google search results (Photo credit: Sean MacEntee)
You say “Semoogle”, I say “Goomantics”. Two made up words; one meaning. Map the terms to one another, and associations to one can be related to the other. Do that within the house that Google built, and you can really traverse the knowledge graph (that was MetaWeb’s Freebase).
Keyword matching is just part of what happens inside the Google machine – and more and more, sense is discerned from context – in aligning content (search results or ads) with the searcher’s intent (their meaning, in terms of identifiable entities and relationships).
Our latest Semantic-Link discussion was interesting in that it touched on some distinct but deep topics that tend to recur in our discussions, namely: usability, privacy and the old standby – the definition of semantics itself.
I won’t spend any more time on the definition of semantics beyond that the consensus (for purposes of this discussion) was that it means “meaning”, with contexts including: linguistic/NLP related word-meaning semantics; and the other being compliance with W3C standards – or architectural Semantics. In essence, the latter is what enables a machine version of the former.
The focus was actually a conversation with guest Nova Spivack, and his more current efforts, including Bottlenose and StreamGlider. (Next time we’ll have to let Nova do more of the talking, as we only really had time to dig into the first of those.) Bottlenose is intended to help people manage and interconnect their interaction across the multiple electronic realms in which they operate. While Nova mentions that the system doesn’t currently make use of W3C standard architectural Semantics, it does use ontologies to relate topics and navigate meaning. This is particularly visible in Bottlenose’s Sonar – which renders a visualization of the active topics, hash-tags, and people around you, with adjustable time-horizon. If you’d like to try it out during the private beta, visit Bottlenose.com and you can Sign Up using the Invite Code: semanticlink.
As mentioned above, two key items arose from the discussion – the matters of privacy, and the question of transparency. In the case of privacy, would it become an issue, from a business intelligence standpoint, that others could more easily see the topics that someone is discussing or investigating – especially if such a tool could cross multiple networks/platforms in finding patterns.
As is often the case in these Semantic-Link discussions, the question of “how much should be exposed about the use of semantics” arose. There is of course a balance between active vs viral evangelizing of semantics, and the cost of exposure is simplicity and usability, while the benefit is flexibility and control, for those who can handle it.
The answer itself is complicated. On the one hand, technologies need to evolve in terms of leveraging semantics in order for people to really benefit from the underlying semantic capabilities. At the same time, those same people we’re talking about getting the benefit shouldn’t have to understand the semantics that enable the experience. Paul Miller, host of the podcast, also wrote about this issue. I’ll add that Investors do to like to hear that their company is using unique and valuable techniques. So too, though, is it the case that any company making use of semantics likely feels it is a competitive advantage to them – a disincentive to sharing details of the secret sauce. .
As mentioned during the podcast, this is a matter of which audience is being addressed – the developers or the masses. And in terms of the masses, even that audience is split (as is the case with almost all other software users). There are the casual users, and there are those who are hardcore – and when we’re talking about masses, there are many many more people would fall into the casual camp. So from a design standpoint, this is where usability really matters, and that means simplicity.
So in the case of Bottlenose, for the time being they’ve chosen to hide the details of the semantics, and simplify the user experience – which will hopefully facilitate broader adoption. There may too be room for a power-user mode, to exposes the inner workings of the black-box algorithms that find and weigh associations between people, places, things… and let users tweak those settings beyond the time-frame and focus adjustments that are currently provided.
Mentioned by Nova was the LockerProject in which personal data could potentially be maintained outside any one particular network or platform. This of course helps on the privacy side, but adds a layer of complexity (until someone else comes along and facilitates easy integration – which will no doubt chip some of the privacy value).
Personally, I’d love to see the ability to combine slices of personal activity from one or multiple platforms, with tools such as Bottlenose, so that I could analyze activity around slivers or Circles (in the case of Google+ usage) from various networks, in any analytical platform I choose.
In the same vein as Word Lens, which I wrote about here just over a year ago, Aurasma too looks through your lens and “augments reality”. What does that mean though? And why is it interesting? At the most basic end of augmented reality, think of those times in touristy areas where you’ve had someone take a picture of you sticking your face through a board, on the front side of which – surrounding the hole you’re looking through – is painted some well-built body that surely isn’t mistakable as yours.
Image via Wikipedia
Add some basic technology, and you have photo doctoring capability that puts a border (or mustache) on your photo, or converts it to a sepia or negative view. Geo-code and/or date-stamp the image file, and integrate with information on buildings, locations, people and/or events that occurred there, and you can display that information along with the image when the coordinates correspond, a la Wikitude. Load up that app, turn it on, and walk around pointing your phone at things, and see what it says about your surroundings. (MagicPlan is an iPhone App, from Sensopia, that is a practical application of related technology, enabling CAD for making floorplans!)
