A Knowledge Recommend System Based on User Model - CiteSeerX

A Knowledge Recommend System Based on User Model Jiagen SHENG, Sifeng LIU

A Knowledge Recommend System Based on User Model Jiagen SHENG 1,2, Sifeng LIU 1 1

School of Economics and Management, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China 2 School of Computer Science and Engineering, Jiangsu University of Science and Technology, Zhenjiang 212003, China doi:10.4156/jdcta.vol4. issue9.20

Abstract The content and searching method of knowledge base are lack of the consideration of user requirements, it is difficult for the user to select the knowledge satisfy with the user requirements from candidate set of a lot of knowledge items. So, we propose a new knowledge recommend system based on user model. According to the corresponding relationship between user model and the knowledge attributes, the satisfaction degree of user requirements for the candidate knowledge can be calculated by the model. To describe our model, the connotation and design method of the user model are introduced, and the architecture of knowledge recommend system based on the user model is presented. Furthermore, the method of knowledge representation and design based on attributes is proposed and the feasibility of the knowledge recommends system prototype is verified too. By using traditional data mining techniques to obtain the user identity and structure of knowledge, the proposed method can be applied to build the knowledge user database to improve the knowledge service quality.

Keyword: User Model, Knowledge, Property, Recommend System 1. Introduction User model is conducive to adaptive search, which can provide personalized and intelligent search service. By recording and learning user search history behavior, user model can more accurately analyses the user search questions. This will improve the correlation between the search results and user needs. In essence, user model improve the degree of user satisfaction by enhancing the relevance degree of user [1,2 ,3]. In addition, the user model is also helpful to locate the user's expectation needs and provide active knowledge services. Then, the model can effectively deal with the various requests about the same search information, which will be different for different user in the depth and width. After the user model is joined, the traditional search process becomes the search process with the user interest, meanwhile, to provide more accurate and friendly search service, the feedback process will aim at the opinion of the individual user. The search process is a process of matching the set of information and the set of user requirements [4, 5]. With the rapid development of science and technology, the new knowledge is discovered and gathered by human continuously, and the increment of knowledge is alarming. It becomes more difficult for people to learn and apply the massive knowledge. Therefore, knowledge process or knowledge design, which make the knowledge adapt to be learned and applied by human, is a very significant meanings job [6, 7]. This paper is organized as follows. In section 2, the related work of user model and knowledge design is detailed. In section 3, the architecture of knowledge recommend system based on user model is provided, and the basic components of each functional module are described. The user model design and structural design of knowledge property are highlighted in section 4. Meanwhile, the structural representation of the knowledge property and design method is also offered in this section. The full text is end and the next step is prospected in the last section.

2. Related work User model is an abstract description for user information. According to the information of user model, system adapts to performance and behavior of user. In general, the user model is from the user,

- 168 -

International Journal of Digital Content Technology and its Applications Volume 4, Number 9, December 2010

