# Presentation: A Visualization of Research Papers Based on the Topics and Citation Network [[paper link]](https://ieeexplore.ieee.org/document/7272616)
<img src="https://hackmd.io/_uploads/SJz5GGWQ6.png" alt="drawing" height="400"/>

## Problem Met in Searching for Research Papers

Through keyword searching, researchers may miss papers that are cited by several papers related to the keyword, because they did not include the particular keyword.

> Keyword search example

## Purpose of the Paper

Problems to solve:
1. When surveying papers in a new field, it is hard to know all the appropriate keywords, and therefore it is not easy for them to determine which papers they should read.

> When delving into a specialized area like Data Visualization in computer science, newcomers often face the challenge of identifying relevant literature due to unfamiliarity with the field's specific jargon and evolving terminology. For instance, a novice might initially search for general terms like "data visualization" or "computer graphics," potentially overlooking pivotal papers that use more nuanced language or specific subfield terms such as "information aesthetics," "visual analytics," "human-computer interaction," or "graphical perception." This lexical gap can lead to a skewed understanding of the field, as seminal works or cutting-edge research might be missed if they are categorized under these more specialized keywords. Thus, the initial hurdle in surveying a new field is not just finding the literature, but also learning the language that accurately describes its various niches and frontiers.

> When delving into the field of Natural Language Processing (NLP) in computer science, newcomers often face the challenge of identifying relevant literature due to the vast array of specialized terminology and evolving keywords. For instance, a beginner might be familiar with basic terms like "machine learning" or "linguistics," but might not yet be aware of more nuanced or recent concepts such as "transformer models," "BERT," "sentiment analysis," or "neural machine translation." This gap in keyword knowledge can lead to difficulties in pinpointing foundational papers or cutting-edge research. As NLP is a rapidly advancing field, new terms and technologies emerge frequently, making it even more challenging for someone new to the field to stay abreast of essential literature without a comprehensive understanding of the current lexicon.

2. Also, new research fields have triggered as fusions of multiple research fields, and researchers need to understand the relations of the multiple fields along their fusion.

Immersive Data Visualization
>The field of Immersive Data Visualization, which combines Data Visualization, Virtual Reality (VR), and Augmented Reality (AR), serves as an excellent example of a research area born from the fusion of multiple disciplines. In this field, traditional data visualization principles merge with immersive technologies to create interactive, three-dimensional representations of data.
>Researchers and practitioners venturing into Immersive Data Visualization need to be conversant not only with the fundamentals of data science and graphic design but also with the intricacies of VR and AR technologies. This includes understanding the hardware and software aspects of VR/AR, such as head-mounted displays, tracking systems, and 3D rendering techniques, as well as user experience (UX) design principles specific to immersive environments.
>Moreover, they must be aware of the cognitive and perceptual aspects of how humans interact with and interpret data in three-dimensional, immersive spaces, which is a departure from traditional 2D data visualization. This multidisciplinary nature requires a broad and often challenging synthesis of knowledge from computer science, design, psychology, and more, making it a complex field for newcomers to navigate and master.

## Related Works (Optional)
Some related studies in this field

## Methodology

#### Some definition
* $\text{Paper}$: display as node.
* $\text{citations}$: display as directed edges.

#### Paper classification

LDA (Latent Dirichlet Allocation)
* consider papers as a mixture of various topics
* find topics from the 20 most common words (excluding non-important words, such as prepositions and frequent used terms.)

#### Network layout
1. Hybrid Force-directed Graph
* displays the nodes supposing that their sizes are proportional to the number of citations.
* enables to place papers that belong to the same research category closely,
* 怎麼分群，請見 [Hybrid Space-Filling and Force-Directed Layout Method](https://ieeexplore.ieee.org/document/4906846)
2. Space-filling network layout algorithm
* avoid the node cluttering and improve the display space utilization
## UI Related
* **Network Layout**
<img src="https://hackmd.io/_uploads/SJG8WfbXa.png" alt="drawing" height="200"/>
<br/>
1. Avoid node overlapping
2. Bundle edges between clusters when count > threshold
3. Avoid edge bundles overlapping node by using Catmull-Rom spline curve
<br/>

* **Rendering and Coloring**
<img src="https://hackmd.io/_uploads/rJcD-MZQp.png" alt="drawing" height="80"/>
<br/>
1. Node sizes are proportional to the number of citations
2. Colors represent the publication years
3. Use gradient color to represent citing directions instead of arrows for readability
<br/>

* **User Interface**
<img src="https://hackmd.io/_uploads/H19ZQMZmT.png" alt="drawing" height="200"/>
> 1. search keyword highlight the papers whose titles include the keyword
> 2. click on a nodes corresponding to a particular paper, the technique displays the details of the paper such as the ACM identifier, title, authors, year, and abstract, on the panel featured by the other tab. At the same time, it highlights the edges of the clicked node, and those of the nodes that are connected to the clicked node.
> 3. When selecting a research category that the user is interested in, the node cluster that has only the research category is magnified in the center of the display.
> 4. users click a paper node, its citation edges are
highlighted. Users can follow these edges and trace them.

## Use Case
<img src="https://hackmd.io/_uploads/rkdHJMWXp.png" alt="drawing" height="250"/>
<br/>
<br/>

> ...

## result
<img src="https://hackmd.io/_uploads/rkdHJMWXp.png" alt="drawing" height="250"/>

## Evaluation

#### Comparison with time-oriented visual representation

Method: By surveying 21 graduate students majoring computer science with 9 problems.

1. The transition of papers amount published in the conference every year.
2. The main topic in the conference.
3. The trend of a research topic by year.
4. The research fields that seem to have a strong relationship with a field you focus on.
5. Much-cited papers on a certain topic.
6. The latest paper on a certain topic.
7. The content trends of papers citing the paper you read (or clicked).
8. Papers whose contents are similar to the paper you read (or clicked).
9. Papers that had a great influence on the paper you
read (or clicked).