Design principles for managing cognitive overload in interactive analysis of corpus data with visualisation

Abstract

The growth and usage of data in this age are no longer like sparks flickering off tinder. Their flames have grown into a wildfire and there is demand to successfully tame them into mediums, so that large datasets can be analysed, especially in fields containing large sets of text-based data. This thesis has concentrated specially on visualising corpus data, especially linguistic, social media and online review data, and formulating design principles to make them successful, so that high element interactivity calculations can be automated when it is preferable to view the datasets at a higher level. Linguistic corpus data may be used by the system to analyse or describe language elements, such as phonetics and phonology. Social media data may be used by the system to identify influential users, the flow of data, and the history of the posters. Online review data may be used for market research, to identify problems and complaints, and to identify who the posters are, to ultimately decide if the posts should be considered when making business decisions. There already exist several visualisation techniques, models, and frameworks that have tackled these above issues, however, they have not been fine tuned to fit the needs of specific users, such as those using text-based corpora data to conduct linguistic research, understand social media, or analyse business reviews. There is also a lack of research work investigating the cause of cognitive load "" the amount of mental strength needed to perform a task "" that the user may experience while using a system and ways the load can be reduced. In this thesis, existing visualisation techniques were reviewed, anticipated visualisation attributes were identified, layouts for the visualisation were implemented and, in the linguistic visualisation, investigations of the system's ability to reduce cognitive overload were conducted and evaluated. This thesis has presented a new framework that connects unique attributes of the data and the problem domain; specific representation of data; and machine learning and decision making algorithms to generate efficient data visualisations and overall better analytical experience for the user.

Date of Award	2019
Original language	English