which is the selection and description of user’s features information, and conversely, the user model serves the user by the way of adapting to user’s needs according to User model. User features described by user model are composed of knowledge state, objectives, background, experience, preferences and behavior characteristics etc. [8, 9]. User model is used to capture and record the interested points of user needs, record and manage the user interest by establishing the model and describing the possible potential demand of interest of user [10]. The user history and evaluation for the system is recorded and used to optimize query behavior and query results in User model. User model is a feature set, used to store and manage the behavior of user history, store and study the knowledge of user behavior and derive associated knowledge. There are two kinds of access ways of personalized user features: firstly, the user’s personalized features are entered manually by user; secondly, intelligent agent is used to track user behavior for obtaining the user's individual characteristics automatically. The building process of user model is essentially a process of knowledge acquisition, and can be viewed as a process of pattern recognition too. Knowledge acquisition mainly includes two ways in user model, i.e. explicit acquisition and implicit acquisition. In the explicit acquisition mode, the user directly takes part in the modeling process and answers the questions from knowledge engineers or knowledge system. In the implicit acquisition mode, user knowledge is obtained by learning and reasoning of system. Explicit mode requires user to be present. Compared with implicit mode, this mode needs more cost and is lack of flexibility. However, the results are more reliable. Both of the two models can be performed by online way or off-line way and can be used to obtain static and dynamic user knowledge. To take advantage of the two methods, static user knowledge can be obtained by explicit model, and dynamic user knowledge can be obtained by implicit method. User behavior is ultimately determined by their cognitive processes. However, the mechanism of cognitive processes of user is very complex and involves many uncertain factors. At the present time, it is impossible to describe the user behavior using an accurate model. So, in recent years, the research of user model aims at the user's behavior pattern and regards user as a black box. This is similar to the research method used in speech recognition, which mainly uses statistical methods but not mechanism. There is natural relation between User modeling and Artificial Intelligence (AI). The primary problem of AI is knowledge representation and a variety of methods of knowledge representation have been proposed in AI field. During the process of User modeling, user can selectively use the methods of knowledge representation according to the actual demand. The common methods of user modeling mainly include logic-based methods, Bayesian methods, neural networks, machine learning and template-based method. Knowledge design is an important research field of knowledge engineering. Matthew Leitch [11] earliest proposed the term of knowledge design in his paper. He elaborated the meaning of knowledge design for human education and learning from human cognition and memory. Matthew Leitch [12] described the design method about the complex object of training materials and processes, and then completed a trial work. Yue Xiaoli and Cao Cungen further research the knowledge design and analyze the connotation of knowledge design, development state of knowledge design and influence on other research areas.

3. Architecture of knowledge recommend system based on user model In general, the maintenance of user model is performed by recording the user interaction experience, such as the time and the frequency of accessing the page, the score of exercise and test. Obviously, the latter is more reliable. In addition, it is necessary for a system that user can set its own related concepts and knowledge state through looking up their own models. This reflects the adaptability of system. Knowledge state is the important object described by user model. Overlay mode is a popular mode adopted by user. To describe the mastery degree of user, corresponding weights are appended to each concept of domain knowledge mode. Therefore, user model is the Concept-Value set, where Concept represents notion and Value denotes the mastery degree of user. In general, Value can be a binary value (known, unknown), or qualitative description (good, general, poor) or a statistical value (0 ~ 100). In some systems, there may be several factors for describing a concept, such as situation of reading, exercise scores, or master degree of the concept derived from other concepts rightly. Then, the overall score of the concept is obtained through the combination with these factors.

- 169 -

A Knowledge Recommend System Based on User Model Jiagen SHENG, Sifeng LIU

There is also a simple model describing user knowledge state, called the stereotype model. It actually a set of user model by dividing users into several groups. For example, a system classifies users as two dimensions. One has four values including Novice, Beginner, Intermediate and Advanced. The other has two values, i.e. master the basics of the initial computer and understand Unix. So user will be divided into 8 different types and concept-value is also adopted in stereotype model. Stereotype model is relatively simple and is easily implemented, but the size of adaptive granularity is not small enough. Usually by connecting overlay mode with stereotype model, stereotype model is used to initialize the value for overlay mode, and overlay mode is used to reflect the adaptation degree to user during the process of system performance.

Editing Labels

Interaction Interfaces

attribute extraction User model database

classification

Index Adapter collection

Glossary database

Figure 1. Figure of Architecture of Knowledge Recommend System Based on User Mode Knowledge recommends system based on user model consists of user model, design and support of knowledge and search engines. From the whole structure of system, bottom layer is the knowledge base based on ontology, and top layer is the user application program, and middle layer plays the part of the connection between the top layer and the bottom layer. The task of the middle layer is to select knowledge from knowledge base according to the request of application program, and service for application programand and users. The middle layer adopt semantic search strategy, meanwhile, it introduces user models and knowledge design to improve the accuracy of the knowledge supply. Knowledge representation based on ontology lays a foundation for semantic knowledge search, and user model and knowledge design can provide more useful knowledge for application program.

