This article deals with the problem of the developing and implementation of the tests and practical tasks generator in the distance learning systems. The object model of the tests and exercises generator is developed based on the mathematical model.
The software module that support practical tasks and academic performance ratings
during the learning process are the most important in distance learning systems (DLS)
[
The practical tasks and tests development is rather hard and tedious process of
tasks (questions) creation, possible answers defining, and correct answers detection
[
The good example of the software implementation for this type of tasks is
Generator of examination tests in software environment «Examination tests delivery
system» [
While DLS testing module engineering, the process of testing is usually understood as a set of statements calls with a certain behavior in the context of a learning model. The process of the automatic test compiling in DLS testing module is a work flow in which a certain logics with the possibility of automation is implemented. It is important to start with a determination of an abstract model, that contains all the necessary data describing the learning aim and explaining the correlation between objects of the specific instances of this model, to automate the compiling of tests’ questions as a work flow.
Let’s say smart-test is the instance of the test, generated by the given template with the automatic test generation process that is described above.
The automatic generation of a smart-test will help to substantially simplify the practical tasks compilation in the mentioned above sciences, reducing required time and guaranteeing the unique character of each task generated, as well as it gives the ability to automate the tests results evaluation without negative human factor effect.
In this work we consider the problem of creation of a smart-test generator abstract
model and tasks for solving the most of the practical tasks of the applied disciplines,
and the implementation work of the smart-tests generator in DLS. The abstract
generator model is a generalized model of mathematic tests [
To build the P – abstract model of the learning task – we use its mathematical
definition [
P = <M, Ф, Q>, where
M – a set of the task input parameters, and is described as М(x1, …, xn), where x1, …, xn – the task parameters;
Ф – a set of the task conditions, described as Ф(φ1, …, φk) , where φ1, …, φk – task conditions;
Q – the task result model, described as Q(q1, …, qm), where q1, …, qm – sets of correlations, defining the results of the task.
This mathematic model is sufficient for further abstractions building and objectrealization with the high-level programming languages.
Let’s agree with the following terms to describe the mathematic model above in a programming language:
P = <M, Ф, Q> – LrnTaskTemplate (abbreviated from Learning Task Template); M – LrnTaskModel (abbreviated from Learning Task Model); Ф – LrnTaskValidator (abbreviated from Learning Task Validator);
Q – TResponse, that is the element of the software module «Solver» LrnTaskSolver<TModel, TResponse>, that will be described below.
The interrelations scheme of the objects described above is introduced at fig.1.
On this scheme ISmartTestModule is a template library of the learning tasks generators for smart-tests compilation.
A template library of the learning tasks generator module is a program class (LrnTaskTemplate), which contains all the functions necessary for building the learning task instances, their correction, and solvings with the use of named model.
Task generator template functions use general (TestSettings) and local
(LrnTaskTemplateSettings) settings, given by a user (test compiler) as input
parameters. These settings may provide the template with generation and validation
parameters of the specified learning tasks as: instruction to use preferable types of
questions [
Learning task model (LrnTaskModel) is the program class that aim to store the input parameters of the learning task. Depending on the specified aims, the learning task model may contain advanced parameters, necessary to describe the task settings. The number of the parameters is not limited by the system.
The next learning tasks generator element is LrnTaskValidator class. The aim of this program class is validation of the model instance to correspondence to the task settings. This program class is also used in reconciliation of the generated learning task model that is described further.
A solver (LrnTaskSolver) is the final element of the template. The aim of the solver is to solve the specified task (achieve results), using the LrnTaskModel parameters with the conditions given in LrnTaskValidator. The results of the solver are the formal expressions in the signature of the certain subject field that may be defined in the LrnTaskResponseModel class inheritor. In particular, atomic expressions may be there – a number, a string, a date. 4
The model of the learning tasks generator template module is shown at fig. 2.
