search engine, for example, is to show him a content that the human user won't see.
And as a human user, that's what I want, finally: the result that matches the more with my request. On the other hand, as a specialist marketing, I wouldn't open my mouth.
During that time, PublicityWeb is in touch with Google US in order to hope, in the next days, a re-validation of the (now corrected) websites. And if some of them seem to re-appear, other ones don't but the agency's website itself, not at all.
Source and more objective details: DataNews, last week.
For those who do not know:
- Google regularly updates a list of instructions that webmasters are supposed to respect.
- Search engine marketing, or SEM, is a set of marketing methods to increase the visibility of a website in search engine results pages (SERPs).
Labels: google
Wednesday, June 07, 2006
Back to the future
google What looked like Google at the beginning ?
The wonderful website Internet Archive lets you travel back in time thanks to its Wayback Machine !
Here is what Google looked like ...
... in december 1998 :
... in 1999, the purified style that we know :
... in 2000, a funny thing :
... in 2001, a tour :
note : not all logos were archived, unfortunately
In 1998, we could read:
For pleasure : google in the time.Google Inc. was founded in 1998 by Sergey Brin and Larry Page to make it easier to find high-quality information on the web. The company is based on three years of research in web search and data mining done by the founders in the Stanford University Computer Science Department. Google Inc.'s headquarters are located in scenic downtown Palo Alto, California.
Google Inc. is not at present a publicly traded company, and we are currently unable to speculate on whether or when our privately-held status might change.
10^100 (a gigantic number) is a googol, but we liked the spelling "Google" better. We picked the name "Google" because our goal is to make huge quantities of information available to everyone. And it sounds cool and has only six letters.
for those who do not know :
Pour ceux qui ne savent pas : Data Mining, also known as Knowledge-Discovery in Databases (KDD), is the process of automatically searching large volumes of data for patterns. Data Mining is a fairly recent and contemporary topic in computing. However, Data Mining applies many older computational techniques from statistics, machine learning and pattern recognition. As usual, wikipedia is full of details. On the french website, we find this anecdote:
The first tests of excavations of data were done historically on million sales tickets of a supermarket, as memorized by the cash registers. At the origin of the popularization of the methods and algorithms, there would have been the description by the Wal-Mart stores of a very strong correlation between the purchase of layers for babies and beers, every saturday afternoon. The analysts realized that men are sent by their housewives to buy the bulky packages of layers for baby. Thus, the shops were reorganized to present the beer next to the babies stuff. Sales climbed out of arrow! This veracious image illustrates the return on investment (ROI) of datamining and more generally of decisional data processing.
Labels: google
Thursday, June 01, 2006
the hOwGee researcher (beta)
the Researcher© smartly uses Google to find stuff hidden on the web.
please let a comment in order to improve the researcher
Labels: google
Sunday, May 14, 2006
5 minutes to create its homepage
google Manual to personalize google by adding feeds :
This page contains last e-mails, live news and last posts of the blogs I like.
If you already use that, it seems old-fashion. Actually, I noticed that many people do not know this kind of system and are afraid with RSS flows.
Of course, it's an example .. Netvibes and others use the same principe and are very good, too.
Step1) get a Google Mail address.
If you don't have, write to me. Anyway, without a gmail address these days, we're nothing :/
Step2) go to google/ig
Step3) log in by clicking on "sign in"
Step4) a default homepage appears. Remove everything you do not need, by clicking on the crosses (Windows-like).
Step5) click on "add content"
Step6) add Gmail and other stuff, if you wish.
Step7) add RSS feeds from your favorite websites (news, blogs, ...)
in "create a section", paste http://howgeen.blogspot.com/atom.xml
(it's an example but let's do it, it's good)
To add directly the hOwGee feed, and if you already have a Google account, simply click here : 
To find others RSS feeds, surf on your favorite websites by looking for this logo :
for those which do not know : RSS feed.
Labels: google
archives >> April - March - February - January -December - November - October - September - August - July - June - May
