molly.com
Monday 30 June 2008
Microformats: Machines Must Do Irony
We were just toying with the idea of a non-serious microformat about attitudes called “Microtude.” Some pointed out that defining attitude for verbal human language as a microformat couldn’t possibly stand; others saw its merit. Then someone said “machines don’t do irony.” Which got me thinking about the semantic web. Lowercase or uppercase, this comment disturbs me.
Semantics means “meaning” – it’s that simple.
If I say “pop” it’s up to you to determine whether I mean a soda or a sound. Semantic differences abound in every language, every culture. Meaning is layered, and that’s part of the beauty of individual languages, whether spoken or programmed. If a semantic web is to succeed, semantics for machines MUST include irony, negativity, or any other negative emotion of the human process.
Yes, I am well aware that the microformats community has worked exceedingly hard to leave out some of those aspects and I applaud that idealism. And I applaud it again.
Still, I maintain that approach is unrealistic. If machines are meant to express ideas to each other as humans might, then our flaws must be introduced to those machines and the software that runs them. Isn’t the idea “human first, then machine?” What I come up with is this: Unless we’re trying to try to make an artificial intelligence define Utopia, I believe we’re stuck solving human problems in human ways. Technology can help, but cannot replace.
Therefore, all that is human must be acknowledged.
Filed under: community, creativity, innovation, professional, society, software, standards, w3c
Posted by: Molly | 04:07 | Comments (21)
But doesn’t a lot of this come around to context? If the meaning of the communication isn’t apparent from the context it’s in then there’s something wrong with the sentence.
The other fundamental question is how do we approach the web as a communication tool? Words are a tiny percentage of the communication that goes on between real people. Words are brilliant at high level concepts and detail, but absolutely inadequate for anything else. Trying to plug in the subtleties of natural communication into any medium using only words is impossible. The solution isn’t to wrap ’semantics’ around words but to write with respect to the limits of the medium.
Written word isn’t the place for ‘natural’ communication, it’s a limited gamut. It always will be, and it’ll always be impossible to semantify pure words by shoe-horning in ‘contextual’ communication.
Semantics, in my view regarding Web content, is to aide in comprehension. I don’t believe it should carried to the extreme of ‘understanding’ particularly as is applies to the human condition. As it applies to that, understanding may be more about ‘having been there and done that’. End of online philosophy … back to reinsertion of tequila IV into jugular.
some of us know “pop” doesn’t mean soda.
I’m 50/50 with you on this; just as i’m 50/50 on microformats in general. I appriciate that without catering for the human condition we cannot hope for meaningful results but is that necessarily the ‘end game’ where microformats are concerned? I doubt it.
They provide a neat way of expressing semantics but, as in your example, they do not express context, which is where they fall over and might seem a little ill-concieved.
That said, microformats don’t really have a human interface as such; they define a sub set of comprehension which could later be mined and used appropriately. It’s not their fault that we can’t express ourselves in a concrete, declarative way,..and it’s not our fault either; it’s just the way it is. As long as the computers themselves can communicate in such a way, why does it matter that we can’t join in?
“If the meaning of the communication isn’t apparent from the context it’s in then there’s something wrong with the sentence.” I disagree. It is OUR place to provide the context. The problem is that context can’t be expressed declaratively, it’s not static and cannot be derived. At best, context could be expressed algorithmically but the heuristics are somewhat difficult!
The missing link is that there isn’t currently a way of bridging the gap between human input, textual or not, and what said input would mean to another human. If that gap didn’t exist would microformats become more useful? Maybe, but if we had that technology would we even need microformats? Probably not.
We’re still MILES off that kind of tech though. The dynamics of lexical analysis capable of parsing every language and dialect is almost insummountable. Moreover, the common-sense-reasoning data alone would certainly consume most, if not all (and more) of the digital storage in existence , even in todays world.
In short, i’m not the biggest fan of using microformats for semantic reasons because i think they will become redundant before they see the light of day and they clearly don’t scale well enough but as far as marking-up an address or contact information (etc,..) discreetely, i don’t see the problem. It’s a ‘band-aid’ fix but a fix (of sorts) it remains.
