Viking

From the same place that gave birth to Siri (the cool app that functions as a personal assistant by listening to what you ask it to do) comes a new search engine called Trap.it. It’s pitched as a “personalized search engine”, but really functions more as a curated information recommendation engine. With trap.it, you create “traps”, which is a stream of recommendations seeded with a keyword search. The idea is that you thumbs up/thumbs down the content given to you, trap.it learns over time, and gives you better results. In concept, it feels quite similar to the now-defunct Twine, built by Nova Spivack. The particulars are very different however. It also bears some striking similarities to Flipboard, although there’s no swiping as this is a web app….

I gave it a few tricky concepts to warm up on. My first test was to get a stream of content about Vikings (the Scandinavian kind, not the football team) – no surprise, given the title of this blog. Since there are two very well defined concepts there, it’s a chance to see how well the learning engine can hone in on the concept, and not the word itself. I got a quick stream of viking stuff, half of which was about the Scandinavians and half the football team. I thumbed down about 4 articles on the Minnesota Vikings, and presto, my stream has become empty of football. Pretty impressive! The main problem now is the stream is mostly full of one news story, about climate change impacting Greenland, with many different sources recapping the same article. So while the concept disambiguation is working, they need a “story disambiguation” similar to what Google News has. I also found it awkward to get more results into the trap – there is an “infinite scroll” option, where as you scroll you get more articles, but my Viking trap could never get more than about 16 or so articles. Not sure whether this is a UI limitation, or whether their base of indexed documents simply doesn’t have more. They also need some way to delete an article from the feed, or mark it as read – the UI suggests that this is a kind of very smart RSS reader, but the basic mechanics of RSS readers are missing….once I’ve read something, I want to get it out of the way.

I tried a couple of other Traps. I’ve been on a binge of listening to “dubstep” recently, which is a trendy new sub-genre of electronic music. I got a great stream of interesting articles, and even found a new musician or two to listen to. The articles were good, but sourced from fairly random-seeming sites – Trap.it is not very clear about how they curate their content sources…..I created a historical fiction trap, and immediately got an interview with Steven Pressfield, one of my favorite historical fiction authors, who has a new book out. Pretty impressive, although likely somewhat accidental? But very cool stuff. This is something to keep an eye on.

(ps – they are pretty responsive to user comments on Twitter @gotrapit – they got back to me almost immediately when I tweeted)

(they may still be in closed beta – GigaOm seems to have some invite codes at the bottom of this article)

I wrote a guest post for BostInnovation on the future of local search. You can read it here: http://bostinnovation.com/2011/03/29/the-future-of-local-search-utility-and-how-todays-services-fall-short/

There’s been a blizzard of social media chatter about Quora recently. Robert Scoble loved it, then hated it. Vivek Whadwa doesn’t get it. Arrington thinks Scoble’s nuts. All of this is focused on whether or not Quora is a great blogging platform, how the voting system works, and so on.

Here’s the real problem with Quora, and why it’s going to struggle, until it solves it.

There’s no information architecture. It’s just a huge pile of unorganized content. The value in a Q&A service is the ability for people to process the answers and find useful information, or at least predict where they’re going to find this information. Hunch suffers from a similar problem – it’s hard to know which ‘Hunch’ is going to answer your question. Until Quora finds a way to put some normalization onto the questions, the pile is just going to get bigger and harder to sift through. And this normalization is a hard problem, particularly for subjective questions, which is what currently predominates in Quora. It won’t be an easy thing for them to fix. Wikipedia (to which many comparisons are made relative to Quora) has a pretty clear information architecture, and a dis-ambiguation mechanism, and it’s data model elements are primarily (simple) nouns (Places, People, Topics) – whereas Quora’s primary content are usually (complex) subjectively phrased questions.
Continue reading

Would somebody please build “menupages for winelists”, on a mobile device. I’d like to be able to walk into a restaurant I’ve never been to, with a wine list as long as my arm, and have the app be able to recommend wines off the wine list. Cross-reference Wine Spectator scores and prices, so I can ask: “what’s the highest rated Merlot”? “Which Italian Red has the best value (price per wine spectator point)”?. Or even just see what the street price of the wine is, and it’s rating. Anything to help me make sense out of that monster list of choices. I have some wine apps I love – my friend Brad Rosen’s Drync for example – but none that I know of has access to the wine list in a venue, the way menupages has access to, and provides searching on, the menus in a restaurant.

