AllThingsD » Web search

Google Buys Phonetic Arts to Make Machines Sound Human

Tricia Duryee — Fri, 03 Dec 2010 22:06:05 +0000

Google is beefing up its voice services with today’s acquisition of Cambridge, England-based Phonetic Arts.

Google’s view is that voice will be critical in making mobile devices with small screens and keyboards more useful. Already, it’s launched a number of services that let people use their voice to conduct a Web search, compose emails, play songs on a phone or get directions.

The acquisition of Phonetic Arts will help in the reverse situation–when the computer speaks to you, a.k.a. voice output, the company said in a blog post.

Voice interaction is a field in bloom. There are safe-driving applications that speak your text messages to you, so you can keep your eyes on the road, and Google’s own translation “speaks” text in multiple languages. Currently, Nuance Communications is one of the leaders in the voice-recognition space.

Google says Phonetic Arts’ team of researchers and engineers will focus on making the interactions less robotic and more natural by using small samples of recorded voice.

Terms weren’t disclosed, but Phonetic Arts will be joining Google’s engineering center in London.

Meet Google's New Search Engine, Same as the Old Search Engine

Peter Kafka — Tue, 11 Aug 2009 11:29:41 +0000

Take that, Microsoft! Meet “Caffeine,” Google’s new search engine…which looks just like the old search engine.

Confused? Don’t worry–the world’s search pros are trying to figure out the difference between plain-vanilla Google and the new version, which Google (GOOG) is previewing via a blog post.

Google engineers tell us that Caffeine is a “a secret project: a next-generation architecture for Google’s web search,” but don’t expect to see any obvious differences in the new approach, even though Google says it will “push the envelope on size, indexing speed, accuracy, comprehensiveness and other dimensions.”

Or so the company says. Put the two engines up in a head-to-head test and you will see slightly different results, but only if you squint at them closely: My vanity search replaces my personal blog with my work site as the top result and doesn’t include image results on the first page, which is probably a good thing.

And sometimes new Google gives you more results, or fewer results, than old Google. But since the vast majority of searchers never get past the first page of results, this doesn’t seem relevant. Same with the supposed speed differences: Google tells me that this search for “BlackBerry Tour” came back in 0.14 seconds, compared to this one, which took a whopping 0.19 seconds. But I’m going to have to take its word on this.

If today was April 1, I’d be wary that Google was pulling our leg here and trying to pull off some meta-prank. Maybe one that involved getting lots of bloggers–or more importantly, engineers from Microsoft (MSFT) and Yahoo (YHOO)–to spend time deciphering the differences between two search engines that weren’t actually different at all. But you don’t do that sort of thing in mid-August, right?

Just How Much Search Share Does Twitter Really Have?

Peter Kafka — Thu, 18 Jun 2009 19:04:16 +0000

Twitter notched yet another milestone yesterday when it finally showed up on comScore’s index of Web search milestones. The catch: It barely registered, pulling down a search share of just 0.001 percent.

ComScore says Twitter logged 30.1 million search queries in May, more than Time Warner Cable (TWC), but not even on the same playing field as search also-rans like Ask.com.

But what if comScore is dramatically undercounting Twitter’s search–not just the standard undercounting that Web publishers always complain about, but something more significant?

It’s a given that comScore is undercounting. I know this because the research outfit told me so: The company confirmed today that it only measures searches executed at Twitter.com. But at least half of Twitter’s users are accessing the service without visiting the site, via third-party clients like Tweetdeck. And within that group of users is the power-user set, which is far more likely to be executing searches, many times a day in some cases, than Oprah fans who just joined the service last month.

So it’s easy enough to conclude that the majority of Twitter’s searches are going uncounted by comScore (SCOR). But how big is the gap? I’ve asked Twitter to share its search numbers, but I’m not holding my breath on that one. (UPDATE: See bottom of post)

In the meantime, let’s do some guesstimating.

Start with this year-old post by John Borthwick of Betaworks, who at the time was an investor in Summize, a Twitter search engine at the time (Twitter later bought Summize outright).

