Firat Batmaz, Chris Hinde Computer Science Loughborough University A Diagram Drawing Tool For Semi–Automatic Assessment Of Conceptual Database Diagrams

Why Diagrams ? Many Diagram Based Questions in HE in Computer science, Mechanical Engineering, Systems Engineering etc... Data Flow Diagrams, Class Diagrams, ER Diagrams

Why Conceptual Database Diagrams? Namely “Entity Relationship Diagrams” Common Extensive practice Range: Simple – Complex Open-ended domain No single best solution

Why Semi–Automatic Assessment? Computer Aided Assessment Coursework and Examination Self-assessment and revision support Enhanced learning Fully–Automatic Assessment (So Far) Not so many studies in automatic diagram assessment. Focus on Grading. "Free" diagram assessment has not been successfully achieved.

Why Semi–Automatic Assessment ? Form the foundation for fully automated assessment in future. The research results have some immediate practical uses.

So, The aim is To computerise the existing manual assessment as much as possible.

The existing manual assessment Marking Slow start Speeds up Repetition of judgements. There are a number of identical diagram components in student solutions.

Identical diagram components in Student Solutions Book Title has Book Copy Reserve Book Title Loan Member Book Copy Student A has Student B Student C Reserve Book Title Member Buy Librarian has Publisher

The Semi–Automatic Assessment System Groups identical segments of the student’s diagrams. Asks the assessor to approve the correctness of a diagram fragment from the each of the different groups. Therefore the assessor would be involved in the marking process only for the number of diagram groups rather than the total number of student diagrams.

Diagram Groups in Student Solutions Book Title has Book Copy has Reservation has Student Book Title Reserve Book Title Loan Student Book Copy Group 1Group 2Group 3Group …

The Diagram Groups The correctness of the grouping depends on the criteria used to match the diagram pieces. entity relationship The smallest diagram piece in each group can be either an entity or a relationship for a conceptual database diagram.

Entity Matching Entities in different diagrams could be considered as matched exactly if they have same entity name and same number of attributes with same name. Pretty tight definition Increases assessor’s involvement Not Sufficient

Entity Matching An Entity From Diagram A Book ID An Entity From Diagram B Book ID Entities are matched.

Entity Matching Book TitleMember Reserve Book Loan Has ID BookMember Reserve Book Copy Loan Has ID Diagram A Diagram B But the entity matching is not valid. Book TitleMember Reserve Book Copy LoanHas ISBN Author Title Copy No Price Purchase Date Ideal Diagram

Entity Matching The definition should also include contextual attributes of an entity. In the KERMIT approach contextual meaning of an entity is given explicitly by forcing the students to highlight the related text in the scenarios.

Entity Matching with Scenario References Book TitleMember Reserve Book Loan Has ID Diagram A Book TitleMember Reserve Book Copy LoanHas ISBN Author Title Copy No Price Purchase Date Ideal Diagram Each book copy has got a unique copy number and their price and purchase date is recorded. --- Scenario

The problems of the scenario referencing Finding a related text to diagram components is not a straightforward task. Direct correspondence sometimes doesn’t exist. Designing a conceptual database model is an iterative process.

Designing A Conceptual Database Model Problem Statements Model (e.g. ERM) Real world view Analyse Describe verify Map verify Teaching And Assessing

The links to the scenario text The initial diagram can have direct link to the scenario text. Intermediate and final diagrams The diagram is modified by applying design rules and constraints. The final diagram can have indirect links to the scenario text requires all the steps between the initial and the final diagram to be seen explicitly.

The links to the scenario text Lecturer Department Has HOD Has Initial DiagramFinal Diagram Manage Staff Department Has Lecturer Head of department DB Scenario

The links to the scenario text Lecturer HOD Initial DiagramFinal Diagram Staff DB Scenario Direct linksIndirect links Merge

The links to the scenario text Lecturer HOD Initial Diagram Final Diagram Staff Direct referenced (DR) components Merge Indirect referenced (IR) components

Intermediate diagrams necessary to be able to match diagram components. Not all intermediate diagrams are important. Initially “Price” is an attribute of an existing entity “Book Title” later on, it becomes an attribute of an entity “Book Copy”. We only consider the previous diagrams leading to the IR-component.

Diagram Modifications which result in IR-components Merging two entities Splitting an entity into two entities Converting a “many to many” Relationship to an Entity

Diagram Modifications Intermediate Diagram Final Diagram Merge Split Convert Indirect referenced (IR) components Events

Self-explanation Student actions must be interpreted to be able to recognise an event. Even then, the interpretation may not be what the student intended. The student needs to explicitly mention their intention during the design.

Self-explanation Self-explanation (SE) is a very effective learning strategy resulting in deep knowledge. The main problem of self-explanation whilst solving the problem is the high cognitive load. The proposed diagram editor is designed to reduce the cognitive load of self-explanation.

Diagram Editor The prototype diagram editor is based on automatic graph drawing. Students enter the component type and name the tool draws the diagram. Students then can focus more on designing than drawing.

Diagram Editor Diagram Modification Section Diagram Display Section Scenario Text Section

Function Button Example Split Event Intermediate diagram "Split" function button box

Function Button Example After Split Event

An Experiment is Required The diagram editor has two aspects. To capture contextual meaning of diagram components to help the examiner during marking. To provide an environment for the student to enter their design. The editor has a very different environment from those of traditional diagram drawing tools.

Experiment People who have studied database design at university level. An introduction session: How to use the editor on one example database scenario. The example scenario uses one of the function buttons. The users are asked to design a conceptual database diagram for a similar scenario.

Results(1) All the participants managed to draw the correct diagram. Although the given scenario didn’t allow them to design the diagram without using the function button, none of them failed to use the editor.

Results(2) The required function button for the design expects an entity name from the user. All participants named the entity differently as expected. Different names for the same entity are not a problem for our approach since contextual information of the component is the main criteria for the entity match and this context is provided by the usage and recording of the function button.

Conclusion This paper proposes a new diagram editor which alters the traditional diagram drawing in order to make the assessment process suitable for semi-automation. This alteration enables the assessor to better understand the student thinking and give accurate feedback to students. The prototype editor provides an environment in which the students can design the database model methodically.

Further works Types of user and scenario are the main factors which could affect the experiment’s results. The users could be students who are learning about conceptual database design, rather than experienced designers. The interface needs to be made more user-friendly. The scenario could be written in such a way that it enforces the use of different combinations of the function buttons. All function buttons for different scenario types will be implemented.

Further works The other part of our semi–automatic assessment is the marker environment. The editor is a beneficial tool only if the contextual information of each component can be used by the marker environment to match them correctly. The implementation of the marker environment and experiments on it are also very important to complete the research.

Questions Please ? Thank you