TV ontologies, how much do they matter?

An interesting and provocative (presumably intentionally ;-) ) post. I’d argue that quite simply, yes, they do matter. Because we’re not really talking about ontologies for TV at all – the examples given above are not really to do with television, the medium, but with the content which is distributed through it. Ontologies to describe the narrative content of TV shows is vital if we’re ever to realise the potential of the Web, not just in terms of desktop or mobile based interaction, but as an information space, which can be accessed no matter what device you’re using.

I’d also argue that although we may appear to be passively consuming TV content, that’s just the mode of television. That is to say, although we don’t actively seem to interact, if you’re engaged in a narrative, and indeed, if you’re having a conversation with friends about the content, then you are exploring, investigating, questioning the content itself.

People don’t watch TV to watch TV. They watch it to be engaged in a narrative, whether that’s news, sport, documentaries or fiction, and once you’re engaged in that narrative, you’re exploring in the same way as you burrow through Wikipedia.

It’s true that viewers don’t want to ‘work’, but that’s more to do with the clunky user interfaces that are forced upon us – the ways we can only follow set ‘user journeys’, when all we want to do is find out own way…

Thanks for the comment Paul, really very insightful.

You said “Ontologies to describe the narrative content of TV shows is vital if we’re ever to realise the potential of the Web” – Imagine you’re sat with your son watching a wildlife show and a turtle pops up. It would be awesome to know a little bit more about turtles if he piped up and asked e.g. who their closest relatives are, why they’re green and all the other amazing stuff kids bug us with. When I think about this and similar situations, I can’t help thinking the following problems with this scenario:

1. With broadcast TV, you’re mostly enjoying a linear narrative, it’s going to be hard to deviate from that without missing something or taking up too much time.
2. Someone somewhere (presumably in a production company) would need to associate the object Turtle with some kind of ‘link’ to other metadata. It’s hard enough right now getting production companies to provide extra stuff around shows, never mind creating valid semantic metadata.
3. For your average wildlife show pulling in possibly 500k people, there’s probably about 1 or 2 people that want to know anything more about turtles. Demand just seems too low and as TV viewing could fragment further, demand might decrease further still.

I’m kind of fascinated by semantic broadcasting, but it really feels too much like hard work for everyone involved, with not enough benefit. Especially when I can go online and type ‘turtles’ into Google.

Really interesting post Mike – some thoughts on your points.

1). Broadcasters are already spending a great deal of thought on how companion devices (laptops, pads, phones) could enhance/support viewing experiences ( http://bit.ly/ciGIZ2 see • What are you working on?) So, as in your example, you wouldn’t need to type “turtles” into Google.

2). Metadata overhead would be large. But the BBC have already (successfully) trailed speech-to-text within Democracy Live ( http://bbc.in/aM01jF ) perhaps the next step is to attribute ontologies/semantics without production/editorial having to do all the heavy lifting of mining metadata out of the broadcast.

3). 1+2 = a sound reason to do #3
If there’s a low friction way to attribute metadata to content, then we should definitely do it. As Paul mentioned, traditional broadcast + the potential of the web + understanding appropriate user needs, in my opinion, is where the sweet spot lies.

Agreed, I probably wouldn’t want my TV viewing littered with uber-red button experiences, but played out synchronously to an appropriate device or contextual to consuming in an alternative way (VOD) – seems like it could be a valuable and exciting experience.

Hi Dik, thanks for the comment.

The second screen stuff sounds awesome, and to a certain extent, that’s what I’m questioning. It sounds awesome to me as a geeky lad obsessed with the things people do with glowing rectangles. But is anyone on the receiving end getting into this? Is there any research that suggests that anyone that doesn’t work in ‘media’ is interested in second screen? Of course, I understand that that’s part of the reason why Vicky’s developing this stuff.

It strikes me that this is fascinated to people that work in digital but when you talk to people that simply enjoy vegging out infront of the telly they are usually confused that you’d mess with something they consider a perfect medium.

Clearly there are some very interesting behaviours being exhibited by viewers with big appointment viewing shows like xfactor and the apprentice. But these things are already happening. And to a certain extent, this is behaviour that is happening without Twitter and Facebook and has been happening over the phone and in the living room since the 60s.

I guess I’m just trying to focus on stuff TV viewers might get excited about rather than stuff designers, creatives and planners get excited about.