Wow. Things are really getting interesting. I’ve been talking for a while now about the convergence of search & travel, search & local, search & mobile/location based services. This is just another proof point (a $700M proof point). It doesn’t really affect Goby directly – we really haven’t been playing in the airfare space, we decided a long time ago to leave that to people who already do it well. The biggest potential losers in this deal are the metasearch players and online travel agencies – Kayak, Expedia, Travelocity and Orbitz, because Google can now build a user experience for flight search that dis-intermediates those companies and connects users directly to airlines for purchasing.

There’s also an interesting tension with many travel advertisers – Google receives billions of dollars in travel advertising revenue, but increasingly travel companies are viewing Google as competition (through things like Google Places pages and Google City Tours), and might take their ad $ elsewhere. The other interesting potential loser is Microsoft, for two reasons. First, Bing Travel is a great differentiator for Bing, and now Google has said they will build a competitor, where previously they had none. Second, Bing Travel (an enhanced and re-branded version of Farecast), was powered by ITA – it will be interesting to see how long that lasts!

Interesting times indeed. As airfares become more and more of a pure commodity, at Goby we’re thinking the action in the industry will turn to finding experiences – the real reason people travel. And Goby is perfectly positioned for that.

(btw – in addition to the obvious, Google is a winner in that they get a huge pool of deep engineering talent, and vertical expertise in the travel industry. They also pick up an interesting and not well publicized asset of ITA, the Needle project).

Microsoft is apparently listening to my blog – they’ve gone and launched the search engine I said Apple should build a few months ago – a media and entertainment search engine! (Or is this their version of me saying “I’m a PC and Windows 7 was my idea?” 8).

Bing Entertainment (http://www.bing.com/entertainment) covers Music, Movies, TV, Games, and Video Games. It has a very attractive, browsable interface with a lot of rich media – for example, the music page shows photos, links to audio clips of the artist, as well as upcoming live performances. In spirit, I love what they are trying to do. They’ve even stolen a page from the wonderful (but defunct) service Lala which Apple acquired. You can play over 5 million songs in their entirety once for free. So if you ever wondered if you’d like Miles Davis’ Flamenco Sketches, now you can find out, for free, without engaging in any piracy: http://www.bing.com/music/songs/search?q=miles+davis&go=&form=DTPMUS.

Or at least, in principle you can. If you click on the play links, you get a “coming soon” message. Seriously, Microsoft? What was the hurry to launch, that you called this out in your press release (http://www.bing.com/community/blogs/search/archive/2010/06/23/a-new-entertainment-experience-for-bing.aspx) as one of the main features, and it doesn’t work? In practice, I find the implementation of Bing Entertainment to be disappointing (in contrast, Bing Travel is done very well). So far as I can tell there is no integration whatsoever between the search bar and the browse panels that are right next to each other. For example, a search for “Lady Gaga” produces the same search results as if I’d gone to Bing proper. It’s positioned as a media search product, but in reality it appears to be an entertainment portal with hard-coded content (and not very deep content), with a disconnected search box right next to it. The catalog appears to be very shallow. A search for the “Archer” TV show was disappointing – effectively the same search as on Bing “proper”. (BTW Archer is an absolutely hysterical parody of James Bond meets Beavis & Butthead meets “The office” – check it out if you haven’t seen it). In the entertainment space, there is a very limited catalog of entities that need to be recognized – how many TV shows are there, after all? Those should be recognized as special searches and treated accordingly. My search for “Archer” should have taken me to an Archer landing page. While it is positioned as a media search product, the search angle is virtually non-existent. The positioning of the product is spot-on, but the execution is lacking. Apple has a huge opportunity here – if Apple makes this product, the user experience is going to sing (perhaps even literally!).

Had a great (and very informative!) visit with Robert Scoble, his coverage is here. He said some great things about us (“more important than foursquare”!), but one of the most rewarding things was that his wife loves Goby as well. It’s one thing to impress someone technical with what you’ve achieved – it’s even more rewarding when people who aren’t in it for the technology see the usefulness of something in their daily life.

Here’s the interview:

I’ve posted on a number of occasions about task-centric search and how it’s the future of information access. I’ve also speculated on what kind of search engine Apple should (and might be) building. Some recent comments from Steve Jobs have strengthened one of my theories and thrown cold water on the other.

