conversation analysis

Report from the first EMCA Doctoral Network meeting

Poster drawing session
Poster drawing session

I’m on my way back from the inaugural meeting of the inaugural EMCA Doctoral Network in Edinburgh this weekend, which has been one of the best PhD-related events I’ve ever had the pleasure of attending. The last word on the meeting by Anca Sterie (one of the participants) at the summing-up got it absolutely right: the openness, intellectual curiosity and thoughtful care of the organisers and the meeting as a whole was unusual and extremely encouraging.

In any case, I thought it would be useful to document how the workshop was put together, because the format and approach was well worth replicating, especially in a field like EM/CA that can only really progress if people have ways of practising and becoming skilled collaborators in data sessions.

Update: Thanks to Eric for the nice photos!

Before the workshop

The organisers sent us a full timetable before the workshop including a list of readings to have a look at. The readings were two methodology-focussed papers from Discourse Studies:

  • Lynch, M. (2000). The ethnomethodological foundations of conversation analysis. Text – Interdisciplinary Journal for the Study of Discourse, 20(4), 517–532. doi:10.1515/text.1.2000.20.4.517
  • Stokoe, E. (2012). Moving forward with membership categorization analysis: Methods for systematic analysis. Discourse Studies, 14(3), 277–303. doi:10.1177/1461445612441534

They also included other authors’ responses to both papers, highlighting methodological differences and challenges.

The sign-up sheet that the workshop organisers mailed around had asked us in advance whether we wanted to do a research presentation or use our data in a data session or both. The timetable made it clear where, when and to whom we were going to be presenting.

Tim Smith introducing the workshop
Tim Smith introducing the workshop

Day 1

10:00 – 11:00 Icebreaker:

After coffee and name-badging we made our way into the old library and were allocated seats in small groups of 5 or 6. It was nice to have our names on the tables like at a wedding or something – not a biggie but it made me feel welcome and individually catered-for immediately.

Tables at the EMCA DN wedding
Tables at the EMCA DN wedding

An orientation talk from the organisers Tim Smith and Eric Laurier got us ready to go, then we did a particularly inspired ice-breaker: 20 minutes to use the flip-chart paper and coloured markers on the tables create a hand-drawn poster about our research summarising what we were doing in our PhDs and noting down things we would like help with or wanted to collaborate on. Then the posters were hung up on a long piece of string wound all the way around the room and we had the chance to mingle, grab a coffee and circulate.

Posters hanging up after the ice-breaker
Posters hanging up after the ice-breaker

 

The string and posters stayed up throughout and had the effect of making the library feel fun, colourful and informal.

 

11:00 – 12:30 Reading Session

We then went with our groups to a few different rooms for a reading session where we discussed one of the texts we’d been sent. In my group we discussed the Stokoe text and the responses we’d read. This was a very useful way to meet each other with a common text to discuss that helped us see how we all had quite different responses and were coming from different perspectives.

The choices of methodology papers and responses had been a great idea for this reason: rather than doing the obvious thing of giving us some foundational EMCA papers this choice provided participants from many disciplines with a way to get involved in issues and debates within EMCA.

13:30-15:00 Data Session 1

After lunch we got our teeth into the first data session. We broke up into different groups of 5 or 6 and those of us who had signed up to present data got a chance to get feedback and ideas from the wonderful collaborative practice of looking at natural data together.

Some participants hadn’t done data sessions before, but there were sufficient numbers of experienced analysts and in my session Liz Stokoe (the plenary speaker) for the day was on hand to facilitate.

In any case, it was fascinating to see the work that people were doing. Geraldine Bengsch had some great multi-lingual data from hotel check-in counters that really reminded me of episodes of Fawlty Towers. It showed how the humour of that situation comedy comes in part from the interactional contradictions of working in the ‘hospitality industry’. The clue’s in the name I guess: where staff somehow have to balance the imperatives of making customers feel welcome and comfortable with needing to extract full names, credit card details, passports and other formalities and securities out of them.

15:00-16:00 Walk

This was a nice moment to have time to chat informally while having a look around the city or (in my case) sheltering from the pouring rain in a nearby cafe.

