Intended to be a demonstration of Clever Sense’s underlying data extraction and machine learning technologies, Alfred creates a personalized “interest graph” that then serves up local recommendations for places you’ll like. But you can also search in a more conventional way.

Of course Clever Sense isn’t the first company to create a “recommendation engine” or local “discovery engine.” Currently Where and Bizzy position themselves in this way. (Where was acquired by eBay and may ultimately turn into something very different however.)

Google is also trying to use social signals to offer local recommendations in mobile.

Clever Sense’s Alfred uses a mixture of Q&A “teaching,” social signals and other techniques to construct your interest graph. I found the teaching part to be less tedious and onerous than on comparable apps. And relatively quickly it was making some pretty solid recommendations.

Clever Sense describes how the app works:

The Extraction Engine built into the Clever Sense Platform curates large amounts of unstructured crawled data by leveraging natural language processing, statistical machine learning, and data mining algorithms.

As it reads through the data, the Extraction Engine learns meaningful concepts that are descriptive of physical items in the real world. These concepts constitute the foundation of the Clever Sense Interest Graph. The engine further leverages social interactions like check-ins, likes, and ratings to enrich the Interest Graph. It calculates similarities via graph-based algorithms using these social interactions.

I spoke with CEO Babak Pahlavan several weeks ago. He gave me a demo and described Clever Sense’s “interest graph” as “Pandora for the real world.” In other words Clever Sense assigns or maps physical places to one another based on styles, characteristics and attributes in the same way that Pandora does for music. However Clever Sense is not using humans to tag or otherwise “curate” data. It’s all done by machine.

Pahlavan also explained that the interest graph is comprehensive, so the app will enable users to visit an entirely new city and get instant recommendations that are similar to places they like at home. That’s a pretty compelling use case I think.

Click here to view the embedded video.

SearchReviews has just launched a search engine with more than 40 million reviews in its system, and plans to hit 100 million reviews by the end of 2011. Those 40 million reviews cover about four million products and come from more than a thousand sites such as TripAdvisor, Amazon, Zappos and others. Those are impressive numbers, but is SearchReviews any good? More on that below along with a look at how it works.

How SearchReviews Works

SearchReviews appears to have the largest searchable database of strictly review content anywhere. Buzzillions advertises having 13 million reviews; Bazaarvoice claims to have “served” 175 billion reviews, but that happens on its client’s sites; Bazaarvoice doesn’t offer a direct search engine of reviews. There’s also Google Places, which has an enormous collections of reviews and sources; but the company has repeatedly declined to share any statistics in the past and didn’t reply to our e-mail in time for publication.

One thing that distinguishes SearchReviews from these competitors is the mix of review content. Whereas Buzzillions has product reviews (shoes, electronics, etc.) and Google Places has place-related reviews (restaurants, hotels, etc.), SearchReviews has both. That offers a somewhat more convenient search experience, especially for mobile users who may be looking for a place to eat and shopping at the same time. SearchReviews launches with apps that are available for both the iPhone and Android devices.

SearchReviews crawls more than thousand websites for review content and aggregates them into its database. The company says it’s adding about two million reviews every week. Right now, it has no formal partnerships with any of the sites that it’s crawling; a spokesman says SearchReviews hopes to do formal agreements now that the site has launched and the review sources can see the search engine in action.

A Look at SearchReviews

Here’s a look at the results for “new york hotels” on SearchReviews:

Clicking on any result brings up an interstitial page that shows the SearchReviews result as an overlay above the actual source of the review. This is how SearchReview says it still “drives traffic and revenue back” to the original source (but those sites may still not appreciate the interstitial).

The site offers a number of tools that you’d expect to find: filter by star rating, by source, and by location. There’s also a tag cloud for further drilling down, or just for getting a sense of the sentiments related to the search results.

Those filters are necessary because there’s no apparent rhyme or reason to how the initial set of results is ranked, and tens of thousands of reviews of New York hotels is useless without further refinements. The filters also reveal what I’d call a flaw in SearchReview’s system: over-reliance on certain sources. Consider the search above for “new york hotels.” Looking at the source list shows that TripAdvisor is the source of more than 99% of the search results, while Yahoo Travel provides less than 100 of the 52,000+ reviews.

