The document describes improvements made to an existing application used to identify important moments in student collaborative chats. The improvements include: 1) Implementing a redirection system to analyze utterance timestamps to identify intense discussion periods, 2) Overlapping graphics to correlate concepts with disputed chat parts to identify more important concepts, 3) Increasing availability by creating a web application and avoiding user intervention for moment detection. The improved application can better identify important moments by considering both concept distribution and dialogue intensity over time.
Physiochemical properties of nanomaterials and its nanotoxicity.pptx
Analyzing student chat conversations
1. 4. Implemented a redirection system to see the voices in the chat context
5. Use the utterances temporal timestamps (“hot” moments identification)
• It doesn't consider the distribution of voices, but only the time when they were input
• Captures the moments when the participants are intensely disputing a subject
• Allows filtering the chat by adjusting the maximum accepted delay between
utterances only highly debated fragments are kept for analysis
100%, 75% 40% 15%
6. Overlap “Hot Moments” and “Sentence Level” graphics (“Hot sentences”)
• See the correlation between the appearances of certain concepts and the disputed
parts of the chat identify more important or inflammatory concepts
Purpose
• Improve existing application for identification of important moments from chats
Existing Application
• Use Polyphony Theory to identify “voices” that are present in chat
conversations, following Bakhtin’s perspective of what a voice should be
• Semi-automatically identify and classify the important moments from
collaborative learning chats based on the interaction of different voices:
• Pivotal moment - in the same utterance, a voice “fades out” while another appears;
• Convergent moment - 2 or more voices appear for the last time in the same utterance;
• Singular moment - several voices appear for the last time except for one which “lives”;
• Divergent moment - 2 or more voices meet and then
they appear in different areas of the conversation
• Relies on the user knowledge to choose the right
voices for detecting the important moments
• Built in Java – only locally available
• Ignores the utterances time stamps
Improvements
1. Increase availability – web app (http://pivotal-moments.ddns.net/moments/)
2. Uses different representations for different types of important moments for
easier recognition
Pivotal moments Convergent + Divergent moments Singular + Divergent moments
3. Avoid user’s intervention for a better detection of the important moments
• The user choses how many voices to be considered (n) and the application detects all
the moments that may be discovered involving any combination of n voices
Identification and Classification of the Most Important
Moments in Students’ Collaborative Chats
Costin-Gabriel Chiru and Remus Decea
“Politehnica” University of Bucharest, Department of Computer Science and Engineering
costin.chiru@cs.pub.ro, remus.decea@gmail.com
Acknowledgement: This work has been funded by University Politehnica of Bucharest,
through the “Excellence Research Grants” Program, UPB-GEX. Identifier: UPB-EXCELENȚĂ-
2016, Aplicarea metodelor de învățare automată în analiza seriilor de timp (Applying
machine learning techniques in time series analysis), Contract number 09/26.09.2016.
Conclusions
• Improved version of an application for the identification and classification of
important moments from the students collaborative chats:
• Also consider the intensity of dialogue in various fragments, besides the concepts
distribution throughout the chat (as in the first version)
• Shows important moments that were automatically identified for various sets of
concepts, besides the ones that resulted from the user’s choices (as in the first version)
more accurate results
• Web application available at http://pivotal-moments.ddns.net/moments/
• The visualization graphics provided by the old version of the application allows:
• Seen the main topics that were debated and,
• Speculating whether the participants reached agreement or not.
• With the help of the “Hot Moments” view the user may also:
• View the intensively disputed parts of a chat,
• See whether the “Hot Moments” overlap with the voices from the chat,
• Discover inflammatory topics, along with the participants that get involved in such talks
• Future work: collaborative analysis tools, so that multiple persons could
participate in the analysis