A lot of this point of view has really come from some testing we’ve done recently where people were totally confused with a prototype we had built. It was a very typical second screen typw of thing. They flat-out rejected it, while we thought it was a winner from the start. Perhaps it was the way it was executed or described it but it’s put a lot of this stuff in a different light for me.

I’d love to hear if the BBC has any insight on this stuff?

“I guess I’m just trying to focus on stuff TV viewers might get excited about rather than stuff designers, creatives and planners get excited about. "

I completely agree. I think that second screen stuff could be interesting, but I also think that too often we’re falling into the same assumptions about the content of non-TV/radio/print media. As per your example, the starting point always seems to be ‘to know more about X’, where X is an isolated thing in the TV content. I guess I’m thinking that’s what Richard means when talking about supporting/enhancing a TV show.

What I’m interested in is taking the exact same content (by which I mean the narrative being conveyed), and adapting that to the form of the medium in question. In the case of the Web, the form of the medium is URIs and hyperlinks (and by extension, ontologies) – so, can we try to represent the narrative of the story through these structures?

Once we’ve worked out how to do that, then we can really start to build interesting experiences not only on desktops/handhelds, but also work with the interplay between different media (e.g. via IPTV..)

Ideally, this data should be captured during the production process, because it’s almost like a parallel process – just as a writer writes a script, could they write (and/or pull together) the URIs and links to represent the story?

It’s a very intriguing subject, though ;-)

(From the BBC R&D Prototyping team)
We’ve just run a live second screen trial with a selected group of Autumnwatch viewers (almost exactly what you described in your scenario) – http://www.bbc.co.uk/blogs/researchanddevelopment/2010/11/the-autumnwatch-tv-companion-e.shtml

We’ll write more when we’ve analysed the data and (hopefully) have some interesting insights.

Thanks Paul, I think we’re on the same wavelength.

Tristan, it would be awesome to see your findings.

Have you guys seen Picklive?

http://mintdigital.com/portfolio/football3s

How about that for real-time narrative? I’m still waiting to see how popular it becomes though.

Hi Mike

Nice post. What follows is mostly opinion…

Firstly to say I agree with almost everyone here; the 10’ (or 3’ if you’re lying on the rug) experience doesn’t lend itself to interaction beyond run, jump and shoot. Which I guess makes this post and accompanying comments more about the “2nd screen experience” than the 1st. Or I think we’re agreed that the future feels more like connected devices than connected device no matter what the IPTV people might choose to say. In answer to your question: “Now, on TV (the 10’ experience) do these ontologies matter?” Probably not at least in consumption mode…

The next question becomes: are “real people” interested in the 2nd screen experience. I know my friends are but then again I know my friends are pretty much wedded to their laptops so that doesn’t tell us much. As you point out big “appointment” viewing shows generate lots of 2nd screen activity, mainly, in my experience, via Twitter. But there’s some evidence that even outside “real time TV” people are watching the 10’ whilst surfing the 2’. Just check Google analytics for Pompeii searches during the 1st tx of http://www.bbc.co.uk/programmes/b009wzbf

Whether this means people are so stimulated by the programme they can’t help but search for more information or whether the programme bores them so deeply they can’t help but simultaneously use their laptops is probably best left unexplored…

But either way it does seem that people are watching and seeking additional information simultaneously. Tho it’s hard to tell the proportion of geeks and hackers to punters

Most 2nd screen apps (at least the ones I’ve seen) seem to take this “more information” model and run with it turning the pull of a Google search into an information push. Personally I’d trust Google to give me a better grasp of “more information” than I would most programme makers. And pushing information at me only really saves me 10 characters and a return key…

So maybe (if we know people are using laptops / phones / whatever) during viewing the problem with 2nd screen is more about what most broadcasters choose to do with it. Sometimes push information might be a reasonable response (as in your wildlife documentary idea), sometimes the programme might demand a different treatment. Personally I think push information would be perfect for something like a food programme (give me the recipe they’re cooking now) but I’d still prefer something Cluedo / game like and social for Midsomer Murders

So IF people do want 2nd screen (because your programmes bore them or overstimulate them) AND you can design a 2nd screen experience that lends itself to the programme format the next question is how you go about building that. As you point out in your first reply to Paul the economics of hand rolling a 2nd screen experience on a per episode / series basis don’t stack up. But that (in my opinion) is where the ontologies start to count. Because no programme is a snow flake. This whodunnit is fairly similar to that whodunnit, this food programme fairly similar to that food programme, this reality tv show….

