APPLICATION OF RECOMMENDER ENGINE IN ACADEMIC

DEGREE AND POSTGRADUATE EDUCATION KNOWLEDGE

MANAGEMENT SYSTEM

Xuan Wang and HongChun Shu

Kunming University of Science and Technology, Kunming, Yunnan, China

Keywords: Recommend engine, Searching engine, Knowledge management, Academic degree and postgraduate

education.

Abstract: Now, Knowledge Management System (KMS) has been used in some universities, although, it does help to

improve knowledge sharing and innovation capacity, another problem appeared: a large amount of data is

submitted to KMS everyday, it is difficult to find proper content only by searching engine. Because we are

not adept at accurately defining our needs with appropriate keywords and searching engine is hard to satisfy

the individuality demand. However, Recommender Engine has emerged as a promising alternative to

searching information due to its ability of recommend content to users based on the source of user’s

information, behavior and collection intelligence. This paper gives an overview of the concept of

recommender engine, then introduce an architecture for building recommender engine in academic degree

and postgraduate education KMS. Base on the project research, this paper proposes a new knowledge

discovery architecture, which combines recommender engine with Searching Engine. The issues and

solutions of Recommender Engine have been discussed in this paper and some recommender techniques

also been proposed to show how a Recommend Engine work in academic degree and postgraduate

education KMS.

1 INTRODUCTION

The development of academic degree and

postgraduate education measures the academic level

and teaching quality of a university. Advanced

education theories, practices and experience should

be applied in the management of postgraduate

education. Now Knowledge Management System

(KMS) has been used in some universities to

improve the knowledge sharing and innovation

capacity. As a result, a large amount of data is

submitted to KMS everyday, such as official

documents, academic papers, scientific research

items, work plans and reports, the information about

enrolment, education, academic degree, employment

and so on. Although, KMS does improve education

management on some level, another problem has

arisen: it is difficult to find proper content only by

searching engine in such huge database. With the

data and content increasing in KMS, finding the just

right information becomes so difficult. 1) The

traditional Searching Engine can just find the

information which we specialize exactly by

keywords, but we can do nothing if we can’t

describe it with appropriate keywords. 2) Most

valuable data and content has just been recorded, but

we need a tool which can make the information we

need appear in our screen automatically. 3) We can’t

use the data and content that is useful and valuable

which was not recorded in the system.

Fortunately, recommender engine can forecast

our interesting information with our behaviors and

then push the information automatically. Because

recommender engine records our behaviors when we

use the system, it can keep the process of using data

in KMS.

We propose a new architecture of KMS which

combines with recommender engine, then analyzes

the composite of the architecture. We also introduce

an example to explain how to use recommender

engine in the academic degree and postgraduate

education KMS.

455

Wang X. and Shu H..

APPLICATION OF RECOMMENDER ENGINE IN ACADEMIC DEGREE AND POSTGRADUATE EDUCATION KNOWLEDGE MANAGEMENT

SYSTEM.

DOI: 10.5220/0003617004550458

In Proceedings of the 13th International Conference on Enterprise Information Systems (KMKSSC-2011), pages 455-458

ISBN: 978-989-8425-54-6

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

2 RECOMMENDER ENGINE

Recommender engine is a specific type of

information filtering system technique that attempts

to recommend information items (books, news,

images, web pages, scientific literature such as

research papers etc.) that are likely to be of interest

to the user. In recommender engine users’ interest is

expressed by rating their items. And recommender

engine uses this data to predict the ratings of items

which the users have not considered yet. Finally,

recommender engine uses the predict ratings to

recommend the most interested items for the users.

A typical architecture of recommender engine is

as follows: Firstly, recommender engine collects the

information which needs to compute. These sources

include the users’ features (e.g. age, profession,

position and department), the item features (e.g.

keywords, genres) and the user-item preferences

data (gathered through questionnaires, explicit

ratings, transaction data). Secondly, recommender

engine compares the collected data with similar data

which collected from others and calculates a list of

recommended items for the users. Many techniques

are used to implement the recommender, such as

Demographic-based Recommender, Content-based

Recommender and Collaborative Filtering (CF)

which includes User-based Recommender, Item-

based Recommender and Model-based

Recommender. This technique is the core of

recommender engine, which decides the accurate

and personalized of recommender. As it was

introduced by Amazon firstly, Collaborative

Filtering is widely implemented. Thirdly,

recommender engine pushes recommendation to the

target user, and the content which given to the user

is not considered yet. With the help of recommender

engine, users can get more information than ever.

Figure 1: Overview of recommender engine.

3 ARCHITECTURE OF

RECOMMENDER ENGINE

3.1 Collection Intelligence

As we know that a large amount of education

information is added everyday, traditional searching

engine only find what we can specialize by

keywords. So knowledge is just stored in the

database or hard disk. And in traditional knowledge

management, collective intelligence locked in the

data which has generated when people search

content, download documents, and interact with

others is ignored.

Obviously, we need a new architecture to mine

the accumulated knowledge. Recommender engine

has emerged as a promising alternative of searching

engine due to its ability to discover items which

users might not find by themselves. Recommender

engine can use collective intelligence by tracing and

recording users’ behaviors, then predicting items.

