Literature Visualization Tool

A visualization tool that allow users to organize 1591 of papers by score, frequency, and year. The interactive visualizations makes literature review much more efficient due to its many functionalities.

https://observablehq.com/d/94783b7756ead148@3099

Getting Started

View this notebook in your browser by running a web server in this folder. For example:

npx http-server

Or, use the Observable Runtime to import this module directly into your application. To npm install:

npm install @observablehq/runtime@4
npm install https://api.observablehq.com/d/94783b7756ead148@3099.tgz?v=3

Then, import your notebook and the runtime as:

import {Runtime, Inspector} from "@observablehq/runtime";
import define from "94783b7756ead148";

To log the value of the cell named “foo”:

const runtime = new Runtime();
const main = runtime.module(define);
main.value("foo").then(value => console.log(value));

Key Features

• Data Cleaning - Used Python for data preprocessing on Google Colaboratory (preprocess_dataset_shao.ipynb)
  • Inner join three different datasets and remove data rows with null value
  • Extract data from .json file and reorganize it
• Literature Review
  • Easy to know the performance of thousands of AI papers
• Research Trend - Easy to know the research topics in each year
• Zoom in and out to prevent data overlap
• Filters for folds, tasks, metrics, datasets, models and can be applied based on user preferences
  • There are 11 folds, 56 tasks, 88 metrics, 245 datasets, 393 models
• Mouse Hover
  • Display paper details by hovering over the data
• Y scale varies with filtered data for a better data visualization

Results