Work

Method of Modeling Questions for Automated Grading of Students' Responses in E-Learning Systems

Year: 2014
Type: article
Abstract: Introduction. Problem relevance. The capability to check a solution of practical problems automatically is an important functionality of any learning management system (LMS). Complex types of questions, implying creative approach to problem solving are of particular interest. There are a lot of studies presenting automated scoring algorithms of students' answers, such as mathematical expressions, graphs, molecules, etc. However, the most common types of problems in the open LMS that are being actively implemented in Russian and foreign universities (Moodle, Sakai, Ilias etc.) remain simple types of questions such as, for example, multiple choice. Study subject and goal. The purpose of study is to create a method that allows integrating arbitrary algorithms of answer scoring into any existing LMS, as well as its practical implementation in the form of an independent software module, which will handle questions in LMS. Method. The model for objects of type "algorithmic question" is considered. A unified format for storing objects of this type, allowing keeping their state, is developed. The algorithm is a set of variables, which defines the responses versus input data (or vice versa). Basis variables (input) are selected pseudo-randomly from a predetermined range, and based on these values resulting variables (responses) are calculated. This approach allows us to synthesize variations of the same question. State of the question is saved by means of "seed" of pseudo-random number generator. A set of algorithmic problems was used to build the lifecycle management functions, namely: initialization create (), rendering render (), and evaluation answer (). These functions lay the foundation for the Application Program Interface (API) and allow us to control software module responsible for the questions in LMS. Practical results. This study is completed with the implementation of software module responsible for mapping the interaction with the student and automated scoring of algorithmic questions. The module supports both standalone mode using the test pages, and the mode of integration with any LMS. Special attention is given to fault tolerance of the module by allowing a continuous state preservation of algorithmic questions in the database. The developed module architecture enables the integration of any types of input and scoring algorithms in LMS. An advantage of the suggested approach is the capability to generate a number of question options, which is necessary to form the homework or tests to be given to groups of students. Application. The results can be used to extend the functionality of e-learning systems for the inclusion of complex question types that support automated scoring of student answers. Further studies. Further research based on this method may include the integration of new student’s answer input methods and their scoring algorithms, as well as creating other algorithms for the system feedback loop implemented in answer () function. For example, one of the urgent problems consists in generation of hints based on the analysis of students’ wrong responses.
Language: English
Open Access status: diamond
OpenAlex ID: W4239791912
FWCI:
Cites:
Cited by:
Related to: 10