First, the cold water. According to Jobs, Apple isn’t building a search engine. Some other folks suggest he’s not being entirely forthcoming, but I suspect he’s telling the truth in a literal sense. In the figurative sense though, Apple has a different vision for how people consume information – it may not be “search”, but it is information finding:

On the desktop search is where it’s at; that’s where the money is. But on a mobile device search hasn’t happened. Search is not where it’s at, people are not searching on a mobile device like they do on the desktop. (more here)

Put one way, “there’s an app for that”(TM). In terms I’ve used before, context is the organizing principle for information access, and what is a mobile app if not “context” embodied? Context can be summarized with a mnemonic: TILT. Task, Identity, Location and Time. If I know your TILT, I know your context. An iPhone app has a pretty good implicit idea about all of those things. Who you are, what you’re doing, and where & when you are. That’s why mobile apps like Yelp or Goby can so effectively answer information needs with just a few taps. John Batelle has expressed some similar ideas here. This is why people use apps on the iPhone rather than search, as Jobs suggests. That’s why Goby manifests very differently on the iPhone than in does on the web, even though the underlying information model is the same.

The broader point: in the future, information finding will be supported by task-centric, contextual search applications, not general purpose “search box + 10 blue links” search engines. They’ll manifest differently on the web than in mobile environments, but share the underlying premise.

Does Blog Search have a future?

Blogs are one of the richest sources of information for certain classes of information. Yet they are frustratingly hard to find or extract information from, and the state of the art (Google, Technorati) feel 100% stagnant. Here are a couple of example use cases I have that aren’t well served by existing tools:

1. Recruiting. When I recruit for a particular role, I’m looking for thought leaders or people with insight and passion. Usually these people have blogs. If I could see, for example, a list of all the people in the Boston area with blogs that blog about web development, I’d probably find some rock star developers. No easy way to do this today.
2. Travel planning. I’m thinking about a stay in southern Utah at a Bed & Breakfast. Who’s blogged about their trips there that might have some good perspective for me?
3. Music Discovery. Great blogs like Aurgasm, Quietcolor or TheMusicSlut are great ways to find music. But how many others like that are out there?

The current serious choices are pretty much limited to using normal search (Google, Bing etc), or using a Blog search engine like Google Blog Search or Technorati. With Google Blog Search, you get pretty much a toned-down version of Google: a search box with 10 results – you can’t really search for *blogs*, you can only search for *posts*, with the relevance ranking determined by some version of PageRank. There’s no real sense of the authority of a blog (other than that of PageRank), and no real opportunity for discovery – just punch in your keywords and hope for the best.

With Technorati, you do get some increased power. You can search for blogs as entities distinct from an individual post, and blogs do get assigned an authority score. But the experience seems to fail as often as succeed. A search for “boston web developer” blogs on Technorati returns three blogs, all with an authority score of 1 ( the minimum) – pretty sure there are more than 3 of this kind of blog in Boston! And there’s no way to sort the blogs by their authority score, at least that I can see. The Technorati blog directory also seems to be mostly limited to “authoritative” blogs – personal blogs (for example my own) seem to have little or no representation. But on long tail topics (say, music reviews of obscure artists), blogs by “real people” are often the only place to find this kind of commentary. Most importantly, there seems to be little innovation happening in Google Blog Search, Technorati, or more generally – the field is stagnant.

What would the characteristics of a good blog search tool be? What’s lacking in today’s approaches?
1. Comprehensiveness. There’s hundreds of millions of blogs (Billions?) – yet Technorati doesn’t seem to find many of them. (Google is more comprehensive, but limited by the “search box + 10 results” interface).
2. Ranking of blogs relative to search query and/or authority of author. (Of course this ranking problem is non-trivial. There are some interesting ideas on authority for twitter accounts which could perhaps be leveraged, e.g. http://thenoisychannel.com/2009/01/13/a-twitter-analog-to-pagerank/)
3. A faceted, searchable directory of blogs supporting discovery. Categorization technology has come a long way. It ought to be possible to categorize every blog against a reasonably detained taxonomy or facet set (say, the Open Directory categories, or something better), with 80% accuracy, across some common facets: topics, locations, age of blog, date of last post, and so forth. Even at 80% accuracy, this asset would be quite effective. And with a little UGC thrown in, the 20% that matter, and are wrong, will quickly get corrected. Using a microformat like hCard, blog authors could document their blog’s metadata quickly and accurately. Blogs also typically have some consistent thematic elements, such as an “About” page or a Blogroll list, that could be mined for interesting metadata. This kind of experience would power a new way to discover fresh and interesting blogs & content.
4. Recommend other, similar blogs. Powered by the facets above, or by a blogroll analysis, or something similar, a recommendation feature for similar blogs could be implemented, based on information readily available in an “almost standard” format.

