Artificial Intelligence-Based Higher Education Students Advising System: The Case of Ethiopia

doi:10.21203/rs.3.rs-2376655/v1

Download PDF

Research Article

Artificial Intelligence-Based Higher Education Students Advising System: The Case of Ethiopia

https://doi.org/10.21203/rs.3.rs-2376655/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Academic advisors assist students in academic matters such as registration, warning, probation and other personal matters. Successful advising systems increase student retention, improve graduation rates and help students meet educational goals. Research in academic advising shows a trend of integrating electronic technologies that advisors must adopt to increase their roles in student education, while resources are limited and the proportion of students increases. The main objective of this work is to design a model of academic advising dialogue system for higher education.

In this research work, a Higher Education Students Advising System (HESAS) has developed as an academic advising dialogue system, to allow advising session to be more centered on students and to help advisors and advisees in making better informed-decision and improved services. The design of HESAS is composed of six main components. These are user interface, spelling corrector, natural language understanding, dialogue manager, response generator and advisement service. The students have used the system during the experimental phases. For the second experimental phase, the system performed well, obtaining close to 83%, on the traditional language processing measures of precision, recall and F1 score.

natural language processing

dialogue system

conversational agents

academic advising

In higher education, academic advisors assist students’ academic, professional, social, and personal matters [1]. Academic advising was initially a response for a need to guide students in selecting their course sequence and later evolved to include assistance in matter. The innovation trend in advising has been towards the use of communication technology such as email, social networking, instant message, podcast, mobile application, online videos and blogs [2, 3]. A quick survey of university websites shows that many have introduced the concept of eAdvising, that is, utilizing electronic means, usually web-based, to offer advising to students [4, 5].

An automated system is better solution for the innovation trend in advising and addresses the concern of not enough advisors, too many students, in an economically challenging environment [1]. Such a system will lessen the burden of academic advisors from several mundane tasks and free up more of their time for the deeper aspects of advising, such as career planning or managing extraordinary personal situations.

To deliver improved academic advising services a higher education students advising system is developed as an academic advising system. The use of technology in academic advising may introduce a greater accountability and may provide better services to the students. Thus, motivated the author to develop dialogue system to allow higher education students to get advisement services anywhere and anytime via their natural language.

Dialogue system or conversational agent (CA) is a computer program designed to carry out an open and coherent conversation with a human. Dialogue systems have employed text, speech, graphics, gestures and other modes for communication. For the purpose of this research work a text-based dialogue system has used and advice students on curricula and extra-curricular related issues such as preparing students to plan, set goals, and make decision.

This research work was primarily motivated by the need to optimize advising, to determine which courses should be offered, and to utilize existing information available in the university registration system, there is no doubt that academics worldwide agree that proper advising is an important factor for students successful progress in higher education. Many studies have been conducted to confirm this matter. As a result, many academic institutes have investigated the use of computer technologies in academic advising to overcome the difficulties experienced with traditional methods. Samples of such studies are presented in this section.

2.1. Expert-based Academic Advising Dialogue Systems

Leung et al.[6] proposed an intelligent counseling system that uses an ontology-based information retrieval engine to suggest relevant information to students. The goal of the research work was to develop a next-generation, student-centered learning environment that blends the strengths of distance learning, face-to-face teaching, and online learning. The system can provide 24 × 7 online counseling services to prospective and current students by simulating the role of a teacher/advisor when responding to enquiries in areas such as career guidance and development, study paths and methods, and program/course choices. The intelligent counseling system has two modules: academic counseling and academic advisement. The academic counseling module handles general queries from prospective students, on questions about career development, program/course information, and learning modes. The academic advisement module deals with specific questions mostly from current students, on program specifics, study plans, and graduation checks. The evaluation of intelligent counseling system yielded satisfactory performance in terms of credibility of its recommendations and usability.

Latorre-Navarro and Harris [7] proposed an intelligent natural language conversational system for academic advising. The main objective of the research work was to deploy an expert-based natural language academic advising system, to manage multiple straightforward advising tasks and allow advisors to engage in non-mundane educative tasks. To allow advising sessions to be more centered on the development of the student, they developed Albertas academic advising dialogue system using an approach of pattern-matching techniques. The design of Albert is composed of four main components. These are the web interface, dialogue manager, academic planner, and database. The advising dialogue manager in Albert includes the natural language understanding and generation system, and the task manager. The dialogue manager (DM) was built around ChatScript (CS), an open source scripting language. The natural language generation component responds an answer that is taken from the knowledge base. The knowledge base has collection of information the system should return in a conversation. The NLU component of the DM produces a semantic representation of the natural language input that is appropriate for the dialogue task. The task manager component represents all the functions Albert executes to complement the dialogue task. These functions include user account management, database management, input validation, automatic updates, a scaling-up routine, and statistical data collection. The academic planner in Albert was designed to offer students guidance and recommendations when preparing their course plans for each term. Students can enter their academic record in the academic planning process and receive recommendations on how to develop their academic plan up to graduation, based on courses completed, course prerequisites and all academic rules. Students used Albert during three experimental phases. For the third phase, the system performed well, obtaining close to 80%. Albert allows unrestricted natural language communication utilizing state of the art natural language processing techniques.