The obvious problem (as you point out) is that ontology development is fairly cheap compared to ontology population. But that I think (and I would say this) is where Linked Data can help TV. As Paul pointed out in his first comment “we’re not really talking about ontologies for TV at all”. We’re talking about ontologies for real life which TV just happens to cover. Many of the concepts involved in this coverage are already described on the web. So you don’t need to describe a turtle; just link from the episode (or segment of an episode) to a description of a turtle elsewhere (DBpedia, Wildlife Finder etc). Compared to the cost of developing a TV programme constructing links to describe it is relatively cheap.

That said I’d probably disagree with Richard on this. Getting from speech-to-text to ontologies/semantics requires decent term extraction and decent term extraction requires some degree of natural language processing and natural language processing isn’t even close. So someone, somewhere has to enter the descriptive data

Even so, the assumption that the “production company” would have to be responsible for “creating valid semantic metadata” seems to miss the point of the web. Since all BBC programmes are described as Linked Data anyone could easily make the claim that any BBC episode / segment was “about” the species turtle. Whether the claim was true doesn’t matter so much as whether it was validated / whose claims you choose to trust

So (at least in my head) there’s a future where punters can use simple authoring tools to make claims and “semantically describe” programmes. Something similar to BBC Buzz (http://www.bbc.co.uk/apps/buzz/about) would provide semantic trackbacks for those claims. If the broadcaster chose to validate that claim it would be “broadcast” along with with the rest of the broadcast metadata (with provenance attached). Whose claims the viewer chose to trust would be a matter for the viewer.

I think there’s probably a wider point here. In a post schedule curated world unless content creators take the time to semantically describe their content / help the audience to help them describe it, it’s gonna get lost in the content tsunami. To my mind:

http://www.bbc.co.uk/blogs/radiolabs/2008/04/being_digital.shtml

still feels like a useful strawman for the future.

And automated 2nd screen apps are not imo the end game; they’re a (semi)useful side effect of decent shared ontologies and decent (publisher and user contributed) descriptive data. But pumping money into hand rolled 2nd screen apps makes as little sense as pumping money into hand rolled programme websites.

Going back to your original post you said, “So, do these ontologies matter at all then? Because if they don’t then TV isn’t going to change very much at all.” To my mind they hardly matter at all in terms of how TV is consumed, but they will make a fundamental difference to how TV is discovered. Which in a post-broadcast world feels much more important.

Awesome comment Michael. Thank you.

Your last point about ontologies making a difference to how people make discoveries is really interesting. I could see how the example I mentioned would far more useful if it was turned in on itself. If you have a science project on turtles and you type something about the reproduction system of turtles into Google and a snippet of a BBC program is linked to directly, now that is really fascinating and it feels like there’s lots of potential there.

As a side note on your point about “speech-to-text to ontologies/semantics”, we had a demo from a chap at Autonomy that somehow managed to munge together a fairly convincing ontology from a body of text. I wonder if something like this might help? Albeit an expensive solution.

I struggle to disagree with the likes of Tristan, Paul and Michael — we’ve had lots of conversations about this stuff, and I think we’re all pretty much heading in the same direction, at least with our thinking.

(For background, I coordinate something called Project Baird — http://projectbaird.com/ — which is exploring this area).

To my mind, RDF in general is an enabling technology: the whole point of it is that you don’t see it. Places produce it, other places (and software) consume it, and links fly back and forth until you’ve got a gigantic Web of Linked Data™.

Now, for linear TV on its own — from a consumption side — this doesn’t do a great deal, but it does give you a means to build things quite easily. If the information is rich and sensibly-structured, then there’s no reason why the applications which consume it have to be limited to “Find out more about this thing” — although, to be honest, that’s quite a handy thing to have to hand for some types of programme.

I think most of us are agreed that the value in whizzy stuff on the “first screen” is… limited. You’re inherently constrained by the context of the medium — a shared device, an expectation of predominantly passive consumption, and it’s some distance away. That said, I don’t rule out there being some natty applications which can work (enhanced EPGs, VoD,news tickers, that sort of thing), and RDF can play a supporting role behind the scenes in making some of that happen. I also don’t rule out somebody smarter than me hitting upon something nobody’s thought of yet which works in a shared passive environment.

