Yuanbo's Project2

Project: Visualization Chinese Poetry

In Class Activity Link

Evaluation Results


Goal:

We want to create a new and unique display environment for literature researchers to understand and read Chinese Poetries.


Background:

Many Chinese poetries are composed in specific environment and express strong emotions. When viewed on screens, it is very hard for researchers to fully grasp the meanings and emotions behind the works. My work places the poetries in an native environment where they are created and help researchers to better grasp the ideas behind. 


My work group peotries based on their similarity in emotions they express.


In Class Activity:

Students can will read poetries in the native environment they are created in VR.


Collaborative Functionalities:

We allow multiple users in a Unity App reading the same poem


Data Format:

The data format should be a simple text file at first, then we will use NLP to do certain encodings and sorting. 


Evaluations:

1)Can our system better help researchers to understand the poetry?

2)Can reviewers find something new about these poetries when viewing in VR?


Documentations:


Project Plan:

Prepare Poetry Dataset

Use Bert to understand Poetry emotions

Create Similarity Matrix for poetries

Use Bert to understand Poetry emotions

Create Similarity Matrix for poetries

Visualize results in Unity VR

Visualize results in Unity VR

Work on environment settings of VR

Ready for in class activities

write wiki contribution

write wiki contribution


Results:

Created a fully functional VR app that can be downloaded here

While this is a preview of the work, please consult this link for full discussion of the methods and results. 

Survey and Evaluation:

Correctness

The participants (13 in total) are asked to read works of the two poets for 10mins on a screen monitor and then read for another 10 mins in VRPoetry. After that, they are given two pieces of poetries by the two poets that doesn't present in both the on screen dataset and VRPoetry. And the task is to identify the author of the two pieces of the text.

Our result shows that about 70% of the participants can successfully make the correct choice in the short 20mins. This is an astonishing number, as such a task normally requires Chinese high school students' hundreds of hours training, while most of our participants have little experience reading poems in Chinese before.

However, it must be taken into account that our participants read the poems in English translated version, while Chinese students read the original text. It might be the fact that the translated version is more direct in expression and use simple languages while the original texts often use certain unusual words.

Preference

On the other hand, we found that about 64% of participants prefer reading in VR than on screen. While this number indicates a general favor attitude towards VR reading, it shows that VR reading is not a widely chosen option.


Problems with VR Reading - Can be solved

We also track participants' feedback on problems with VR reading. Some of them can be solved by implementing better reading designs. These includes:
1) Make the board closer to the reader

2)Provide possible interactivity with the text

3)Provide strong contrast between text color and background


Problems with VR Reading - Cannot be solved

These problems are generally hardware problems, or we call "Oculus problems". These includes:
1) Increasing pixel values. The texts are not clear enough

2)The headset is too heavy to wear for long time reading