2.2. Web-Based Higher Education Advising Systems

Albalooshi and Shatnawi [8] proposed a multidisciplinary web-based higher education advisory system that offer standard, advanced, and configuration management services that can easily be accessed by students, advisors, educational managers, and alumni. The standard services was based on retrieving and organizing students’ information in direct queries that require retrieval, formatting, and organizing of students’ data. Basically, the authors have five services in this category. These are student transcript and registration information, courses to be registered next semester, student’s graduation progress, student progress dependency graph, and alumni services. An advanced service, in this category the system was not only does retrieving students’ information but also it allows students to interact with the information such as GPA simulation. Educational managers were able to retrieve useful information such as students’ statistical information and statistical information for managers. Configuration management services these services was enable the system manager to configure the business rules and system parameters, or, on the other hand, to update, delete, or insert new records to the database, i.e. if a new degree program was introduced. A multidisciplinary web-based higher education advisory system was evaluated by several users (students and advisors). The evaluation result was satisfactorily positive. The system provides the user with valuable, accurate information; can solve many advising issues. However, a Multidisciplinary Web-Based Higher Education Advisory System was only designed to perform simple tasks. In addition to that the system wasn’t ease of use and user friendliness.

Feghali et al. [9] proposed online advisor system that provides students with a very basic interface that contains no intelligence: it is an online transcript, showing the semester in which the student is registered with a listing of the courses took with their respective grades. The Online Advisor creates academic schedules semester-to-semester and year-to-year by organizing information from many sources. The system provides information needed for course planning in an understandable and visually appealing way. In particular, the system displays the major and overall average, indicates which major, university and distribution requirements have been satisfied and which need to be completed. It displays the number of credits completed and the number still needed for graduation. A database management system through a web application acts as an interface between applications and the Online Advisor. The online advisor is not ‘intelligent’ on its own. It only uses the rules that are stored in the database and displays them with the user in mind. The online advisor system consists of three pages: the summary, the degree checklist and the degree plan pages. The summary is the one page that is mostly used by advisors and students that gives a quick snapshot of the most current situation of a student’s file. The Degree Checklist displays the courses required from each student in order to fulfill his degree requirements. The “degree plan” pages allow the student or the advisor to select the courses that he/she thinks must be taken in a specific semester. The usability of the online advisor was assessed by several users. 79% of users stated that they were satisfied with the Online Advisor. 90% rated the Online Advisor as effective and efficient. 75% rated the Online Advisor as useful and helpful. However, the online advisor system was not intelligent.

The design of HESAS used template-based approach for natural language understanding (NLU) and response generation (RG). The template-based approach has chosen, due to, the simplest and the most common, we only need to preassemble a sentences or prompts. The dialogue management used information state update approach. The information state update approach contains many features for designing a dialogue agent such as asking clarification questions, making a proposal, rejecting misunderstood questions and suggests plans.

Generally, most of the design of dialogue systems has similar components or modules. These components can be summarized as language understanding, dialogue manager, and response generation. Even though, the general basic architectures of dialogue systems are nearly similar, they substantially differ in their internal functionalities depending on the algorithm employed and the aims of the dialogue systems were designed. As many of the dialogue systems design, HESAS also share common components such as dialogue manager, response generator and knowledge base. As a result, it consists of spelling corrector, dictionary, natural language understanding subcomponents and advisement service. The natural language understanding component performs three types of analysis of user’s statement: syntactic, discourse, and semantic analysis as shown in Fig. 1.

3.1. Spelling Correction

The user interface is responsible for accepting the user’s input and provide an output. It converts an internal representation (which user may not understand) to the user understandable form. The job of spelling corrector component is concerned with detecting misspelled words in the user’s input and possibly assisting users in correcting them with the use of dictionary. HESAS is a text-based system, thus spelling errors are very common, yet the users expect the system to handle them effectively.