Aurasma adds to this, by integrating image recognition (think: word recognition, but visually, picking up defined items) and rendering associated audio, video, animation, what have you – much like scanning a QR code would launch an associated action – but in this case, like WordLens, will do it in place on the image. Take a look:
The reality is that behind the scenes, with text, image or voice recognition, any action could be defined to be launched upon encountering triggers. Going further, imagine using multiple criteria or triggers to launched actions – tweaking the criteria for different scenarios. For example, a coffee company logo could spawn a video themed “start your morning with a cup” if the logo is seen early in the day, a “get a mid-day boost” if it is in the afternoon, or “keep your mind sharp tonight” if it is in the evening (adding “to get your studying done” if the geocode also indicates that the location is on a college campus. The mantra of late has been “context is king”. That’s context.
What company, technology or issue caught your attention in 2011
Are we “there” yet?
What are people watching for 2012
Notable attention grabbers were: schema.org and its impact on who pays attention (i.e. SEO space); linked data (and open data); increase in policy maker awareness of the need to pay attention to interoperability issues; commercial integration of technology (ontologies plus nlp capabilities) to leverage unstructured content; and of course Siri (a key example of such integration…).
In terms of where we are in the progression of the semantic technology realm, the general sentiment was that Siri represents the beginning of inserting UI in the process of leveraging semantics, by making the back end effort invisible to the user. And looking forward, the feeling seems to be that we’ll see even more improved UI, stronger abilities in analysis and use of unstructured content, greater integration and interoperability, and data-driven user navigation, and Siri clones.
During the recording of the December podcast of the Semantic-Link (as of this writing, soon to be posted), I emphasized the general need for enablement of the general public to begin contributing and consuming linked data – without having to have much, if any, technical wherewithal. The real explosion of the Web itself came as a result of wysiwyg authoring and facilitation of posting content and comments by just about anyone with a web connection. Similarly, de-tech-ification of where the web is going from here is what will pave the way to getting there.
There are standards and tools now for the related underlying componentry, and what is needed is user-interface development that will usher in the explosion of linked-content generation and consumption (as web2.0 did before).
Toward this end, Andreas Blumauer writes about a new version of PoolParty’s WordPress plugin that extends an in-page Apture-like approach, to use and contribute to the LD ecosystem. This (coupled with other elements such as SKOSsy) is an example of the type of UI gateway that is needed in order to enable the general public to participate – with systems that generate and digest the linked-data-age information currency.
If this isn’t one of the coolest things you’ve ever seen…
You probably thought it was Jetson’s material that someone could speak one language into a phone, and you could hear it in a different language on the other end. Pretty great stuff, translation on the fly. Think about looking at something that is written in a different language, and being able to able to see it in another, without having to go look it up somewhere!
Del.icio.us has been a reliable web-based bookmarking resource that has not only enabled bookmarking in the cloud, so bookmarks could be accessed from any computer you happen to be using, and sharing them with others. It facilitated multi-tag classification of them, so that you could zero in on what you’re after by triangulating with related words in your own naming convention, and breaking free of traditional, hierarchical folder storage structure.
It has been a great resource for researching the language that others use to describe the topics and pages you are interested in, and has allowed, if not encouraged, the development of worldviews and, though some scoff at the word, folksonomies.
While this news is a shame, the truth is that the resource has not been leveraged, and some say it has been neglected, since it was acquired in 2005, which happens to be when I began making use, about 5,000 bookmarks ago. I recently signed up for Pinboard (which has a one time cost of about $7 right now, and offers an auto logging archive for an annual subscription). The thought of paying for something you’ve done for free might bother some, but Barry Graubart makes an interesting point in his post on this subject when he says “remember that the lack of a business model is what required Delicious to sell to Yahoo, who neglected it.”
I’ve long been compiling material for my (in progress) book entitled “Long Term Short Sightedness”. I’ll have to be sure to save some room to write about this decision by Yahoo!
In the context of marketing and advertising, we’ve heard more during the last year or so, in reference to the semantic web and semantic technology. What does Semantic Advertising really mean? One interpretation – the one we’re not talking about here – is the selling of something by calling it semantic, which some have done in order to ride momentum (which I call “meme-entum”) of the space to sell something based on a loose association with the concept of “meaning” or “intent”. So what are we talking about?
The strategy in the space has long been driven by word association, and more and more-so on an automated basis. At a time, placement was done on an entirely manual basis – and automation of keyword matching increasingly became the basis for new business models. That is, after all, the basis of what we now think of as contextual advertising – the alignment of what the user is looking for with the other things they encounter on the page.