@Molly:
I should clarify my comment, because it wasn’t meant to disturb!
“Machines don’t do irony” should be expanded to “it’s very difficult for machines to parse the level of intended irony in a given human statement”. As @Matt argues, we shouldn’t try and encapsulate all layers of meaning into our statements. We must leave some of it to context. Semantics means meaning, pure and simple, but meaning cannot exist in a cultural vacuum. I suppose I’m making a “death of the author, birth of the reader” type argument. The problem of making machines able to understand the vagaries of culture is surely as difficult as @Stimul8d describes. What would be ironic in one culture is not in another. Communication between people relies on shared cultural understanding – that’s hard to establish in plain text, which is of course why people use emoticons (and in other contexts, this is why people use manners). Do we want a machine that can second-guess the correct culturally-determined response to a given statement? My recommendations on Amazon aren’t necessarily that accurate. Sometimes I just want something out of the blue. I think people often irony as a defence mechanism to prevent themselves being second-guessed.
There’s a paradox here, surely, which is that microformats are by definition human and machine readable, yet part of what makes us human is that we are “not” machine, that we can’t necessarily be second-guessed. I too privilege human concerns over machine, but that’s why I think it’s actually more human to leave some things unsaid and left to cultural context.
Molly,
Semantics is meaning, not emotion. The statement “semantics for machines MUST include irony, negativity, or any other negative emotion of the human process.” implies that a computer can know what to do with emotion. It may know what styles to apply to content marked as having some emotion, but the computer can’t know what irony, sadness, happiness or any other emotion really means… It’s a silly argument about what meaning can a machine understand…
All that aside, the real question I have is why do you (or anyone) want a web tool that understands some emotions based microformat embedded in a markup language like XHTML? I get that your microformat wasn’t serious, but if it were, what would really be gained by writing tools to parse it?
Emotion: “incidents of coordinated changes in several areas, including what has been called the ‘reaction triad’ of physiological arousal, motor expression, and subjective feeling, in response to either an internal or an external event of significant importance to an individual”. How is that not representable in code? I do it every day, it’s just written differently.
Emotions are nothing more than conditioned responses, internal or external, generated from an internal state to serve some environmental purpose. I see delegates/events, encapsulation, method calls and return values. Maybe i’m over simplifying but don’t overestimate humans, we are predictable, programmable machines. We respond to events based on the knowledge we have availible to us at the time. Not to be the misanthropist but all that seperates us from machines is the complexity of the system.
As far as seeing the benifit of expressing irony or sarcasm in a digital format is concerned; ever recieved a text message that seems a bit abrupt or rude which turned out to be completely innocent? Consider how differently you would respond if your mobile phone could push that information to a speech engine which would read the message with a sarcastic tone, or not, as required. Cool yeah? Well why not? I would be able to hear the tone you used the word “silly” in. (Sorry, i just think it’s a little rude)
I think there are better ways to achieve this kind if thing, albeit in the future, but i’m glad people are addressing it. Personally, i think the best , and most immediately implementable way of expressing meaning in text is the use of aural stylesheets. If you move the emphasis from one word to the next in a sentence it’s easy to see how it offers a much better representation of intent. Just a shame they’re such a pain in the arse.
A few things:
1. This is why the Semantic Web (Upper Case, w3.org type Semantic Web) is focused heavily on bioscience, and really fairly serious applications. Airline bookings, academic bibliographies, metadata, library catalogues, restaurant menus and scientific research tend not to be irony-laden or emotion-driven to the extent of excluding automation. The W3C Semantic Web initiative is not trying to taxonomise and ontologically map human emotions – we just want to ensure we never have to spend hours cross-referencing plane bookings with hotel bookings with trains and car rentals and restaurant opening hours.
2. The name “Semantic Web” is a misnomer too. It is not a drop in replacement for the Web. The Semantic Web is already a success in certain areas. Some large biomedical research companies are already using the technology developed for the Semantic Web, and using it to better research drugs and treatments that save lives. I’d be quite happy to live without the Semantic Web going any further if it means we cure a biomedical biggie like cancer or AIDS as a result of the technology originally intended for the Semantic Web. Similarly, if the Semantic Web doesn’t recognise the irony in my e-mail or the sarcasm in my Twitter posts, but allows me to be able to do really dull things like book train tickets or borrow library books or stay up-to-date with politics, I’d consider that a success. Bleh to utopianism – we’ll get there – maybe – by baby steps.
