Learning-Oriented Question Recommendation Using Bloom’s Learning Taxonomy and Variable Length Hidden Markov Models - Transactions on Large-Scale Data-and Knowledge-Centered Systems - page 33

Fig. 2. Question-topic mapping

is the knowledge dimension and

C = { remember, understand, apply, analyze, evaluate, create } (4)

is the cognitive process dimension. A more detailed explanation of these con-
cepts can be found in [ 2 ]. Similarly to the topic mapping, we define the learning
objective mapping as

M ˆ ↆQ×L , ( q, ˆ ) ∈M ˆ ⃒⃐ q is mapped to learning objective ˆ. (5)

In contrast to the topic mapping M ˄ , we allow questions to be mapped to a

single learning objective. Intuitively, this means that a question can refer to
several topics, but a single learning goal. In general, questions refer also to a
single topic. For simplicity, we have only dealt with single topic assignments in
our experiments and, in the following, we consider

M t au to map each question

to a single topic.

Question Projections. Based on the mapping relationships, we define the
following projection functions:

1. The topic projection - a function that projects a question on the topic
space using the mapping

M ˄ :

p ˄ : QₒT ,p ˄ ( q )= ˄ ⃒⃐ ∃ ( q, ˄ ) ∈M ˄ (6)

2. The learning objective projection - a function that projects a question
on the learning objective space using the mapping M ˆ :

p ˆ : QₒL ,p ˆ ( q )= ˆ ⃒⃐ ∃ ( q, ˆ ) ∈M ˆ (7)

2.2 The Learning-Oriented Recommendation Model

In the following, a novel learning-oriented question recommendation technique
is introduced that aims at improving the user's learning experience based on the
following intuition.