No doubt TripAdvisor is a valuable and trusted source, but SearchReviews will need more variety if it’s to add value as a reviews search engine. The problem is somewhat worse on the iPhone version because those filtering options aren’t (yet?) available; on the search shown earlier in this article for the LG BD550 Blu-ray DVD player, the 20 reviews I saw were all from Walmart.com.

One of the impressive aspects of SearchReviews is that it indexes review content, not just business and product names. So, someone looking for “spicy thai food seattle” can find reviews that very specifically mention those words.

SearchReviews has a couple extra features I should mention: a tool that lets searchers connect with Facebook friends to ask for recommendations and a widget for publishers to embed SearchReviews content on their sites and share in the site’s advertising-based revenue.

All in all, SearchReviews is a bit rough around the edges right now, but with some further development it looks like it could become a promising source for consumers seeking out review information.

There’s a relatively elaborate point system (a “currency”) used to pay for answers and as a reward for participation otherwise. Because it’s globally available in English we’re likely to see broad participation from outside of Africa in Baraza.

The more interesting question in my mind is whether this will become a model for a larger Q&A effort at Google. The company course owns social search/Q&A site Aardvark, which it hasn’t done much new with as far as I can tell. I did think I saw a LivePerson-like implementation of Aardvark for AdWords support but now I’m unable to find the page.

Google offered an early Q&A service (paid) in Google Answers, which was shuttered after the rise of the free Yahoo Answers. There are numerous sites and services (at least 10), including Facebook and Quora, that offer “social search” or Q&A functionality. None have emerged as a replacement for traditional search or really lived up to the promise of human Q&A otherwise.

Related posts:

• The Rise Of Help Engines: Twitter & Aardvark
• Will Quora Challenge Google? No. Is It Useful? Yes
• Ask Comes Full Circle With “Q&A” Offering

Also in this category is the yet to launch “personal assistant” Siri, which taps artificial intelligence rather that real people and is starting on the iPhone but will extend to the PC as well.

ChaCha, which is free and ad supported, can accept text or voice queries and provides a text response back. It uses part-time human agents to answer questions. ChaCha began life as a “social search engine” on the PC and morphed into text-based mobile answers service. More recently the company launched mobile and online coupons to diversify its revenues.

Competitor kgb is also text-based but today launched iPhone and Android apps that offer a game-like experience around the core answers offering. Its model requires consumers to pay $0.99 per question and it doesn’t have ads. However the iPhone and Android apps provide access to a range of features and content for free.

Peer-to-peer help or answer engine Aardvark, which now is calling itself a social search engine, has multiple entry points: Web, Twitter, IM and the iPhone.

Mosio was also an early entrant in this segment but has shifted its focus somewhat to the enterprise market. It still maintains a mobile answer community however.

There are and have been many “Q&A” sites on the PC, exemplified by Yahoo Answers. There are also numerous “social search engines.” They’ve had varying degrees of success but are not widely regarded as true alternatives to Google, Yahoo or Bing. However on a mobile device, where there is less screen space and less patience, there is an opportunity to develop a differentiated offering. Accordingly, these mobile answers services are more directly — or by default — positioned as alternatives to search engines. And consumers may be willing to consider them.

I recently did an informal comparison of a number of these services and Google and found that there was no clear winner; each had strengths but none emerged as a clear victor. Yet these mobile,  human-powered competitors must provide a richer, faster or more engaging experience than conventional search engines more generally if they hope to succeed over the long term.

Large corporations have had various “buzz tracking” tools at their disposal to monitor “brand chatter” but these have not been available for SMBs. Marchex has now created a comprehensive reputation management tool aimed at the SMB market that enables monitoring of virtually everything being said about them online via a single dashboard. I’ve seen the demo and it’s a very impressive offering. The search tie-in is that the tool is built on top of Marchex’s OpenList search and reviews aggregation technology. During the beta period SMBs can get direct access but after that it will apparently only be available through Marchex’s partners (e.g., YellowPages.com) and not directly to SMBs. I think that’s unfortunate because it’s a valuable tool and service that SMBs would directly pay for.