16:00 – 17:00 Plenary

Liz Stokoe presenting CARM
Liz Stokoe presenting CARM

Liz Stokoe then gave a plenary talk that was unusual in that it focussed mostly on her methods of explaining EMCA to non-academics and on the challenges of making EMCA research both public and open to commercial opportunities in communication training with CARM. This kind of talk just doesn’t get done often enough in academia in general. In many fields academics treat discussing ‘impact’ as a necessary evil reserved for review boards, grant justifications and job applications – but never even mention it as part of PhD student training. Stokoe’s talk was open and honest about the challenges and conflicts in this process and it was really useful to see how someone had learned – through repeated efforts – to explain this kind of work to people effectively in non-academic workshop environments.

Although the talk didn’t really relate to the papers we’d read in preparation for the meeting I actually thought this was a much more useful talk in the context than a purely academic presentation. Also, we still had plenty of time to ask Liz questions about her academic research afterwards and over a very nice dinner for all participants.

Day 2

After meeting up for a coffee at around 9 we split up into different groups of 6 again, this time for the presentation sessions.

0900 – 10:00: Presentation Sessions

Chandrika Cycil presenting her data
Chandrika Cycil presenting her data

I was presenting so I only got to see one other presenter: Chandrika Cycil who had some fantastic multi-screen data from her research on car-based interactions focussing particularly on mobile uses of media technologies. There were some lovely recordings of a family occupying very differently configured but overlapping interactional environments (i.e. front-passenger seat / back-seat / driver) together. It was fascinating to see how they worked with and around these constraints in their interactions. For example, the ways the driver could use the stereo was really constrained by having to split focus with the road etc. whereas the child in the front passenger seat could exploit unmitigated access to the stereo to do all kinds of cheekiness.

I also got some really nice feedback and references from my presentation on rhythm in social interaction that I’ll be posting soon.

10:00 – 11:30: Data Session 2

I was also presenting in the final data session. This session – and the meeting as a whole – strongly reaffirmed my affection for the data session as a research practice. There’s no academic environment I’ve found to be so consistently collaborative, principled, and generous in terms of research ideas generated and shared. So – as always in data sessions – I got some amazing analytic ideas from Mengxi Pang, Yulia Lukyanova, Anna Wilson and Lorenzo Marvulli that I can’t wait to get working on.

11:30 – 12:00: Closing review

In the last session we had a chance to give feedback and start planning the next meeting – I think the date we arrived at was the 27th/28th October. If you’re a PhD student working with (or interested in working with) EMCA, and didn’t make it to this one – I strongly recommend putting the dates in your diary for the next one!

Report from the first EMCA Doctoral Network meeting Read More »

S.P.A.M.D.: a turn-analytical mnemonic

While trawling through the many piles of conversational data being thrown at me while I’m happily ensconced as a CLIC visiting graduate student at UCLA, I’ve made a little mnemonic device for myself to help me tackle a transcript turn by turn:

Turn # (lines)
– seq:
– pos:
– act:
– mrk:
– des:

For each turn of talk, I’m asking myself:

  1. Which turn number is it?
  2. Which lines does it occupy?
  3. What is it in its local sequence? (an FPP, an SPP etc.)
  4. Which position in the sequence does it occupy?
  5. What action does it implement? (If any.)
  6. What is it marked by? (If at all.)
  7. How is it designed/shaped?

Although there’s always lots more to ask of any turn, especially in terms of its use of conventional formulations, detailed lexical/syntactic/prosodic features and how the turns in or across sequences interrelate, this seems like a reasonable set of initial questions to ask when looking at a turn for the first time.

I’m OK with the acronym S.P.A.M.D. but would welcome any suggestions for other turn-analytic question labels beginning with M or even better – with E so I can complete the set.

S.P.A.M.D.: a turn-analytical mnemonic Read More »

Conversation Analytic Transcription with CLAN

I have been looking for software tools and a sensible workflow for making Conversation Analytic style transcriptions, and I haven’t found any really useful resources that weigh up the pros and cons of different approaches.

