Imputing KCs with Representations of Problem Content and Context
Z. Pardos, und A. Dadu. Proceedings of the 25th Conference on User Modeling, Adaptation and Personalization, Seite 148--155. New York, NY, USA, ACM, (2017)
DOI: 10.1145/3079628.3079689
Zusammenfassung
Cognitive task analysis is a laborious process made more onerous in educational platforms where many problems are user created and mostly left without identified knowledge components. Past approaches to this issue of untagged problems have centered around text mining to impute knowledge components (KC). In this work, we advance KC imputation research by modeling both the content (text) of a problem as well as the context (problems around it) using a novel application of skip-gram based representation learning applied to tens of thousands of student response sequences from the ASSISTments 2012 public dataset. We find that there is as much information in the contextual representation as the content representation, with the combination of sources of information leading to a 90\% accuracy in predicting the missing skill from a KC model of 198. This work underscores the value of considering problems in context for the KC prediction task and has broad implications for its use with other modeling objectives such as KC model improvement.
Proceedings of the 25th Conference on User Modeling, Adaptation and Personalization
Jahr
2017
Seiten
148--155
Verlag
ACM
Reihe
UMAP '17
citeulike-article-id
14390619
isbn
978-1-4503-4635-1
citeulike-linkout-1
http://dx.doi.org/10.1145/3079628.3079689
priority
4
posted-at
2017-07-10 11:07:28
citeulike-linkout-0
http://portal.acm.org/citation.cfm?id=3079689
comment
(private-note)The idea is to deduce problem metadata from application context. If two problems were applied in the same context, they should have similar metagdata.
Using skipgram (World2Vec approach).
Evaluation: withdraw some metadata and try to predict it.
Try to see whether problems with he same KC clustered together - answer not quite.
But using NN with 100 nodes in hidden layer, allows to rise precision over 80\%
In the context and in the problem there is about equal amount of information about the problem.
Using teacher expertise! Sequences defined by the teacher...
%0 Conference Paper
%1 citeulike:14390619
%A Pardos, Zachary A.
%A Dadu, Anant
%B Proceedings of the 25th Conference on User Modeling, Adaptation and Personalization
%C New York, NY, USA
%D 2017
%I ACM
%K content-modeling context domain-model umap2017
%P 148--155
%R 10.1145/3079628.3079689
%T Imputing KCs with Representations of Problem Content and Context
%U http://dx.doi.org/10.1145/3079628.3079689
%X Cognitive task analysis is a laborious process made more onerous in educational platforms where many problems are user created and mostly left without identified knowledge components. Past approaches to this issue of untagged problems have centered around text mining to impute knowledge components (KC). In this work, we advance KC imputation research by modeling both the content (text) of a problem as well as the context (problems around it) using a novel application of skip-gram based representation learning applied to tens of thousands of student response sequences from the ASSISTments 2012 public dataset. We find that there is as much information in the contextual representation as the content representation, with the combination of sources of information leading to a 90\% accuracy in predicting the missing skill from a KC model of 198. This work underscores the value of considering problems in context for the KC prediction task and has broad implications for its use with other modeling objectives such as KC model improvement.
%@ 978-1-4503-4635-1
@inproceedings{citeulike:14390619,
abstract = {{Cognitive task analysis is a laborious process made more onerous in educational platforms where many problems are user created and mostly left without identified knowledge components. Past approaches to this issue of untagged problems have centered around text mining to impute knowledge components (KC). In this work, we advance KC imputation research by modeling both the content (text) of a problem as well as the context (problems around it) using a novel application of skip-gram based representation learning applied to tens of thousands of student response sequences from the ASSISTments 2012 public dataset. We find that there is as much information in the contextual representation as the content representation, with the combination of sources of information leading to a 90\% accuracy in predicting the missing skill from a KC model of 198. This work underscores the value of considering problems in context for the KC prediction task and has broad implications for its use with other modeling objectives such as KC model improvement.}},
added-at = {2018-03-19T12:24:51.000+0100},
address = {New York, NY, USA},
author = {Pardos, Zachary A. and Dadu, Anant},
biburl = {https://www.bibsonomy.org/bibtex/270d0111c9c0135791d642a81d6cf0da5/aho},
booktitle = {Proceedings of the 25th Conference on User Modeling, Adaptation and Personalization},
citeulike-article-id = {14390619},
citeulike-linkout-0 = {http://portal.acm.org/citation.cfm?id=3079689},
citeulike-linkout-1 = {http://dx.doi.org/10.1145/3079628.3079689},
comment = {(private-note)The idea is to deduce problem metadata from application context. If two problems were applied in the same context, they should have similar metagdata.
Using skipgram (World2Vec approach).
Evaluation: withdraw some metadata and try to predict it.
Try to see whether problems with he same KC clustered together - answer not quite.
But using NN with 100 nodes in hidden layer, allows to rise precision over 80\%
In the context and in the problem there is about equal amount of information about the problem.
Using teacher expertise! Sequences defined by the teacher...},
doi = {10.1145/3079628.3079689},
interhash = {72014761ee1ca1cdb9c07c36d2971dc0},
intrahash = {70d0111c9c0135791d642a81d6cf0da5},
isbn = {978-1-4503-4635-1},
keywords = {content-modeling context domain-model umap2017},
location = {Bratislava, Slovakia},
pages = {148--155},
posted-at = {2017-07-10 11:07:28},
priority = {4},
publisher = {ACM},
series = {UMAP '17},
timestamp = {2018-03-19T12:24:51.000+0100},
title = {{Imputing KCs with Representations of Problem Content and Context}},
url = {http://dx.doi.org/10.1145/3079628.3079689},
year = 2017
}