Borthwick reports seeing a huge number of search queries on Twitter on the opening day of Apple’s (AAPL) 2008 developer conference, topping out at an average of 190 queries per second. Tease that out over a full day, and you get 16.4 million searches in 24 hours.

For argument’s sake, let’s say that most of those searches occurred in an eight-hour stretch before, during and after Steve Jobs’s pronouncements that day, and knock that total down by two-thirds, to something like 5.5 million queries.

Steve Jobs pronouncements are rare things so it would be wrong to assume that Twitter sees similar usage patterns every day. But then again, Twitter has had an insane growth spurt in the last year: The most recent comScore traffic numbers peg monthly visitors at 32 million world-wide, up from a couple million a year ago.

See where this is going? Again, for argument’s sake, let’s say that Twitter’s peak traffic a year ago is now close to daily traffic today, and extrapolate that 5.5 million query guesstimate out for a month: You get something closer to 165 million queries.

Want to tweak any of my assumptions above? Be my guest. But no matter how you cut it, I’m sure that Twitter’s real search numbers are going to be several times higher than comScore’s number, at the very least.

Again, this matters in the end because Twitter’s most compelling investment thesis is that it can provide real-time search. And for that to mean something, the company is going to have to start registering as an actual search competitor at some point, not just to Time Warner Cable but to Yahoo (YHOO), Microsoft (MSFT) or even Google (GOOG). So how close, or far away, is that from happening?

UPDATE: Twitter cofounder Biz Stone responds, but declines to hand out any numbers. No surprise. I am a bit surprised to see him play down the importance of search at Twitter. I wonder if his investors are also surprised.

We don’t share absolute data such as total requests or queries per day but we do look at the whole ecosystem when we measure these things (not just Twitter.com).

Also, we are focused on the sharing and discovery of tweets so comparing Twitter to web search is interesting but not necessarily how we would measure success.

Web Searches That Really Bear Fruit

Katherine Boehret — Wed, 14 Jan 2009 00:18:02 +0000

There’s nothing more frustrating than a fruitless Web search — or one that returns results that distract you from your original goal. Search giant Google knows this all too well and realizes that there’s a chance you might switch to another search engine if you get tired of poor results.

This week I tested two free tools that attempt to make your Web searches more relevant by learning from users’ reactions to search results: Google’s SearchWiki and Surf Canyon Inc.’s namesake tool for Web browsers. These two don’t necessarily compete against each other; in fact, they can be used in tandem. But after initially entering a search query, SearchWiki requires additional work on the part of the user that many people may not want to do. Surf Canyon works automatically as you go, sorting results according to real-time user behavior.

[ See post to watch video ]

SearchWiki depends on people to rank their own search results by promoting favored URLs to the top of a screen and knocking others to the bottom. It is available to most people who are logged into a Google account, and these user preferences are remembered if the same searches are performed at other times.

This sorting is done using elegant animation; preferred URLs float to the top of the screen when selected and unwanted results disappear in a magic-trick-like poof when removed. Comments about a link can be typed into a word bubble beside the URL and all comments are available to the public, labeled as posted by “Searcher” unless you create another nickname for yourself. People can also add preferred URLs to a search-results page if, for example, they know a better link about something than those that show up.

Google’s SearchWiki

But who wants to do all this work? Google (GOOG) says your votes don’t influence the way other Google users see search results, nor do they affect your search results if you aren’t logged into Google. You can see the number of votes a URL got from fellow voters, as well as comments made about the URL — but only after you select a link at the bottom of the search-results page. If you promote a URL, you’ll automatically see what other people think about this link.

For your efforts, you’ll create a small collection of results that are saved in your account, sorted by date and time should you ever want to revisit them. This could come in handy in some circumstances, such as if you were researching a topic and you forgot to save Web pages as you went. Google confusingly calls these “SearchWiki notes,” though they really include all of the links you voted on, as well as typed-in notes about links.

SearchWiki is a tough sell because most of us are already trained to surf the Web quickly, skipping ahead and back through links without taking the time to rank those results or comment on them. And it only works with Google searches.

