An unofficial blog that watches Google's attempts to move your operating system online since 2005. Not affiliated with Google.

Send your tips to gostips@gmail.com.

October 8, 2008

Machine Translation and Speech Recognition at Google

Google's technologies for automatic translation and speech recognition already have visible results: you can translate texts in 35 languages at translate.google.com or use voice to find a local business with GOOG-411, but Google intends to expand their use. You should be able to translate an email written in a foreign language or find answers to simple questions by voice.

Mike Cohen, who leads Google's speech technology efforts, and Franz Och, machine translation researcher, chat with Alfred Spector, VP of Research and Special Initiatives at Google, about two technologies that might seem unrelated to Google's core competency. Both statistical machine translation and speech recognition are search problems and Google's computer infrastructure can process large amounts of data that are needed to build language models. Another big advantage for Google is that it has popular services that generate a lot of useful data.

"When we first created GOOG-411, we had no speech data. Because we had so much query data here at Google (textual queries that people had typed to Google Maps), we could already train a pretty good language model. Now, obviously, text is a little different than speech and now that we've also trained on speech, we have better performance than we had back then, but even out of the box we could get good performance on that problem because we had so much textual data," says Mike Cohen.

October 6, 2008

YouTube Theater View

YouTube shows a new option for long videos: a theater view that enlarges the player and fades out the rest of the page. Randall Munroe's talk at Google and Steve Jobs' speech at Stanford are two examples of feature length videos that should display the new option.


You can restrict YouTube's search results to feature length videos by appending &longform=1:
http://www.youtube.com/results?search_query=site:youtube.com&longform=1

It's not clear how YouTube filters feature length videos, but most of them are from the Authors@Google series, Standford University, Carnegie Mellon University and Vancouver Film Studios.

If you want a YouTube interface that looks more like Google Video, try the YouTube Googler Greasemonkey script.

{ Thanks, Daniel. }

Better Answers in Ask.com

Ask.com went back to its roots and enhanced the snippets for search results if the query is a question. Snippets are usually excerpts that include your query, but Ask.com replaced them with the actual answers, as you can see if you search for [how to tie a tie]:


"Presenting direct answers to your searches, front and center, has always been at the heart of the Ask.com experience, and we push further down that path today with the introduction of three new answer technologies: DADS, DAFS, and AnswerFarm. These technologies take both structured and unstructured data, and - instead of delivering a title and description for each document - they deliver answers," explains Ask's blog. This is a clever idea, but Ask.com only shows at most 2 results with enhanced snippets.

The new feature is part of a broader update that mixes specialized search results with organic web results, much like Google's universal search. Ask.com goes one step further and almost eliminates standard, as you can see if you search for [Madonna]: in the top 10 results, you can find 4 web search results, one direct answer, image results, news results, event listings, video results and an encyclopedia result.

Google Spreadsheets Redesign

One more step towards a consistent Google Docs interface: Google Spreadsheets has been redesigned and it now includes the same old-fashioned menu like the word processor and the presentation app.

"We were hoping to accomplish three things with it: make it faster, make it more consistent with our siblings (documents and presentations), and give us more room to add features without clutter," explains Google.

The toolbar includes some of the most frequently used features, but I don't think Google did a job at ordering them: for example, the cell formatting options are too prominent, while pasting and sorting have been neglected. It's great that the formulas can be added without switching to a different tab and you can share the spreadsheet without opening a new page. The most stylish UI element is the chat box that can be loaded by clicking on "(name) is viewing...".


If you don't remember the previous interface, here's a screenshot:


A very cool new feature is the full-screen mode: press Ctrl+Shift+F repeatedly to alternate between two full-screen views and the standard view. The same shortcut lets you switch to the full-screen mode in the word processor, but you need to press Esc to return to the normal mode. Where is the consistency?

{ Thanks, Kevin. }

October 4, 2008

Audio Knols

Google Knol tests a cool feature that automatically converts articles to audio files, which can be played or downloaded. For now, the option is available for a small number of knols, like this one about multiple sclerosis.

"We are experimenting with Audio Playback as an option for some knols, starting with a handful of English language featured knols. If this experiment is successful, we may make Audio Playback available to more knols in additional languages, and additional features," mentions a Knol help page.


Hopefully, Google will add support for text-to-speech conversions to other services like Google Docs, Google Reader or Google Book Search.

October 3, 2008

Google Tests Image Search Ads

TechCrunch reports that Google started to experiment with displaying ads next to image search results. This isn't a surprise, since Google announced in May that it intends to test display ads.

There are different formats and positions for the ads, which combine a small image with a standard text ad. Steve Poland spotted a blended ad that looked deceivingly similar to an image result.

"The big insight of Google wasn't text ads; it was that the ads should be conducive to the format. We were doing text-based search that was all textual. Visual ads don't work in that format," explained Marissa Mayer in February, when Google started to test video ads next to web search results.


Google Image Search prepares to become more useful by adding options to find similar images, recognize faces and objects. The new features will increase the site's popularity and will attract more commercial queries that could be monetized using display ads.

No More Definition Links in Google Search?