4. Structurization of knowledge information and design of knowledge model 4.1. Knowledge design process Knowledge design is a subject of studying how to choose, organize and communicate knowledge with the appropriate manner in the knowledge space, and make knowledge more easily perceived, digested and used by human. Its purpose is to offer appropriate knowledge to the proper person through proper way in the right time. We are living in the period with “rich information but lacking of human attention”. On the one hand, all kinds of knowledge emerge as a flood. On the other hand, the time and effort of study for human are very limited. How to make human can acquire their required knowledge

- 170 -

International Journal of Digital Content Technology and its Applications Volume 4, Number 9, December 2010

in a limited life? First, knowledge selection from the ocean of knowledge is needed for the specific user. Second, the extracted knowledge should be processed and makes it be easy to be learned and mastered by user. Therefore, knowledge design can be interpreted as knowledge process, of which raw material source is knowledge source, and the parameter is user model, then the system can provide knowledge products with invariable connotation and optimized format. knowledge source knowledge design

knowledge source

User model

Design comment

Figure 2. Schematic Diagram of Knowledge Design According to the application of knowledge design in practice, several principles of knowledge design are proposed： (1)Knowledge design is based on user model and user demand. Knowledge design services for knowledge user, aiming to supply knowledge for user in a more convenient form for human learning, grasping and using. Knowledge design is a cross subject with perception science and computer science, and only according with human cognitive process can the way of knowledge supply be suited and effective. (2)Knowledge design is a repeatedly iterative process. Knowledge design is a dynamic process, including designing, production, using, estimating and redesigning. The knowledge redesign is the key step of knowledge design. The factors of knowledge redesign mainly include the following aspects. First, the change of user model is one of the factors. User model is the foundation of knowledge design, reflects continuously change of user feature and needs repeatedly knowledge redesign to meet the user demand. Second, knowledge acquisition is really cost and it is impossible for human to completely grasp the designed knowledge, so the knowledge redesign is inevitable. (3)Evaluate of knowledge design is an important reference element of knowledge redesign. The evaluation of knowledge design is a scientific evaluate of knowledge design. Knowledge redesign is a process of amend the original design.

4.2. Formalized model of knowledge design process Knowledge design process can be formally described as follows: Problem space of knowledge design:

M  input / output set I / O, process set P, the state set Q, mapping set  . State conversion occurred by mapping, showing sub-processes:

Pi＝Qi  Qi  1  i＝1， 2n  .

P  Set  Pi  i  1, 2 ... n  . Output is obtained through the input set and the process set: I  P  0 . Combination of sub-processes form the design process:

According to design method of parameters, at each stage of knowledge design, we provide the following parameters as the design conditions and constraints. Design is complete while all constraints are satisfied.

-P  Parameters  P1, -Vr  Value Ranges 

, Pn ;

V1,

, Vn , where Vi  vi1,

- 171 -

, vin ;

A Knowledge Recommend System Based on User Model Jiagen SHENG, Sifeng LIU

-C  Constraints  c1,

,cm ;

-R  Requirements  r1,

, rk ;

-Pr  Preferences  pr1, , prj ; -Cf  Global Cost Function . Solution space of knowledge design =< to meet the knowledge requirements and design constraints of the knowledge production 0:| knowledge needs Iknowledge function Q1knowledge organization Q2knowledge representation Q3 >.

4.3. Design of metadata tables of knowledge and knowledge user Table 1. Metadata Table of Knowledge User data type explanation

number

field

1.

WorkerID

Varchar2(10)

员工 ID

2. 3. 4. 5. 6. 7.

