dcla13 discourse, computation and context – sociocultural dcla
DESCRIPTION
My DCLA13 talk at LAK13 in Leuven. The images should all be CC licensed with links provided in the speaker notes on the slides. I'd recommend looking at the other slides from this session (see http://www.solaresearch.org/events/lak/lak13/dcla13/ ) particularly those on context - this presentation provides a theoretical perspective on context, which some of the other presentations were showing really interesting examples of in empirical (and well theorised) work.TRANSCRIPT
Discourse, Computation and
Context – Sociocultural DCLA Revisited
CC-BY-NC http://xkcd.com/1085/
Simon Knight @sjgknighthttp://people.kmi.open.ac.uk/knight/
Karen Littletonhttp://www.open.ac.uk/education-and-languages/main/people/k.s.littleton
Acknowledgements/References
This talk is based on:Knight, S, and Littleton, K. “Discourse, Computation and Context – Sociocultural DCLA Revisited.” In 1st International Workshop on Discourse-Centric Learning Analytics. Leuven, Belgium, 2013. http://oro.open.ac.uk/36640/
Thanks to the anonymous reviewers, my supervisors Simon and Karen, Yulan He and Rebecca Ferguson for helpful comments and conversations.
Context in context
1. Research around discourse for learning2. Particular emphasis on context3. Current work in DCLA4. Remaining challenge5. Moving forward
Discourse for learningsupporting individuals’ subject
learningTrack subject language
supporting psychological development – the development of oral language and reasoning
skills
Track argument behaviours, structures, and language
promoting whole class and small group understanding or
commonality
Track social interaction
enabling sharing of ideas that can be improved together (both whole
class and small group)…
Track co-constructive artefact development
Context in learning discourse
Common Knowledge• Shared Perspective• Built through discourse and joint action• It both constitutes a context, and is a dynamic
facet of context
Temporal ContextMercer, Neil (2008) “The Seeds of Time: Why Classroom Dialogue Needs a Temporal Analysis.”
• Education is not a series of discrete events• Dialogue mediates this long term process• Yet this aspect is not well theorised or
empirically studied
Common Knowledge
• Dialogue as a representation of context• Dialogue as a dynamic co-constructed context
Background Dynamic
Historic Fluid
Based in communities of practice
Built on co-construction within groups
Context in learning discourseExploratory Talk and Accountable Talk• Critical constructive
engagement• Justifications given• Active participation• Reasoning is visible, and talk
is accountable – participants Interthink
E.g. of exploratory talk
A: OK (reads) so is she a good role model, and why?A: Yeah I think she isB: Wait though, we have to say why, why is she a good role model?A: Hm, she’s a good role model because she helped people…B: ….she helped them to get better A: Yeah, and she helped them in like, a new way?B: Right, she started a new method
DCLAsupporting individuals’ subject learning Subject
vocabulary use
supporting psychological development – the development of oral language and reasoning
skills
Rhetorical marker use
promoting whole class and small group understanding or commonality
Social Network Analysis
enabling sharing of ideas that can be improved together (both whole class and small group)…
Temporally driven discourse
analysis
What do we want?
• Domain general analytics• For peer talk rather than peer/teacher or
peer/conversation agent• With a focus on common knowledge, and
exploratory talk – related to positive outcomes
Combining DCLA
Education is interested in progress – getting from here to there. Methods can be combined to:1. identify the apt-concepts for any particular
discourse (row 1)2. understand the network of intercolutors and
their contributions – conceptual & rhetorical interthinking resources (rows 2 and 3)
3. and understanding how these discourses are related across time and location (row 4)
The Challenge for context…
• Moving beyond understanding where someone else is
• To understanding how they got there too (and who they travelled with)
• And the (joint) artefacts they created, left behind, and plan to create
What is happening v. what is being done
E.g. of exploratory talk
A: OK (reads) so is she a good role model, and why?A: Yeah I think she isB: Wait though, we have to say why, why is she a good role model?A: Hm, she’s a good role model because she helped people…B: ….she helped them to get better A: Yeah, and she helped them in like, a new way?B: Right, she started a new method
Disambiguation - A ‘basic’ challenge in DCLA
Example problems in DCLA & Education
• “Context is a problem, isn’t it?” Rhetoric and anaphora are linguistic problems (as are speech acts such as ‘can you pass the salt?’)
• “Stop that!” “is that right?” spoken by a child or teacher is rather different
• “What were the key medical developments?” Asked at start/end of lesson, or of different students
• “The wiglywoo” Words (or non-words) take on new meanings through on going dialogue
• “should think about x” Can be an imperative, a hypothesis, a question/suggestion, etc.
Documents v. Conversations
• Documents – relatively stable snapshots of distribution of topics
• Conversations unfold, and topics are renegotiated
Thus Introne and Drescher:• Analyse sequences of replies to understand word
clusters, changes, merges, splits to model co-occurrence as opposed to modelling based on dictionaries or other corpora
Methods as Context
The data we train on, and the processes we use have an impact on classifications and how we can treat them
• Training sets require human coders – their methods are an aspect of context
• Document splitting which treats each turn as a document may gloss temporal aspects of context
• Smoothing across turns may rectify this (cf current evidence)
• Other methods for temporal analysis
Methods as Context
• Topic modelling for knowledge flow (and cohesive ties) (domain, and possibly social network specific) – Transactivity
• Plus exploratory talk markers (domain-general)
Discourse as Context
Discourse:1. exists within a particular context and
mediates it, and2. creates context, as a dynamic, collaborative,
discursive property.
You Are Here
Your Route/ landscape
Going forward (with…)
Parallels in Pragmatic Web?
Move from:1. syntax (logical forms and symbolic structures) 2. semantics (meaning of symbols) – preoccupation
with standardisation based on ontologies3. pragmatics address the evolving contexts and
practices of creating, using, and developing epistemic artifacts.
Language in action: The shift “from…language as a tool of representing the world to its view as a means of interacting with the world.” [22:39]Context, language in action, community networks
A note of caution
• DCLA has novel potential for new assessment practices
• However, we should tread cautiously…“in developing tools with limited – but perhaps unstated – views on the nature of language use for learning”
• Once we can detect – how do we support? (offline work)
Thank you for listening
@[email protected]://people.kmi.open.ac.uk/knight/