Information Technology Reference
In-Depth Information
specific concept, etc.), while search engines are usually queried to simply retrieve
information.
The work presented in this paper aims at improving question recommenda-
tion for QA systems by addressing these two aspects. Our main objective is to
leverage the functionality of QA systems towards new learning techniques and
use the wisdom of the crowds in order to convey useful information and guide
the learner on a meaningful learning journey. For this purpose, a domain-specific
topic-taxonomy and Bloom's learning framework is employed, whereas for mod-
eling the order in which questions are selected, variable length Markov chains
(VLMC) [
6
] are used.
The rest of the paper is structured as follows: Section 2 presents the knowledge-
base with the domain-specific and learning taxonomies; Section 3 introduces the
new learning-oriented recommender model; Section 4 gives an overview of the eval-
uation results and, finally, Section 5 makes a summary, draws some important
conclusions and presents future work objectives.
2 Approach
Aiming at improving the learning experience of users when interacting with a
QA system, question recommendation, in the context of this paper, refers to
recommending questions to users who ask them and are interested in learning
about a particular domain. The question-answer dialog with the system should
allow the user to navigate through a meaningful and useful chain of answers that
can enrich the users knowledge about a particular domain.
In order to account for the learning process or the order in which ques-
tions are selected, a probabilistic graphical model based on variable length
Markov chains [
6
] is constructed and trained on the users question browsing
history. This is not a novel approach; it has been successfully adopted for
query recommendation [
10
] as well. In this paper, we attempt to adapt and
improve this approach for question recommendation by considering two rel-
evant question features: the questions topic (or subject) and learning objec-
tive. The learning objectives, in the context of Blooms learning taxonomy
[
2
], refer to a classification of educational goals (e.g. summarizing, classify-
ing, recognizing, etc.). Current conceptions about learning assume learners as
active agents and not passive recipients or simple recorders of information.
This shift away from a passive perspective on learning towards more cogni-
tive and constructionist perspectives emphasizes what learners know (knowl-
edge) and how they think (cognitive processes) about what they know [
2
].
Therefore, the learning taxonomy is defined based on two dimensions: the
knowledge and the cognitive process.
For this purpose, two taxonomies are considered: a domain-specific taxon-
omy that contains possible question topics and Blooms learning taxonomy, a
classification of existing learning objectives. In the following, we will present the
knowledge base behind the recommendation model.
Search WWH ::
Custom Search