The spelling corrector uses dictionary lookup techniques for detecting non-word errors. Because it is simple and quick task, it just simply lookup each word in a dictionary; any unfound words are considered as misspellings, So that, the dictionary contains list of English words. The author used an existing dictionary and on top of that, added list of technical terms. These technical terms were continuously updated throughout the experimental phases. Totally, the dictionary consists of 15,827words. Algorithm 1 shows how the spelling corrector detects the spelling error and fixes it. Algorithm 2 shows tokenization.

Algorithm 1: Spelling Corrector

Algorithm 2 Tokenization

3.2. Natural Language Understanding

The job of natural language understanding component is concerned with the analysis of the utterances a user makes in a dialogue. In order to perform analysis on the user’s utterances, the NLU component should consist of three subcomponents. These are syntactic, discourse, and semantic analysis.

3.2.1. Syntactic Analysis

The syntactic analysis consists of determining the grammatical relationships among the words in the user’s utterance. It is the first stage in the analysis of a user’s statement, the syntactic analysis, starts with tokenization of the user’s statement, that is, division of the input in a series of distinct lexical entities. There are two distinct steps in the implementations of syntactic analysis: part-of-speech tagging and parsing. The Algorithm 3 illustrates POS (part of speech) tagging.

Algorithm 3 A Part of Speech Tagging

3.2.2. Discourse Analysis

Discourse analysis consists of determining the relationships among multiple sentences. An important component of discourse analysis is reference resolution, the task of determining the entity denoted by a referring expression.

3.2.3. Semantic Analysis

Finally, the natural language understanding component performs semantic analysis, which is determining the meaning of the sentences. Typically, this consists of representing the statement in canonical formalism that maps statements with similar meaning to a single representation and that facilitates the inferences that can be drawn from the representation. WordNet, a lexical database has been used to provide lexical semantics for the words occurring in parsed sentences [10]. This system employs ChatScript (CS) which includes WordNet along with many predefined concepts, which allows easily handling synonymous expressions. Conversely, the definitions for the technical terminologies the users must be used hand-coded (a written program).

3.3. Dialogue Management

The dialogue manager is the heart of HESAS because it coordinates the activity of all components, controlling the dialogue flow, and communicates with the knowledge base. The dialogue manager controls the communication between modules/components. Modules/components are not able to pass messages directly to each other, for design as well as technical reasons. The design of the system is such that the dialogue manager is the mediator of all communication between system modules, and in this way is able to control all messages passing and thus the order of module execution. Algorithm 4 shows how the HESAS dialogue manager works. The algorithm keeps a queue of conversation acts it needs to generate. Acts are added to the queue based on grounding and obligations of the system.

Algorithm 4: A Dialogue Management Algorithm

The DM requires extracting the required information from the knowledge base (KB), that is, database (DB). The knowledge base is a database that stores advising information. Generally, the KB has information related to curricula and extra-circular issues. It consists of information related to curriculum for department of computer science, legislations of Ethiopian public universities, higher education glossary, introduction, default answers (no response matched) and information about HESAS.

3.4. Response Generation

The Response Generator (RG) is the final component of HESAS, which constructs the message or output to be given to the user. The output or response to be given by the user is given by the dialogue manager. The dialogue manager in turn extracts the response from the knowledge base of HESAS and sends it to RG. The job of RG is responding answers to the user.

3.5. Advisement Service

The Advisement Service component is concerned with registration, retrieving and organizing of the students’ information. Basically, Advisement Service delivers two services. These are student transcript and registration information and course enrollment service. With this advisement service as the main feature, the author referred to this system as the Advisement Service (AS) system.

This section presents the implementation of the components and the experimentation aspects of the proposed system, HESAS.

4.1. Implementation

For developing the prototype of the proposed system, the author used programming languages such as PHP, HTML, JavaScript, and CSS. Development tools such as XAMPP, and ChatScript. Figure 2 shows the conversation between student and HESAS.

4.2. Evaluation Result

The experimentation of HESAS has been done into two phases. During phase 1, the author analyzed the user log files and the data shows that, as usual in electronic text communications, students prefer to ask questions using the minimum amount of words possible, instead of through a Standard English sentence. Therefore, HESAS should include methods to allow this user preference, while concurrently allowing students to use complete expressions. Phase 2 introduced the system performance measures such as precision, recall and f1-measures to evaluate HESAS. Finally, obtained the results of precision 82%, recall 84%, and f1-measures 83%. The results show that the system performance metrics are all close to 83%. These results are estimated minimums; given the error is an estimated maximum.