Lorenza Mondada’s very useful presentation on using ELAN for transcription does the most concise job of summarising the main choice point in this decision:

Transcription and representation of the flow of talk and multimodal conducts:

  • Transposition from time to space
  • Representation of time is crucial
  • Two formats exist :
    • The list format (ex. CLAN, Transana,…)
    • The partition format (ex. Praat, ELAN, ANVIL,…) –> based on an infinite timeline
    • For a CA perspective on talk, the list format is more adequate for the representation of sequentiality; however, for a multimodal analysis of various simultaneous lines of action, the partition format is very useful
  • These formats have analytical implications

So I began looking at various list-format transcriber options: CLAN, Transana, and Transcriber were the ones I checked out.

Transana didn’t seem to work under Linux at all, so that was a non-starter – even though there were Unix python sources available they looked more or less abandoned to me.

Transcriber was actually in my apt repository! which was a nice surprise. I installed it and got it up and running in minutes. Unfortunately, it looked terrible, used ancient audio devices in Linux, and felt very awkward to use.

I decided to use CLAN for the following reasons:

  • It’s saves human-readable text files I can munge and edit in vim (or any other text editor)
  • it uses key-commands for almost everything (little mouse-work necessary)
  • clean, stable and simple interface and media player integration
  • It’s highly modular, separating a windowed transcription system from command-line-centric analytical tools

Basically, it has a very unixy-philosophy to it (specialised tools, loosely coupled) and it’s a joy to use.

Here’s my workflow:

Currently I am enhancing some existing transcriptions from the BNC using the original audio from the Audio BNC, which I wrote about in more detail here.

score-viewer.png

First I search for the rough transcription I’m after using Matthew Purver’s‘s SCoRE tool. Using my favourite text editor, I munge this into a text file with one turn per line, and no turn numbering.

clan-annotation-interface.png

Then I copy and paste this into CLAN’s text editor, which I’m running under WINE – there isn’t a unix version yet. The image above shows a partially complete transcription, along with the audio track below. In order to show just how useful this system is for both transcription and for enhancing existing text transcriptions, I’ve made a short screencast:

Finally, I run the ‘indent’ tool on the resulting .cha file which aligns all the overlap markers and other semi-diagrammatic elements of a CA transcription. For more information on the various utilities included with CLAN, check out the CLAN user manual.

complete-annotation.png

The resulting annotation looks pretty good in CLAN, while being both editable, searchable, and allowing timed viewing and adjusting of the linked media – either using text editing or CLAN’s integrated media browser/editor. The output (a .cex file) can just be copied and pasted into a word/libreoffice document:

libreoffice-copypaste.png

Of course before publication, the CA-transcription style will still need to be painstakingly rendered in LaTeX, which is no fun at all. I guess a LaTeX export option is my only feature request for the very impressive CLAN toolset.

Conversation Analytic Transcription with CLAN Read More »

Can you have a conversation with your TV?

In the Internet of Things, what kind of thing is your TV? And what kind of thing are you?

Researchers designing ‘Social TV’ used conversation analysis to look at how people interact while watching TV (Oehlberg, Ducheneaut, Thornton, Moore, Nickell, 2006){{1}}. They found that viewers integrated the TV into the conversation, making space for it to ‘take turns’ in much the same way as any other conversationalist.

Although the TV is traditionally quite a selfish conversationalist, speaking more than listening (unless you explicitly shut it up with the remote), the prospect of it becoming a more accommodating conversational participant as a networked device is intriguing. For example, when your Boxee remote app on your iphone detects an incoming phonecall, it will pause your TV. That’s not an especially complex interaction, but it points towards the possibility of the TV at least brokering conversation in a more fluid way by letting other devices (your iphone in this case) participate in turn taking.

It would be interesting to see whether a TV that politely paused itself when it detected a sufficiently high level of chatter in the room would encourage or discourage further communication. I suspect the former – like a schoolteacher adopting patient silence with pursed lips while the children settle down. It would be a fun thing to experiment with though.

But the real promise of networked TV is that it becomes more than just a turn-taker, and begins to participate more actively in the conversation.