Figure 2: Architecture of Recommender engine in KMS.

And the excellent advance of this architecture is

that the collective intelligence is to be recorded in

KMS by recommender engine. As we know,

collective intelligence is an important part of

organizational knowledge, and we can use it to

predict accurate items for special users and discover

more knowledge which we would never find it by

traditional technology.

3.2 Recommender Engine Versus

Searching Engine

Although, traditional searching engine has

advantage in mining knowledge, we shouldn’t

abandon it in KMS. So recommender engine and

searching engine must be used together to provide

searching service. Here, we propose a new

architecture of KMS (Figure 2).

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

456

Figure 3 shows that recommender engine has the

same importance of searching engines. Both engines

get data from KMS which has the same architecture

as before. We also see that recommender engine

uses the index of searching engines to increase

systemic efficiently.

3.3 User’s Interested Data Collection

It is believe that the more the recommender engine

understands the users the more accurate it can

recommend. So first of all, we adopt recommender

engine in KMS which collects users’ interesting.

There are two methods to collect the data. The first

is explicit method which collects interesting data by

asking the user directly. Sometimes the user gives

his rating to special items or content. The second is

implicit method which collects interesting data by

tracing user’s activities or behavior. This process is

completed automatically, but the implicit data must

be filtered before used to predict because the

behavior can’t reflect the interesting of users.

However, all the users of the KMS have the duty

to contribute their knowledge to KMS. So we can

make some regulations to require users to submit all

the interesting data to the engines. It can not only

increase the accurate of recommender but also

record the collection intelligence in the KMS.

3.4 Security and Privacy

The problem of security of items is also existent in

the recommender engine. To protect the security of

data, a secure architecture is needed to filter the

recommender information. In addition,

recommender engine knows the interests and

behaviors of the users. Most of these data is privacy

and the engine needs to protect those data.

4 RECOMMENDER

TECHNIQUES

The techniques used by recommender engine can be

classified based on the information they use.

According to the features of the academic degree

and postgraduate education, this paper pays more

attention to these recommender techniques as

follows: Rule-based Recommender, Non

personalized-based Recommender, Content-based

Recommender and Collaborative Filtering-based

Recommender.

4.1 Non-personalized Recommender

Non-personalized Recommender is simple because it

uses any information of the users to compute the

recommender items. It recommends items just base

on the importance or popularity of items. The

knowledge management can use these techniques to

push new contents, popular contents and important

contents to each user. All the users are categorized

by department, and the users in same department

may have the same interest. So the engine can

recommend items based on the department.

4.2 Content-based Recommender

Methods use the information about item features and

the ratings a user has given to items (Thomas Hess,

2009).When user searches information in KMS, his

or her behavior is recorded by the Recommender

Engine. The engine uses the individual information

to predict items which have the similar attribute to

the ones preferred in the past. The underlying

assumption of the Content-based Recommender is

that those who interest in the past tend to interest the

similar in the future.

4.3 Collaborative Filtering-based

Recommender

Obviously, both recommender technique discussed

above ignore the contribution from others. And the

collection intelligence is not considered.

Collaborative Filtering-based recommender is used

wildly in most of the e-commerce web sites.

Collaborative filtering is a method of making

automatic predictions about the interests of a user by

collecting taste information from many users. There

are three main techniques can be distinguished: user-

based, item-based, and model-based approaches. But

these approaches can be reduced to two steps:

1. Look for users who share the same rating

patterns with the active user whom the prediction is

for.

2. Use the ratings from those like-minded users

found in step 1 to calculate a prediction for the

active user. The following example of CF explains

how recommender engine works.

If user A likes item A and item B, and user b like

item A and item B, we can discover that user A and

user b have same interest. So if user c likes item c,

we can recommend item c to user A.

In the real recommender engine, all the above

and other approaches are combined together to

provide recommender service. There is an open

APPLICATION OF RECOMMENDER ENGINE IN ACADEMIC DEGREE AND POSTGRADUATE EDUCATION

KNOWLEDGE MANAGEMENT SYSTEM

457

source project named Mahout which currently

provides tools for building recommender engine

through the Taste library.

Figure 3: User-based Collaborative Filtering.

5 CONCLUSIONS

The architecture of the academic degree and

postgraduate education KMS has been discussed in

this paper. Recommender engine is important as the

searching engine in this KMS. Both engines are

combined together to provide searching and

discovery service. Recommender Engine will help to

find new contents which user have no idea but

interesting about. Furthermore, an important feature

of recommender engine is that they can record and

use collection intelligence to predict interesting

content to users.

This paper is a precursor to discuss the

application of recommender engine in the academic

degree and postgraduate education KMS. With the

developed of recommender technology, the

recommender’s accuracy, scalability and

performance will be increased. And education

managers will profit by finding new interesting

knowledge to improve their efficiency.

REFERENCES

Thomas Hess, Recommender Engines Seminar Paper,

2009: Introducing Apache Mahout Grant Ingersoll.

http://www.ibm.com/developerworks/java/library/j-maho

ut/index.html

http://en.wikipedia.org/wiki/Recommender_system

http://en.wikipedia.org/wiki/Collaborative_Filtering

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

458