So to put it simply: What is it that is new and different? What is it about the inner workings of an advertising mechanism that makes an offering semantic or not. What are the drivers and opportunities around these differences? What is real? These are some of the things we’re looking to learn about in detail at the panel discussion that I’ve been helping to organize for Internet Week in New York – the title of which is Semantic Advertising.We’ll leave it to our moderators to dig into the nuts and bolts of the subject with the experts that have been gathered. Going into the discussion though, here are some of the questions I’m thinking about:
Since keyword matching is, well, keyword matching: what are the main differences between straight-up contextual advertising that uses keyword lookups relative to its semantic brethren?
Does the addition of keyword frequency, and therefore the statistical analysis of the text, make the matching on a ranking basis qualify as semantic?
Going beyond simply enhancing alignment, predicated upon statistical assumptions, is it the further use of NLP to not just extract concepts to be matched, but to determine the intent by the terms used – to better tune matches when words have multiple potential meanings? Many of us have encountered the unintentionally matched ads – which can be disastrous for a brand. What can really be done now, and how?
Further on the NLP side, there is the potential for sentiment detection – so even when the correct meaning of a term is understood, determining whether its use is appropriate for matching would be based on the positive or negative connotation of its use (think here in terms of whether you would want your airline advertised next to a story about an aviation mishap, for example).
Going at the question from the “semantic-web” side, is embedding (and detection of) metadata on the page just a different flavor of Semantic Advertising – or should we be calling that Semantic Web Advertising instead? This seems less prone to interpretation errors, but the approach relies upon metadata which is largely not yet there. (Because of the markup related aspects of this point, I wanted to call this post “Mark(up)eting and (RDF)ertising”, but was talked out of doing so).
Is there a difference in strategy and/or scalability when considering whether a semantic approach is more viable when done within the search process, as opposed to on the content of the page being viewed?
If ads to be served are stored in semantically compliant architecture, does that itself provide any advantages for the service provider? And would doing so give rise to the service being referred to as Semantic Advertising? Does this even enter into the eqaution at this point?
Would increases in the amount of embedded metadata shift the balance of systematically enhanced ad selection and presentation of sponsored content – from one web-interaction phase to another?
I’m looking forward to the panel – to open my mind regarding these and other factors that come into play – and what elements and trends will be necessary for the viability of the various possible directions here.
Image by buschap via FlickrEvery now and again, you find yourself doing something that, when considered in the context of other parts of your life, can reveal interesting things about yourself or people in general. I drew one such parallel after spending a weekend with my family hurtling through the air at Hershey Park.
For those of you who have never been there, it has 11 roller coasters – and if you’re not into them, well, you’re going to be standing around a lot waiting for your kids to come back saying “that was great”, “let’s do it again”, or something about a snack. I’ve done lots of things that some would consider daring – including rock and ice climbing, and flying in a glider, sea kayaking, whitewater rafting… but I’ve just never been one for rides that are supposed to give you near-death experiences, because frankly, I don’t generally like feeling that close to “the edge”.
Since my wife and kids were so gleeful following each of their wild rides – especially Farenheight (pictured above) and Storm Runner, with their loops, corkscrews, and in the case of the latter, acceleration from 0 to 70 in 2 seconds – I was compelled to understand their sense of the differences from one ride to another – and how each impacted their senses.
In the back of my mind, I think I was secretly trying to ascertain which of the elements I could personally have handled – given that each person has a different sense of what “scary” is, and you cannot just rely on someone saying “it wasn’t so bad” or “you could handle that”. Key factors which weigh differently by individual include: height, speed, drops (number, length, steepness), banks, roughness, re-direction and mis-direction… (among surely many more categorizations of the coaster-phile).
On the “handle-it” scale, I had already managed Lightning Racer, Wildcat, and Comet, and had dipped my toe into the “beyond” with SuperDooperLooper (and a few others) – and their death defying drops and loops. To “push my research”, I finally succumbed to the pleas of my family, and (somehow) joined them on Great Bear, which added “hanging from the rail”, and “corkscrews” to my repertoire.
At the end of Great Bear though – which, by the way, I survived – I realized that I had only managed to do so by staring at a bolt – which attached the seat in front of me to the chassis holding the seats on the rails, from which we were hanging.below – for the entire ride. A friend commented days later that the only way to do a ride like that is to give yourself over to the ride. I fear (among obvious other things) that I did not do this. While I can say to my 10-year-old, who finally stopped calling me a sissy, that “I did it”, I cannot say that I experienced all it had to offer – but I survived.
The parallel I alluded to above is not related to the two tracks on each coaster – but to life in general – work, activities, people, relationships… If you give yourself over to these experiences, you can get out of them all they have to offer, and help others do the same. This too reminds me of the same parallel that went on in the back of my mind following a whitewater kayaking trip in the winter of 1978-79, during which a professor said to me “we need to make sure we’re moving faster than the water if we have any intentions about steering”.
If you hang on, fixing your gaze on a bolt to simply survive, you’re going to get tossed around and may not enjoy the “ride”. In the case of the Great Bear, the truth is that the ride may never get another chance to show me its stuff.