If you like the idea of more personalized Web searches but would like to use other search engines or don’t want to do extra work, you might like Surf Canyon. Once downloaded, this tool displays bull’s-eyes beside certain results to show that Surf Canyon has found additional related hits. Clicking on this bull’s-eye reveals those suggested links, pulled from deeper down in the search results, and these links might have bull’s-eyes of their own. This cascade of data goes on and on as an algorithm studies which of the returned results you do or don’t choose.

You might be deterred from using Surf Canyon because it must be downloaded before it works on Internet Explorer or Firefox. (A version of Surf Canyon for Apple’s (AAPL) Safari browser is due out within a month.) This tool works with Google, Yahoo (YHOO), Microsoft Live Search (MSFT) and Craigslist, and just started working with LexisNexis’s LexisWeb.com legal-search engine.

Surf Canyon might not seem to be doing much at first, but it changes and reflects your preferences as you make them. For example, a search for “Obama dog” originally returned results about how the President-elect and his family are narrowing their search for a puppy. But as I opened more links related specifically to Mr. Obama’s daughters, more results appeared on screen about Sasha and Malia. Each time I hit the browser’s Back button to return to the original search page, Surf Canyon offered a new set of relevant URLs.

I tried looking at Craigslist.com for last-minute inauguration tickets, and one hit listed an inauguration-appropriate dress that someone was giving away free. The Surf Canyon bull’s-eye appeared beside this result, and when I selected it, three more dress listings appeared.

Surf Canyon recently released an option for users who want long-term personalization, found at my.surfcanyon.com. It lets people select sources from which they prefer to receive news, shopping, research, or sports and entertainment results. Individual sites not listed on this page can also be added to a list of sources to use; likewise, sites can be added to a blacklist so results never come from them.

Unlike Google, Surf Canyon doesn’t save your history or usage profile. And if you haven’t created personalized preferences using the link above, it responds solely using your as-they-happen signals, like when you choose one link over another.

Google’s SearchWiki is asking users to do extra work, which may not be practical for many users. But if you do use it, this tool’s personalized, saved results could be a real boon. Surf Canyon worked well for me with multiple search engines, retrieving data from result pages I likely wouldn’t have opened. Either way, your days of futile Web searching are numbered.

Edited By Walter S. Mossberg

Email us at mossbergsolution@wsj.com. Find this and other columns and videos online free at the All Things Digital Web site: http://walt.allthingsd.com

Boost Microsoft's Lousy Search Market Share and Win Prizes!

John Paczkowski — Wed, 01 Oct 2008 18:00:32 +0000

[ See post to watch video ]

Boost Microsoft's Lousy Search Market Share and Win Prizes!

John Paczkowski — Wed, 01 Oct 2008 18:00:32 +0000

[ See post to watch video ]

New From Microsoft: Live Search SearchGimmick!

John Paczkowski — Wed, 01 Oct 2008 16:52:52 +0000

Microsoft’s Live Search Cashback–“The Search That Pays You Back!”–must have had at least some short-term positive effect on Microsoft’s search business because the company is augmenting it with another rewards program. Now, in addition to receiving Cashback rebates on certain purchases of products found through Microsoft’s live.com Web search, users can win prizes as well. The company’s new SearchPerks rewards program awards Live Search users points for every Live Search query. And those points can later be redeemed for prizes like ringtones and Xbox games, or donated to charity.

Quite a gimmick, though Live Search Senior Director Frederick Savoye says SearchPerks, like Cashback, is actually part of Microsoft’s (MSFT) grand plan to “innovate the [search] business model.” If that’s so, the company might want to rethink it. Because Cashback really hasn’t done much to bolster Microsoft’s laggard search service, which remains a very distant third in the search market. According to Nielsen Online’s MegaView search ranking for August, searches on MSN/Windows Live declined 23.8 percent year-over-year. Its August 2008 share of the search market: 10.7 percent.

A 23.8 percent YoY fall in searches doesn’t seem like much of an innovation to the search business model. And with its search market share in decline, Microsoft clearly needs something more than a steady stream of gimmicks to stem the bleeding. Said Collins Stewart Internet analyst Sandeep Aggarwal: “Internal initiatives at [Microsoft] need time to lift off and prove their merit, but sooner or later the company will have to face the realities and decide again if a combination with [Yahoo] can speed the process.”