The list of the main functions of the tasks generator is below: 1. GenerateModel is a method of learning task model instance creation (LrnTaskModel). 2. ReconcileModel is a method of generated learning task reconciliation. This function is necessary to solve the problem, when the system generates similar tasks at the first stage of task generation in the set of the final smart-test. The correction algorithm must be implemented locally for each type of the learning tasks. The aim of the correction is to change the learning task model by making it unique comparing with other tasks in the set of the tasks in the smart-test. 3. CreateQuestion is a method of the question description creation. This method creates a program class instance that describes the question. This class model is shown at fig. 3:
Here is the description of model properties:
─ Title - title-text of the task. This property consists of the short text, that explains
students the subject of the question;
─ Description - the description of the task that will be shown to the student. An
author of the specified learning tasks module implementation should consider
possibility to create description text on one of the system languages (for example,
Russian, Ukrainian, English);
─ Tip - a hint text, that the student may use answering the question. This field is
optional and may be left empty. Usage of the tip is fixed and processed by the
system;
─ QuType - type of the question. An author of the specified learning tasks module
implementation may use one of the multiple questions types, mentioned in the
specifications [
Here is the example of learning tasks templates implementation on “Quadratic
equation” topic [
Executing the solution from the parameters of the mentioned model, this solver returns the object of the QuadraticEquationLrnTaskResponse type (fig. 9), that contains a pair of numerical values: x1, x2: 7. Template 2 solver - QuadraticEquationLrnTaskSolverU, which implements functions and properties of the abstract program class LrnTaskSolver (fig. 10). Executing the solution from the parameters of the mentioned model, this solver returns one numerical value x: 8. Test setting class - QuadraticEquationTestSettings implementing the program interface ILrnTaskTemplateSettings (fig. 11):
In this example it is supposed to specify valid values in the test settings for quadratic equation coefficients: a, b and c.
Now let us turn to the algorithm of creating smart-test and learning tasks using the object model, mentioned above.
The process of the smart-test creation begins with the connecting necessary modules to the system interface. We should mention that in this example the user has just one module: QuadraticEquationTest. The user should point in the module settings the templates to use (there are only two in this example – Template 1 and Template 2), and the settings of each of the template. The test compiler also sets number of questions: N, which should be generated. One of the parameters, that the user should specify, is the parameter of the rate set, which specifies the coefficients of usings of module templates in the smart-test.
When all the settings are made by the user, a smart-test is triggered after each request of a student in the testing group of the distance learning system. Below you can see the algorithm of the test instance generation (fig. 12).
A number of questions estimation. Assume N - a number of questions, needed to generate the test, T = <t1, …, tk, q1, …, qk, s1, …,sk> - a set of used templates from all the connected modules, where ti – the instance of template class, qi – rate of using of i-template to the all number of templates T, si – template settings, made by the test compiler. Example. Assume N = 20, T = <t1, t2, 0.3, 0.7, s1, s2>.
Then the final test will consist of 20 questions, based on the templates t1 and t2. 30% of the questions (6 questions) will be generated based on the first template. 70% of the questions (14 questions) will be generated based on the second template.
Learning tasks modules creation. Below you can see the algorithm scheme of the learning task module creation and correction (fig. 13).
As one can see from the scheme above, model correction is implemented immediately after it is generated. It allows, first of all, using rough method of the model values generation, using the system random number generator. Then in the correction function the model parameters are changed not to coincide with the other models, generated in the cycle before.
Generation of the final questions sets (TestQu). The process of the final set of questions creation, based on the already generated and adjusted ones, occurs with the same enumeration as a generation of the learning tasks models - first all the templates of the connected modules are enumerated, and then the number of this template questions is searched through.
The sub-program of the learning task model question instance creation is shown at fig. 14.
Thereby, smart-test is initializes and registers in the distance learning system, and then it becomes ready for students to pass it. 6
The structural model of the software module for the smart-tests generation and use, based on the method of automatic generation with the assigned tests or practical tasks template is build on the basis of mathematical model.
The software module classes, algorithms of learning tasks and questions models generation and verification, and smart-test answers processing are described. The described model was tried and tested on the examples of the smart-test for the exact sciences: mathematics, physics, chemistry, computer science. Also it seems possible to make smart-tests in the field of some other disciplines: some biology sections, psychology, music, sociology, etc.
This method of automatic test generation, in the long view, may be used while creation of adaptive tests with the use of smart-tests.