3. One of the things I have found really depressing is that so many people seem to think that e-mail is a “low bandwidth” communication medium, and that people won’t be able to express the full, rich range of emotion and meaning that they can in real life. Bullshit. Just means we aren’t teaching people to write expressively. Are you telling me that we are lacking something having to read Jane Austen *in a book* rather than watching her give us a podcast (or the Victorian equivalent)? Phooey. Some ideas translate well into words, as do some people. Those people who don’t do so well in a written medium are just having a hissy fit about it. If you can’t display your ideas and emotions clearly in an e-mail, don’t just curse the whole of written culture – take a damn writing course or pick up a writing manual. Add class=”irony” to your paragraphs and spans if you want, but there are good in-built techniques in our language – and if you use them properly, there is no need to make it explicit that you are being ironic or goofy or bad-tempered as it’ll show quite clearly.
Morris–
That was an excellent post. Thank you.
And here I was saying that the W3C has no desire to implement a taxonomy of human emotions, only to find that a bunch of researchers in Japan have managed to convince the W3C to let them run an Incubator Group (or an XG as we call them) on creating just that. I’ve blogged it in dismay: http://tommorris.org/blog/2008/07/01#When:22:29:49
Whoa, I see, I read and I wonder.
Molly, you can do this forever, say something and again and again and no one hears or see. I can give a comment to you and no one reads between the lines.
Semantic open web = Peaceful open society.
The times are a changing and it’s time for dissent.
http://www.youtube.com/watch?v=KXCnzQl_rDg
Thank you,Molly:
you was always inspiring me.
Emotion is not irrational.
“Emotion is not irrational.”
Thoses who have irrational thoughts would think emotions are meaningless.
@Stimul8d
The primary emotion which drives us all is fear. Do you say that you do not fear? If so, then you have suppressed your emotions to such an extend that you have numbed them out.
People that numb out are those that may be suffering from Post Traumatic Stress Disorder. These feeling may be buried deep but at any times they can be triggered causing panic or rage to happen (flight or fight response). The worse part of this is at this time we a unable to identify the pool of rising emotions, so again we instinctively by habit suppress them. Yes, it’s a continuous cycle of self destruction.
Stimul8d, you wrote.
“We are predictable, programmable machines”
My question to you is who does the programing? I know the answer. Do you?
Irony is not dead, but it is a bit stunned, at least in written communication (and so is it’s kissing-cousin sarcasm). Somehow, Swift managed quite well with his ‘modest proposal’, but many modern readers seem to have trouble extracting non-literal and shaded meaning from textual communication. On the other hand, I am seeing increasing visual literacy, and more people understand visual irony (for example in the ‘inappropriate’ use of logos and brands).
Is this just a trade-off between literacies?
If so, we just need more punctuation to make textual irony visually explicit, such as http://en.wikipedia.org/wiki/Irony_mark .
In the absence of additional punctuation, we’ve been making do with combinations of existing punctuation marks such as the smiley (or for that matter, the venerable emphatic question ‘!?’, the trailing question ‘…?’, and so on), but that is really only a stopgap until we develop a standard punctuation primitive for negation or reversal.
However, we do have a little-used convention for rhetorical questions, enclosing the question mark in parentheses like this ‘(?)’. Perhaps this could be extended to ‘(.)’ for sarcasm and irony, and ‘(!?)’ for mock outrage.
Consider this my modest proposal(.)
@Alan Gresley
Where do i say i don’t fear or suggest that emotion IS irrational? Our emotions are rationalised by our own experiences. The fact that i believe that this can be modelled in code doesn’t mean i have PTSD!!!
“My question to you is who does the programing? I know the answer. Do you?”
If my words (“We are predictable, programmable machines”) have caused you offense then that’s you’re own fault. I presume you mean God does the programming. I generally don’t get into religious conversations because everyone has a right to belive what they will. Maybe you ought to too.
thanks
thnaks
That was an excellent post.