Develop/Update Web Interface for Natural-Language Processing Extracted Interaction Network

[cannin]

Background

This builds on the work of from #109 and found here: https://github.com/cannin/ihop-reach, where the Pathway Commons team has an extracted network of biological interactions from scientific literature (http://reach.nrnb-docker.ucsd.edu/) and a prototype interface has been developed.

Goal

The goal is to expand on the web interface and the data generation pipeline developed in 2019. Specific items can include, but are not limited to:

Frontend

• Pagination of results
• The search interface can be sluggish; improve the search interface
• Convert the existing Tampermonkey script (created as part of link outs from the site) into a Chrome plugin: https://gist.github.com/cannin/4a30c1e3551b83f059c2206e491a082a

Backend

• Incorporation of additional datasets to all users to parse the interactions better. For example, importance characteristics of the journal or citation counts of the given publications (e.g., annotating articles with journal rank from https://www.scimagojr.com/journalrank.php)
• Update the data generation pipeline software to make updating the dataset more maintainable
• Develop Docker containers for any additional datasets that are to be included to help maintain them.

IMPORTANT: A good proposal would highlight two possible expansions one front-end change using Javascript (e.g., search interface optimization) and one back-end that incorporates additional data using Python (e.g., extraction of journal ranks for articles and integrating into the database)

Difficulty Level

Difficulty level 1. Skills necessary for this project are widely applicable to many other projects.

Skills

• Python
• Javascript

Public Repository

https://github.com/cannin/ihop-reach

Potential Mentors

• Augustin Luna (@cannin)
• Benjamin Gyori (@bgyori)

Contact

Augustin Luna (aluna AT jimmy.harvard.edu)

[DipanshiB]

I would like to take this up if its open for contribution at the moment? I have worked with GraphQL, Node.js applications with MongoDB database. @cannin

[cannin]

Sorry for the delay. @DipanshiB if still interested send me a message at the email in the project description, let me know a bit about the prior work you've done, your programming background, and give me a quick idea of how you might fix the search interface to make it faster.

[PritiShaw]

Hi, I am interested in this project. I am confident in ReactJS, Python, Flask, Node, Express, MongoDB. I am well aware of linux environment and I also have certification in basic Google Cloud Platform (from Coursera). I went through the work done last year. I can see that, for search, lunrjs has been used. Although I am not aware of the requirements of the project like which details are important, I think we can improve the search output by displaying vital information and only display those names which have the searched term as substring ( might fireback if there is a typo in the search key) On the first load, search takes time, I suspect the size of the search index to be the root cause of the problem. As it is mentioned in the site that the application has over 3million sentences, I think using API based search can help us improve the search experience. Are there any competency challenge tasks prior to the application period? Can I start developing the Chrome extension?

[PritiShaw]

Are there any competency challenge tasks prior to the application period? Can I start developing the Chrome extension?

I made the basic Chrome extension using the provided Gist. Requesting code review for my work available at https://github.com/PritiShaw/ihop-reach/tree/Chrome-ext-Highlight-PMC

I tested in Chrome, Chromium on Ubuntu19 and Chrome on Windows 10(results below)

[cannin]

@PritiShaw that's really nice! I updated the project description to give a better idea of what should go into a successful proposal.

[cannin]

@PritiShaw if still interested in the project please send me a quick email and i'll make some additional comments

[khanspers]

Active project for GSoC 2020, closing here.