Name City Gender Birthday EntryDate Note

Varchar2(30) Varchar2(20) Varchar2 (2) Date Date Clob

员工姓名 员工所属工作城市 性别：M 表示男，F 表示女 出生日期 入职时间 员工备注信息，可设为员工图片

number

field

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

MetadataID MetadataNum Name KnoFrame RelevantKno EditorID EditorDate Language Version MediaTyep UserType Knowledge class Sup-concept Sub-concept

Table 2. Metadata Table of Knowledge data type Varchar2(10) Varchar2(30) Varchar2(20) Varchar2 (2) Varchar2(20) Varchar2(20) Varchar2(20) Varchar2(10) Varchar2(20) Varchar2(20) Varchar2(20) Varchar2(20) Varchar2(20) Varchar2(20)

explanation 知识 ID 知识编号 知识名称 名称结构 相关知识 编辑人ID 编辑日期 语种 版本 媒体形式 角色/用户类型 知识分类 上位知识条 下位知识条

5. Conclusion When the information provided by knowledge catalogue is relatively simple, it becomes difficult for user to select appropriate knowledge from a lot of candidate knowledge set. Aiming to help user overcome the problem, we proposed a new knowledge recommend system based on user mode to help the user to overcome the problem. According to the corresponding relation between the user model and knowledge attributes, the method lays a foundation for calculating the satisfaction degree of user requirement. The main contributions of this paper include: 1) offering a user-oriented knowledge recommendation system framework; 2) designing the metadata tables of knowledge and knowledge user, which can be used to guide the design of the knowledge recommendation system. In the future, we will study the following two aspects: 1) the corresponding relation model between knowledge and user; 2) the application and development of knowledge recommend system in the library management.

6. References

- 172 -

International Journal of Digital Content Technology and its Applications Volume 4, Number 9, December 2010 [1] Zhongzhi SHI. Knowledge Discovery [M]. Beijing: Tsinghua University Press, 2002 [2] Stefani A, Strappavara C. Personalizing access to Web sites: The SiteIF project [EB/OL]. Http://wwwis. Win. Rue. Nl/ah98/Stefani/Stefani. Html, 1998-06-24 / 2004-03-12 [3] Sorensen H, Mc Elligott M. PSUN: A profiling system for usenet news [EA]. CIKM'95 Intelligent Information Agents Workshop [C]. Baltimore: ACM Press, 1995:205-211. [4] Sowa J F. Knowledge representation: Logical, philosophical, and computational foundations [M] .Brooks/Cole: Division of Thomson Learning Inc, 2000:51-54 [5] Pratt KB, Tschapek G. Visualizing concept drift [A]. Proceedings of ACM Conference on Knowledge Discovery and Data Ming [C]. Washington, DC: ACMPress, 2003:735-740 [6] Hongfei Lin, Xuegang Zhang. Text structure analysis method Based on concept [J]. Research and Development of Computer [7] ZengxiangLu, Hongchao Guang. Network information filter by bookmark services [J]. Journal of Software [8] Tao Zou, Jicheng Wang, etc. Design and implementation of data collection system based on web [J]. Intelligence, 18 (3) :195-201 [9] Resaick R, Varian R. Recommender Systems. Special Issue of Communication of the ACM, 1997:40 (3) [10] Huiling Zheng. Research and development of university library personalized information recommendation services [J]. Library Journal of Henan [11] Pazzani M. Billsus D. Learning and Revising User Profiles: The Identification of Interesting Web Sites. Machine Learning, 1997: (27) [12] Mladenie D. Text - Learning and Related Intelligent Agents: a Survey. IEEE Intelligent Systems, July / August 1999:14 (4) [13] Middleton. Stuart. Shadbolt, Nigel R. Roure. David C. De. Ontological User Profiling in Recommender Systems. ACM Transaction on Information Systems, 2004:22 (1) [14] Xiaolin Xu, Xirong Que, etc. Information filtering technology and personalized information services [J]. Computer Engineering and Applications [15] Huanqing Xu, Yongcheng Wang. User interest model based on weighted conception network [J]. Shanghai Jiaotong University, 2004 (1)

- 173 -