One of the easiest ways to find definitions for words and idioms was to do a Google search and then click on the words from Google's blue bar. Google linked to dictionary.com until 2005, when it switched to the more comprehensive answers.com. Unfortunately, the feature is no longer available and users will have to find other ways to get definitions.


Here are some other ways to find definitions using Google:

1. search for [define word] and you'll get a short definition for your word, most likely from WordNet. To get other definitions from different glossaries and dictionaries, click on "web definitions for...". Example: define cup.


2. a shortcut for finding a list of definitions from the web is to use the define: operator. Just search for define:cup and you'll get a lot of definitions from a variety of source, but not all of them are authoritative.

3. subscribe to the Merriam-Webster OneBox to add a special search result with definitions.


According to Search Engine Land, Answers.com got 20-25% of its traffic from the Google definition links in 2007 and there wasn't any paid deal between Google and Answers.com. The only other important search engine that shows definition links is Yahoo, but it doesn't use a third-party service.

{ via Google Blogoscoped }

Update: The feature is back. Maybe it was just a technical glitch.

October 2, 2008

Google Homepage Time-lapse

You can go back and see how Google's homepage evolved in the past 10 years. The video uses cached versions of Google's homepage from Internet Archive.


Google's special site for the 10th birthday shows 17 of the most interesting homepage interfaces and Philipp Lenssen analyzes some of them. My favorite Google homepage is the one from April 1999 which only included a search box, two buttons and a link to "More Google!". If you like it, there's a slightly different version at google.com/ie_rsearch.html.

Google linked to many pages that no longer exist, but they're still available at Internet Archive:
* some credits from 1998
* press coverage from 1998-1999
* older versions of Google's logo
* Google's affiliate program
* an old Google tour
* the special Katrina page.

The Invisible GoogleUpdate.exe

Most of the recently released Google software uses live installers that download the required files from Google's servers and enable an auto-updater. Whether you want to install Chrome, Gears, Lively or Google Earth's plug-in, Google first downloads a small setup file that needs an Internet connection to obtain the software.


The most important reason why Google doesn't provide offline installers for these applications is the auto-update software, which has a separate development cycle. Instead of integrating an auto-update feature in all the applications, like in Picasa, Google Desktop or Google Toolbar, Google decided to create an independent tool that deals with keeping Google software up-to-date.

"GoogleUpdate.exe is a software component that acts as a meta-installer and auto-updater in many downloadable Google applications, including Google Chrome. It keeps your programs updated with the latest features. More importantly, GoogleUpdate allows your Google applications to be rapidly updated if security flaws are discovered," details Google.

Web applications can be easily updated and everyone gets the latest version, but desktop software needs to be updated manually or by the software itself. Google chose to automatically update most of its applications, in many cases without providing an option to disable the updates and without informing the user when a new version is installed.

GoogleUpdate automatically runs in the background when you start your computer and it connects to Google's servers every few hours to check if there are updates and to report some usage data.

"When GoogleUpdate communicates with Google servers, it sends IDs of GoogleUpdate-managed applications on your computer and general usage information for these applications. GoogleUpdate also uses its own, randomly-generated unique ID number to accurately count total users. This information includes version numbers, languages, operating system, and other install or update-related details, such as whether or not the applications have been run."

GoogleUpdate is installed as a system service, which can be disabled from the services console, but there's no transparent option to uninstall it. It also installs the GoogleOneClick plug-in for Firefox, IE, Chrome that allows Google to launch the updater directly from a web page. If you kill GoogleUpdate.exe from the task manager, a scheduled task will reopen the service when your computer is idle. Google says that the service is uninstalled a few hours after you uninstall the last application that uses it. "Google Update Service uninstalls itself when there is no Google software using it. It may take a few hours after uninstalling Google software for Google Update to uninstall."

To make things even more difficult for those who manually remove the updater, "you may find that your Google programs no longer function properly and, in many cases, you may see GoogleUpdate return automatically".

Even if the software's intentions are noble, it's unfortunate that Google doesn't inform users about the updater and it doesn't provide an option to disable GoogleUpdate or to ask before downloading updates.

October 1, 2008

Google News for Blogs

Google Blog Search's homepage has been updated and it now includes a list of popular stories, which are categorized and clustered similarly to Google News.

"Blog Search uses a set of algorithms to try to determine the most popular stories in the blogosphere. We consider factors such as a blog's title and content, as well as its popularity throughout the rest of the blogging community. Then we display groups of posts that are closely related," explains the FAQ.


It seems that Google Blog Search didn't select some authoritative blogs to be listed on the homepage and almost any blog post could be included in a Blog Search cluster. There are 11 categories: politics, US, world, business, technology, video games, science, entertainment, movies, television and sports, but they aren't equally represented.

Google shows a chart for the number of blog posts that discuss a story:


Unfortunately, Google is not able to detect spam posts, plagiarism, forums or blog comments. Google Blog Search doesn't provide yet feeds for the top stories and the search results aren't clustered.


The new Blog Search homepage is a decent alternative to the technology-focused Techmeme and to Technorati's front page. Hopefully, Google will filter the low-quality posts and it will integrate the service with Google News.

{ Thanks, Jonathan. }