Framework of proposed recommendation approach

Generally, as mentioned in section 2.2.1, recommender systems consist of three main parts: input sources, recommendation method and output recommendations. The framework of learning resources recommendation is shown in Fig.5.1. It consists of the following parts.

(i) Input sources:

According to different part of the proposed framework, input sources consist of two parts: learner profiles and learning resources. Learner profile is made up of (i) keyword map based preference, (ii) learning process based preference (this part consists of learning process and learner rating calculated based on learning process) and (iii) the preference of learner relationships (comparisons) based on analysis of social interactions.

Chapter 5: Recommendation of Learning Resources 83

Learner process Learning resources

Learner keyword map profile

Learners’

comparison profile

Learner rating profile Content filtering based on

keyword map

Collaborative filtering based on learner’s relationship Hybrid filtering

Recommending Method

Input sources

User interface Output

recommendation

Fig.5.1 Framework for recommendation of learning resources

Recommendation of learning resources consists of three main parts: input sources, recommending method and output recommendations. Input sources (deep grey) consist of two parts: learner profiles and learning resources. Learner profile is made up of (i) keyword map based preference, (ii) learning process based preference (this part consists of learning process and learner rating calculated based on learning process) and (iii) learners’ comparison based on analysis of social interaction. The recommendation method (light blue) contains three stages: (1) content filtering based on keyword map; (2) collaborative filtering based on learner’s relationship and (3) hybrid filtering. We provide top N recommendation list for output recommendation (green).

Learning resources can be used to help learners to examine their understanding from diverse perspectives, making connections across related concepts, and bridge the gap between their theoretical understanding and practical knowledge [Jeong and Hmelo-Silver, 2010]. On the other hand, learning resources are different from music, movies and etc. and the definition is as follows.

Definition 5.1 : Learning resource means any digital educational resource used within the e-learning environment, it could be a course, a web page, a simulation, i.e. all known formats of digital educational resources regardless their granularity. In our case, learning resource consists of web page, note and comments.

84 Chapter 5: Recommendation of Learning Resources

lr4 lr7 lr1 lr5 lr8 lr12

u3

1 2 3 4 5 6

} , ,

{ _{a ab abj}

b wp note c

lr =

} ,

,..., ,

, , , ,

{

_{3 4 3 7 3 1 3 12}

3

= < u lr > < u lr > < u lr > < u lr >

LP

_u

Fig.5.2 Learning resource and learning process

Learning resource consists of web page, note (abstract & remark) and comments. Learning process is sequence of learning resources. For typical learning resources sequence is defined (e.g. book chapters: 1, 2, 3, ... , n). For informal learning resources (e.g. websites) sequence is not defined. Recommender system doesn’t consider this sequence.

Proposal considers and estimates this sequence.

Learning resource consists of web page, note (abstract & remark), comments (Fig.5.2). Learning resource b expresses as follows: lr_b={wp_a,note_ab,c_abj}. Where

wp

_ais the web page a,

note

_abis the note which relates to web page a, and

c

_abj is the comments which orient to

note

_ab. b is the primary key.

For example (Fig.5.2),

wp

a is the content (地球の環境問題について 地球温暖化、日本、中国の環境問題など……温

暖化・砂漠化、森林保全・違法伐採対策、生物多様性・動植物保護、オゾン層保護、南極、

廃 棄物 ・ 化学物質な ど 地 球規 模、局地的な 国際環 境条 約な ど) of the web page (http://www.seidokanri.jp).

note

ab is the content which consists of abstract (地球環境問題とは環境問題の中でも…) and

remark (地球温暖化の原因でもある温室効果ガス…)

c

abj is made up of 6 comments such as (地球温暖化の原因…温暖化の影響などについて), (工 業化、自動車の普及による…) and etc.

Chapter 5: Recommendation of Learning Resources 85

Learning process is sequence of learning resources. Learning process reflect learners’ learning interests, intents, and experiences. In our research, the definition of learning process is as follow.

Definition 5.2 : Let U ={u₁,u₂,...,u_m}denotes the set of all learners and LR={lr₁,lr₂,...,lr_o} denotes the set of all learning resources (items). A feature function, f_ui(LR), represents learner u_i’s preference for the learning resources in LR. A learning process is a sequence of tuples

>

=< _{i k}

ui u lr

lp , . Learning step is the index of tuple in a learning process. Each learner u_i may have a learning process LP_ui to express her/his preferences. Let lp_ui^j denotes the jth learning step of learner u_i and in this learning step the learner u_i learned learning resource lr_k. Then, the learning process of learner _ui can be represented as _LPui={lpui^j_}ⁿ_j=1.

For example, learner u₃ has performed 7 learning steps and learning process can be represented as }

, ,..., , , , , ,

{ _{3 4 3 7 3 1 3 12}

3= <u lr ><u lr ><u lr > <u lr >

LP_u . In this learning process, tuple

< u

₃

,lr

₄

>

’s learning step is 1; tuple

< u

₃

,lr

₇

>

’s learning step is 2; tuple

< u

₃

,lr

₁

>

’s learning step is 3;

>

< u

₃

,lr

₁₂ ’s learning step is 7. The same learning resource can be accessed multiple times in separate learning steps (Fig.5.2). In our e-NOTEBOOK system, learning processes are generated after learners wrote note or wrote comment (Fig.3.2).

The learner profile is as follows:

) ,

, ,

(

_{i i relationshipi ratingi}

i

LP KeywordMap A A

Learner =

Where, LPi is the learning process of learner i.

KeywordMap

_i

=< K

_i

, R

_i

>

is the keyword map of learner i based on learning process. Arelationship

[

ri rin

]

i = ,1,..., , is the relationship of learner i and other learners, and this part is estimated by learners’ comparison. Arating

[

ri rmi

]

i = 1,,..., , is the rating which calculated based on the learning process of learner i. n is the number of learner. m is the number of learning resource.

For example,

]) 0000 . 0 ,..., 0276 . 0 , 1638 . 0 [

], 0167 . 0 ,..., 4667 . 0 , 0167 . 0 [

, } 6955 . 0 , , ,...,

2.8627 , , {

}, 51 . 0 , ,...,

72 . 0 , {

}, , ,..., ,

({

_{5 4 5 98}

5

>

>

<

>

<

>

<

>

<

<

>

<

>

<

=

伐採 洪水 ガス

地球 地熱

温室

lr u lr

u

Learner

86 Chapter 5: Recommendation of Learning Resources

(ii) Recommendation method

The recommendation method is the heart of the recommendation of learning resources. It aims to provide both accurate and justifiable recommendations. The recommendation method contains three stages: (1) content filtering based on keyword map, (2) collaborative filtering based on learner’s relationship, and (3) hybrid filtering.

(iii) Output recommendation

Learners interact with e-NOTEBOOK system through its interface. Provided functions for example is meant to help a learner to keep track of the learning process he has performed, record the log data, provide recommendation of learning resources as navigation and etc. Here, we provide top N recommendation list.

The following sections will introduce our proposed recommending methods.

ドキュメント内 COLLABORATION AND RECOMMENDATION FOR GROUP LEARNING USING IN e-NOTEBOOK SYSTEM (ページ 96-100)