Perhaps not quite as actively as that, but there are some amusing parallels between some of the proposals for future TV services and the TV-becoming-flesh in Cronenberg’s Videodrome.

For example, some of the most interesting social TV research I’ve found so far – namely the notube project has looked at ways of leveraging network technologies, linked data and Semantic Web strategies for enriching TV viewer’s experience.

Their ‘beancounter’ application uses a variety of sources including social network platforms, to aggregate data about what you’re watching and creates a detailed, machine readable profile of your habits. This profile can then be used to generate better recommendations, or even help to inform and improve your experience of viewing and discussing media. The really difficult bits of this problem – like trying to figure out what is actually being talked about are dealt with gracefully, using ‘good enough’ systems that evaluate multi-lingual natural language text strings and suggest concepts that they may refer to. Ontotext’s LUpedia service provides this entity recognition function for notube. {{2}}

But in a home, where the TV is in a shared space, does the TV learn from each member of the family separately? From listening to discussions at BT, I’ve learned that the ‘problem’ of knowing who is watching the TV, in order to recommend relevant and appropriate content is not going to be solved by having each watcher cumbersomely log in to the TV. Nor is it going to be entirely solved by logged-in or sensed 2nd screen companion devices (not everyone will have one). BT’s immediate strategy will apparently involve a more complex watershed, where what people watch at certain times of day will inform assumptions about who is watching, and what kinds of content to recommend. Communications companies just don’t have this level of access to monitoring our individual behaviours within the home, and there are probably serious privacy and consent implications that will be significant barriers to granting it to them.

And anyway, I’m more interested in what kind of device the TV becomes when it learns from our collective viewing habits, aggregate viewing behaviours and networked discussions. Does the networked TV begin to develop a compound user profile of it’s own? A unique combination of a household’s various proclivities? Is it like the family dog, which everyone interacts with individually and collectively, and is then seen as having a personality, to some extent nurtured through this process.{{3}}

This brings me back to the idea of the TV as a conversational participant. If it can develop a profile, and start to build a model of the various areas of interest and domains of knowledge that a user is interested in, can it participate in a conversation in a more complex way than turn-taking? One of the key ideas of conversation analysis is the notion of ‘repair‘, in which the contingent meanings of utterances between conversational partners are narrowed down and cross-checked for mutual comprehension through all kinds of gestural or verbal cues and repetitions.

Can the TV begin to engage in this level of conversation? Can it’s profile of established interests be used as a source of recommendations that might clarify a misunderstanding of something that has just been said, for example, correcting the misidentification of an actor by people chatting about what they’re watching together on facebook{{4}}. Or could it relieve the co-watcher’s burden of responding to annoying whispered questions during films (‘why is she holding that chainsaw?’) by delving into more complex layers of in-programme dramaturgic medatada and providing some suggested explanations{{5}}.

Can the profiles of individual viewers and their shared TVs then be evaluated quantitatively for similarities over specific periods of time as a measure of the effectiveness of this kind of conversational grounding with various types of content and TV format?

And at what point do we reach the threshold of complexity, fluency and multi-valency beyond which these kinds of interactions with your TV can be thought of as a conversation?

[[1]] Oehlberg, L., Ducheneaut, N., Thornton, J. D., Moore, R. J., & Nickell, E. (2006). Social TV : Designing for Distributed , Sociable Television Viewing. Theater. [[1]] [[2]] I think of this as graceful because it addresses a hugely complex set of contingencies in a simple and contingent way, by issuing a query to a good enough service via a standard API, that does something useful, and assumes that in the future, when there is more linked, semantically enriched data, and more advanced inference services available, the API can just be plugged into those. [[2]] [[3]] I’m aware that this idea that pets do not have Disney-like anthropomorphic personalities is not popular, especially with British people, and I’m not backing it up with anything other than my own supposition that this is the case, and that your animals would eat you in a second if they were hungry enough and you were incapable of defending yourself. [[3]] [[4]] In the lingo of conversation analysis this would be called ‘self-initiated self-repair’ [[4]] [[5]] Other-initiated self-repair [[5]]

Can you have a conversation with your TV? Read More »