But — and this is the conclusion many of those who aren’t in the business of manufacturing TVs or pushing hybrid “platforms” have reached — the scope for the really clever stuff seems to lie on the second screen, and in this environment, the better the supporting data, the smarter the applications can be, and TV ontologies go a long way towards providing rich data which is usable in this way.

The powerful aspects of RDF are the links and its, er, genericness. By using it as an abstraction layer, you can pull lots of stuff together without the need to build separate apps for every single programme which gets produced. We’ve been playing with a whole range of prototypes, ranging from the entirely generic (i.e., the “Find out more”-type affairs — http://emberapp.com/nevali/images/tablet-ui-project-baird-4/) which are based purely on the programme topics and have mechanisms to direct you to other applications or sites depending upon what type of a thing each topic is, to the more specific, such as a “Whodunit?” voting interface which got rapidly repurposed to be a virtual clapometer for BBC Question Time with approximately no code changes needed.

I saw (but not took part in) the Autumnwatch prototype, and it definitely looked nice. I have a feeling the results will probably tally with the conclusions we’ve reached, which are generally along the lines of [all to an extent inter-related]: (a) if information presented on a second screen is too transient, it won’t work (it’ll feel like you’re being deluged from all sides); (b) there’s a risk in being too pushy or demanding too much attention — a second screen is still secondary to the TV, and so will be abandoned if it gets in the way of the primary content being consumed; © some people will want to explore there-and-then, while others will want to save information for later reading, either in the dull bits, or after the fact; (d) links are useful — if you know your friend is watching the same programme, it’s useful to be able to send them a link to what you’re looking at so that they can share the experience; (e) people like to be able to go back over the information they were presented after the programme’s finished and review it; (f) you really need a sync mechanism with your receiver, especially if you’re got a PVR and so can pause live TV, or are watching a previously-recorded programme.

So, uh, yes. Sorry for the ramble. My tuppence or so, at any rate :)

Oh! I meant to say, regarding “enrichment” of text — have a gander at LUPedia: http://lupedia.ontotext.com/

Hi Mike

I admit I have a vague suspicion of anything Autonomy related. It’s what the intelligence services use to construct meaning from unstructured text (phone calls, emails etc). It’s also the tech behind a famous broadcaster’s website search. Neither case a cause for resounding celebrations :-)

In the interests of impartiality there are a couple of other options for term extraction: IBM Languageware is the “enterprise solution”; GATE is the academic (and much cheaper) alternative. But they all require fairly intensive human supervision. There’s no magic bullet to get you from text to structure… no matter how good the tech demos.

In terms of using ontologies for SEO I think there’s probably more long term mileage there than in 2nd screen stuff. Tho meant to say in earlier comment:

Where broadcasters have content that retains value past the point of first broadcast (cos it’s entertainment / education / information / whatever value doesn’t degrade with time) there is an argument that they need to do everything in their power (including second screen investment) to promote that first broadcast as “appointment” TV. Cos post first broadcast that same episode is gonna be somewhere on Bit Torrent at which point the broadcaster loses control of the content. So maybe 2nd screen stuff could be seen as a loss leader for appointment viewing?!?

Thanks Mo, I think I’ll be revisiting your comment for reference!

Michael, regarding Autonomy, the fact that they spent a ‘Security Expert’ to demonstrate the product simply because we were working in the public sector spoke volumes.

It’s perhaps an unrealistic aim, in terms of budgeting, but there is a point about hand-rolling websites alongside programmes. Ideally, the web-of-narrative-information should be constructed during the production of the show – the equivalent to the writer writing the script. Thus the website becomes hand-rolled as part of the natural process of adapting the narrative for each medium. In other words, the Web presence doesn’t just provide the ‘find out more’ experience, but it’s equal and equivalent to the experience of watching a programme… Personally I think that’s really what we should be aiming for, and then build the extra stuff on top. But then, I’m not the one with the purse strings… :-/

As promised, we’ve just published some of our findings around audiences, interactions and content for second screen experiences – http://www.bbc.co.uk/blogs/researchanddevelopment/2011/04/the-autumnwatch-companion—-de.shtml