Verticalized Blog Search Engines might also provide some task-centric capabilities. As I’ve written before, the future of search is about providing task-centric search capabilities. In music for example, The Hype Machine has some very interesting behaviors it can support, simply by virtue of being focused on music.

The obvious question: what business or investment model would support this kind of vertical search engine? In the Goby world of travel and entertainment, there’s a long history of various ways to monetize that kind of content. In the “pure content” world of blog search, it’s less clear – a pure page-view based CPM ad model isn’t likely to work. If the New York Times can’t make that kind of model work, a startup probably can’t either. Perhaps some form of interest-based, downstream ad retargeting approach might get enough leverage that it could get to critical mass. Alternatively in some domains a “freemium” model might work, where additional tools (say for recruiters or brand managers looking for a competitive edge). Given the scale of the problem, it’s not clear a bootstrapped company could take this on – the infrastructure requirements (bandwidth for crawling, servers, etc) probably require a non-trivial level of investment.

What blog search tool do you use? Do you use a blog search tool? or just Google? Is anyone innovating in the area?

There’s a convergence coming, between the worlds of search, travel, local, and social. It used to be that if you were traveling, you used a guidebook and map and talked to the concierge, then you graduated to TripAdvisor and Expedia (and if you were adventurous, Kayak). People’s use of search engines tended not to intersect with their travel planning. In recent years of course Google has become a de facto part of the travel planning experience – although by no means a perfect one. And some search engines have introduced travel products (notably Bing Travel). And for planning your weekend, search engines have historically not been of much use at all – they don’t understand the concept of time or location very well (“this weekend” is just a few keywords to them), and don’t understand your task (when I search for beaches on Cape Cod, why do I get back results for restaurants with the word “beach” in them?). Robert Scoble has some thoughts on this subject, here. Google appears to be moving in this direction, with their rumored acquisition of ITA, which powers many airfare metasearch sites including Kayak. Their abortive attempt to acquire Yelp shows how search & local are converging as well.

But there’s a new game in town – social/local gaming, in particular with things like Foursquare and Gowalla, that combine social gaming with local-search-like results, allowing people to broadcast where they are and what they’re doing. There’s an evolving “stack” of technologies, including location databases and engagement tools, nicely summarized by Chris Dixon. (I disagree with his assertion that location databases will become commoditized – the information is too hard to come by, and companies like InfoUSA make hundreds of millions in revenue providing this kind of data. Not to mention the startups like SimpleGeo and Locationary and for that matter Goby, that are tackling the problem, but I digress).

This kind of engagement is going to have a profound impact on how people plan travel and figure out their weekends. DeepDish Creative (http://deepdishcreative.com/wordpress/2010/02/foursquare-for-tourism/) is talking about how destination marketing organizations can leverage these tools to promote their destination. But I see two problems with this generation of tools as they apply to this problem:

1. They are after-the-fact. I tend to engage with Foursquare after I’m already AT someplace – Foursquare isn’t really involved in my decision process, it simply records what I’ve already decided. As a result, it has limited use (not no use, just limited use) in making decisions.
2. These tools only recognize a limited set of entities, primarily businesses (in fact, primarily restaurants). It’s hard to check in at a U2 concert, because it’s an event, and it’s hard to check in at the Grand Canyon, because it’s not really an entity, it’s a generalized (and off-the-beaten-track) place. God help you if you want to check-in on a hiking trail!

Addressing those last two elements would create a resource that will not only appeal to my vanity & let me broadcast what I’m doing, but more importantly help me decide.

The key need here is a semantically meaningful database of things, to key all your features off of, and search tool to find & organize them – not just a pile of URLs. The system needs to know that Yo La Tengo is a band playing at the Fillmore on the 23rd of April, with a date and a location – not just a pile of keywords without any meaning. Any system like this needs to cover hotels and restaurants as well as non-business entities like hiking trails or concerts, and once you leave hotels/restaurants, this information is hard to come by. Once you have the database of entities, it is straightforward to build a platform for people to engage with their networks, in the context of that content. Once you have a strongly categorized, rich database of things to do, and a strong network of people telling you what they are interested in, you can provide compelling recommendations as well as support discovery. And, strangely enough 8), that’s where we’re headed with Goby – we plan to be right at the intersection of this convergence.