Discovering User Experience Variables from Textusing Computational Semantics Approaches
Nielson Norman group defined user experience as, “all aspects of the end-user’s interaction with the company, its services, and its products”. Many researchers have investigated what criteria ensures good user experience. With the vast development in information technology, we could easily access...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Language: | English |
Published: |
2015
|
Subjects: | |
Online Access: | http://ir.unimas.my/id/eprint/10764/1/Wendy%20T.pdf |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Nielson Norman group defined user experience as, “all aspects of the end-user’s interaction with
the company, its services, and its products”. Many researchers have investigated what criteria
ensures good user experience. With the vast development in information technology, we could
easily access to user generated data such as reviews that explain user experiences. However,
mountainous of reviews provide too much information and at the same time contain noises.
These motivate this study to present a novel solution to automatically analyze reviews and
predict the underlying user experiences. We believe that this solution provides insight into the
behavioral aspects of those reviews where most of the time, we cannot observed them directly.
In our study, we have proposed a Computational Model for User Experience (CompUX) that
able to predict user experiences from reviews. We have choosen five main user experiences:
Perceived Ease of Use, Perceived Usefulness, Affects towards Technology, Social Influence, and
Trust. We have created an UX semantic space to learn the semantic meaning relationship of
words and documents by incorporating the state of the art distributional semantic models: Latent
Semantic Analysis and Paragraph Vector as part of the CompUX. Next, by mapping reviews to
their semantically similar measurement items (derived from behavioral science) using the UX
semantic space, we could infer user experiences from reviews. Based on the results obtained,
the model performed better than random prediction and we were able to achieve macro average
F-Measure of 0.31. |
---|