The press release summarizes the content and data being collected:

Marchex Reputation Management was built using proprietary local search technology from Marchex’s Open List local search network. Covering more than 8,000 sources and currently containing more than a quarter billion pieces of meta-data, Marchex Reputation Management monitors and reports on a specific business’ online footprint, including its user reviews and news, blog and social media mentions. In addition, Marchex Reputation Management ensures the accuracy of information in existing online business listings found on general search sites, local search sites and directories, such as consistent business name, address and phone numbers, as well as making recommendations as to where to add new listings for more coverage. This information is refreshed daily.

Marchex Reputation Management blends Marchex’s heritage in local search with its proven ability to partner and deliver local advertising solutions. Specific features include:

• Reviews: Charts break out the percentage of positive, neutral or negative reviews in an easy-to-read display.
• Mentions: Charts demonstrate activity by reviews or mentions to tie the information back to particular changes in the business or marketing efforts by month.
• Business listings: Ensure business listings are accurate and visible on top online consumer destinations.
• Keyword identification: Top keywords and phrases are highlighted to identify the terms that differentiate the business and act as an alert to any customer service issues.
• Competitive marketing analysis: Compare a business to other local businesses to analyze how its reputation and marketing efforts stack up.
• Ability to engage: Share positive news and reviews with customers and employees through email, Twitter, Facebook, Digg or other sources.

The tool can be used to determine where business listings exist online and how consistent and complete they are; but it’s most valuable, as the name suggests, to track reviews and what’s said about a business. SMBs can also compare themselves to others and direct competitors. And there’s the ability to share reviews via Twitter, Facebook, etc, which takes this beyond simply a monitoring tool into the realm of outbound marketing:

I spoke with Marchex EVP of Product Engineering Matthew Berk, whose OpenList was acquired by Marchex in 2006 and is the basis of this tool, and he said that SMBs in the beta test are figuring out that they can use this not only for monitoring of reviews but also to glean information for later marketing purposes. It has long been the case that proactive SMBs have used online reviews to improve their service or identify problems but there hasn’t been an easy way for them to get all the relevant information in a single tool or dashboard.

GetListed does a good job of helping businesses identify where they show up online and provides a number of services for SEO purposes. But in terms of monitoring online ratings and reviews of local businesses and building a suite of services around that, there isn’t anything else in the market (to my knowledge) that does this.

Click here to view the embedded video.

We saw the power of opinions posted online in our global study earlier this year about trust in advertising, and the point came up again in our recent findings. Social media is becoming a core product research channel. Almost 15 percent of Socializers most trusted information they found on blogs when researching new purchases online, while nearly 20 percent trusted most the information they found on message boards.

Then the provocative question is asked: “So are social networks replacing portals or search engines?” Nielsen then says:

Perhaps. Regardless, if we don’t understand and address people feeling increasingly alienated by the amount of information on the Internet, and the need for a human guide, yes, your favorite social network (or something like it) will become the next great content gateway.

But it’s not an either/or, zero-sum situation. There may be some number of queries that people pose to their networks before or in addition to their use of search engines. Overall, “word of mouth” on social media sites is a complement to search engine usage. No doubt social media tools will continue to grow as content sources; however I don’t think we’ll see Facebook or Twitter replace Google or Bing any time soon.

But in what is a first (to my knowledge) Idearc’s Superpages has brought true local search to Twitter through Twitter.com/sp411. Twitter users can now conduct searches of the Superpages database and get results as direct messages on  Twitter to their local queries. (They can also retweet and share them, which makes it that much more interesting.)

Here’s the process:

1. You follow sp411 and then it will automatically follow you a few seconds later
2. You then send a direct message to sp411 (“d sp411″) with a query and location. Example: pizza in Seattle, “d sp411 pizza Seattle”.
3. Results will appear in an all the Twitter notification places (email, SMS and direct message).

Here’s what it looks like:

Just a couple of seconds after the direct message is sent you see something like this:

If you click through on any of these links you are taken to a business profile page on Superpages’ site:

After the initial set up process, in which you follow sp411 and it follows you, the service proves to be quite useful especially if you’re looking for a specific business phone number or address. It’s also an example of Superpages proactively taking its data and local search capabilities out to where users are, not unlike what many companies have been doing for some time on Facebook.

