COVID-19 Search

Author: Da Yin

This project is a front-end website for a search engine designed for COVID-19 related information. Click here to visit the website.

Main Features

• Instant Search: Any typings will receive response immediately.
• Advance Search: You may customize the search by changing the page size or sorting metric.
• Event exploration: The related events are calculated by k-nearest neighbors on pre-trained graph embeddings.
• Region Search: Searching for locations such as Beijing or China will get the infection data.
• Entity Search: Searching for entities such as Coronavirus will get the entity knowledge.

Techniques

Front-end website is bootstrapped with React Create App written in mainly Typescript. The search engine is largely implemented with a simple text search on MongoDB. The segmentation on Chinese texts uses jieba.

The back-end services is mainly provided by AMiner COVID-19 including the infection data and the news data. The entity search dynamically calls the API from XLore.

The related events are calculated using k-nearest neighbor based on graph embeddings generated by ProNE [1]. The graph is constructed by connected news and segmented words.

[1] Zhang, Jie et al. “ProNE: Fast and Scalable Network Representation Learning.” IJCAI (2019).