The fundamental objective of this research work is to provide students with an automated online advising experience that is as close as possible to traditional human interaction. In this research work, a model for dialogue system for advising higher education students has proposed and called as HESAS. HESAS is designed based on template based approach. It composed of six main components. These are user interface, spelling corrector, natural Language understanding, dialogue manager, advisement service, and response generator. The implementation of HESAS is done with ChatScript, PHP, JavaScript and CSS languages.

The system offers higher education students an instrument to potentially increase student integration, retention, satisfaction and performance, by providing additional advising services without the need for additional personnel. This achievement is possible by allowing academic advisors to allocate resources to tasks typically classified as developmental and educative advising, instead of repetitive time-intensive tasks related to prescriptive advising.

Acknowledgments

The author would like to thank Yaregal Assabie(PhD), Associate professor of Computer Science department, Addis Ababa University, Ethiopia.

Author's Contribution

All this work is done by the author of this research work including literature review, proposed system design and implementation of the system.

Funding

Open Access funding

Availability of data and materials

The developed system and knowledge base of the system available from the corresponding author on reasonable request.

Ethics approval

Ethical approval was obtained from Addis Ababa University Ethical Review Board, Ethiopia (032/2011). All methods were carried out in accordance with relevant guidelines and regulations.

Consent for publication

Not applicable

Competing Interest

The author declare no competing interest

V. N. Gordon, W. R. Habley, T. J. Grites, and Associates, Academic Advising: A Comprehensive Handbook, National Academic Advising Association, 2nd ed. San Francisco: Jossey-Bass, 2008.
M. J. Leonard, “Advising Delivery: Using Technology”, in Academic Advising: A Comprehensive Handbook, National Academic Advising Association, 2nd ed. San Francisco: Jossey-Bass, pp.292-306, 2008.
National Academic Advising Association, “Advising Technology Innovation Awards”, retrieved from https://www.nacada.ksu.edu/programs/Awards/EPublication.htm, Last accessed on June 2021.
H. S. Hart, R. B. Hussey, M. J. Leonard, J. Levin, and S. M. Winck, “eLion: Penn State’s Comprehensive Web-Based Academic Advising System”, in The Mentor: An Academic Advising Journal, retrieved from http://dus.psu.edu/mentor/bookstore/elion-advising-system, Last accessed July 2021.
Academic Advising and Career Center, “eAdvising”, University of Michigan-Flint, Flint, Michigan, retrieved from http://www.umflint.edu/advising/eadvising.htm, Last accessed on July 2021.
C. M. Leung, E. Y. M. Tsang, F. S. S. Lam and D. P. C. Wai, “Intelligent Counseling System: A 24 x 7 Academic Advisor,” EDUCAUSE Quart. 33, no. 4, 2010, retrieved fromhttp://www.educause.edu/edUCaUSe+Quarterly/edUCaUSeQuarterlymagazineVolum/intelligentCounselingSystema24/219101, Last accessed on June 30, 2021.
E. M. Latorre-Navarro and J. G. Harris, “An Intelligent Natural Language Conversational System For Academic Advising”, International Journal of Advanced Computer Science and Applications (IJACSA), Vol. 6, No. 1, 2015.
F. Albalooshi1 and S. Shatnawi, “HE-Advisor: A Multidisciplinary Web-Based Higher Education Advisory System”, Global Journal of Computer Science and Technology, Vol. 10, Issue 7, Ver. 1.0, September 2010.
Feghali, T., Zbib, I., and Hallal, S., “A Web-based Decision Support Tool for Academic Advising”, Journal of Education Technology and Society, Vol. 14, No.1, 2011, p.82-84.
Allen, J., Ferguson, G., and Stent, A., An architecture for more realistic conversational systems. In IUI ’01: Proceedings of the 6th international conference on Intelligent user interfaces, Santa Fe, New Mexico, 2001, United States, pp. 1– 8. ACM Press.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Artificial Intelligence-Based Higher Education Students Advising System: The Case of Ethiopia

Status:

Version 1

Abstract

Figures

1. Introduction

2. Related Work

2.1. Expert-based Academic Advising Dialogue Systems

2.2. Web-Based Higher Education Advising Systems

3. Overview Of The Proposed System (Hesas)

3.1. Spelling Correction

3.2. Natural Language Understanding

3.2.1. Syntactic Analysis

3.2.2. Discourse Analysis

3.2.3. Semantic Analysis

3.3. Dialogue Management

3.4. Response Generation

3.5. Advisement Service

4. Experiment

4.1. Implementation

4.2. Evaluation Result

5. Conclusion

Declarations

References

Additional Declarations

Status:

Version 1