And in case it isn’t self evident, this also makes Twitter/sp411 a local-mobile search tool as well.

There are a couple of other Twitter bots out there but this is the first local search tool on Twitter. I would imagine, as people see and start to use this, we’re going to see a bunch of vertical and other search tools employing similar functionality. And as they start to emerge they make Twitter a much more interesting and useful site.

If you’ve used search.twitter.com, you’ll be familiar with FriendFeed’s real-time search — the interface is almost exactly the same. One difference that FriendFeed is quick to point out is that its results update automatically and scroll down the page, without the need to click or refresh. The other, bigger difference, of course, is that FriendFeed’s search pulls in data from a wide variety of sources (including Twitter).

The FriendFeed search includes options to save your searches or view them in a smaller, pop-out window. You can also post your search on other social sites, or grab some code to embed the search elsewhere.

But one thing that doesn’t seem right is the randomness of search results.

That’s a close-up of the image I used above showing four results that are in nothing close to chronological order. All the searches I tested today had the same problem.

Earlier this month I profiled the new OneRiot Twitter search engine, and now via Louis Gray, we learn about Twazzup, another new entry into the Twitter search race. Where OneRiot is focused on sorting through the Twitter noise and helping users find just the links and web content that people are talking about, Twazzup takes everything Twitter has to offer — noise and all — and lets users find what interests them most.

Here’s a look at the Twazzup interface (you can click to see the larger version on Flickr) after a search this morning for “U2″:

A stream of tweets shows up in the main content block on the left. These are shown in real time, although it appears that Twitter is running slow this morning — both Twazzup and Twitter’s own search engine are showing tweets about 15-20 minutes delayed as I write this. Twazzup shows recommended search refinements in small, gray boxes above the real time stream; a series of U2 concert tickets went on sale today, so some of the refinements include “vegas,” “houston,” “tickets,” and “muse” (one of the opening acts).

Twazzup’s unique content is to the right:

• Popular Tweets – this needs work; the example in the screenshot is from a coupon service and has nothing to do with U2. It’s there because ‘u2′ appears in the shortened URL. (Seems like an invitation for spamming via custom short URLs.)
• Top Trendmakers – accounts whose tweets are driving conversation
• Related Photos – related photos from Twitpic, updated in real time
• Most Popular Links – links most often included in tweets related to the keyword

Aside from the “Popular Tweets” issue I mentioned above, Twazzup provides a good Twitter search experience — and one that’s different enough from others to make it a valuable addition to the real-time search engine field.

OneRiot, a real-time search engine for web and video content, has launched a new Twitter search engine with a unique angle: Rather than focus on what people are saying, the search engine focuses on the web pages people are linking to.

“Our approach is, you’d go to Twitter for the conversation,” says OneRiot GM Tobias Peggs, “but you’d come to OneRiot for the content.”

How It Works

One Riot indexes tweets, looking for messages with embedded links, then crawls and indexes the content being linked to. The OneRiot algorithm includes spam analysis and has checks for timeliness, relevance, and a “hotness”/freshness factor for URLs that are being talked about a lot in recent tweets. Here’s what it looked like yesterday on a search for iPhone:

As you can see, the focus is on content — on the URLs people are discussing and sharing on Twitter. Search results are displayed in the familiar Title-Snippet-URL format. But Twitter is a social site, so OneRiot’s search results also include some social elements for anyone who wants to dip into the conversation. Each result includes a reference to how recently the URL was shared on Twitter, how many tweets mention it, and who first found it on Twitter. Clicking on the “Shared in (N) tweets” link displays the tweets that mention this URL:

(you can click for the larger version of both images on Flickr)

It’s a unique approach to Twitter search that should prove particularly beneficial to the growing business community on Twitter. In my short time testing the service, I found that it does bypass a lot of the noise and chatter, and leads more quickly to relevant content. And there’s a convenience factor, too, in that OneRiot’s search expands shortened URLs, so you can see the actual content people are talking about without having to click first on a TinyURL or some other link.

An alpha version of their Twitter search engine is available now at twitter.oneriot.com. The company plans to get user feedback for at least a few weeks before eventually moving Twitter search to the main OneRiot.com search engine. You can message @OneRiot on Twitter to give feedback.