The first expert.ai Natural Language & Text Analytics API Hackathon is officially in the books. After our panel of judges reviewed more than 50 unique and innovative natural language-based applications, they found that one stood above the rest: The Peer Reviewers App.
The app was developed by 29-year-old software developer Peter Asamoah and 21-year-old software developer Timothy Yirenkyi who are based out of Maidstone, United Kingdom and Kumasi, Ghana, respectively. Though neither had any prior experience with natural language technology, they entered the hackathon with the intention of experimenting with it and expanding their artificial intelligence foundation.
The Story Behind the Peer Reviewers App
The Peer Reviewers App is a web application that uses AI knowledge extraction to perform document similarity with the intent of recommending peer reviewers for academic and scholarly journals and articles. The goal of this application was to demonstrate how the peer reviewer selection process can be automated accurately by leveraging the knowledge extraction and document similarity capabilities of the NL API.
This project was inspired by a previous one in which they used a “Bag-of-Words” model to search for matching documents. This competition gave them an opportunity to experiment with the expert.ai NL API and evaluate how its knowledge extraction feature compared to their former approach of extracting relevant words based on their volume of occurrences in a document.
Not only did knowledge extraction prove more efficient to their application, but it increased correctness which, in turn, enhanced query accuracy and improved search results.
How They Built It
Timothy and Peter delegated their duties according to their individual strengths. Timothy is a front end and mobile developer who mostly uses JavaScript and TypeScript, so he chose to build the front end of the app using the Angular framework with most of the development done via TypeScript. Peter is a back end developer that uses mostly .NET framework languages such as C# which is what he used to build the back end.
To leverage the NL API, the duo opted for the REST approach, which involved building a simple client to communicate with the API. From there, they used the API for key phrase extraction and knowledge extraction of topics discussed in the parsed content. In addition to the extraction capabilities, they found the scores returned as part of the response to be particularly helpful in ranking the topics and key phrases by relevance and importance.
This not only enabled them to accurately classify key phrases and topics of submitted articles, but further refine them into “related topics” to add more depth and specificity by which reviewers can be matched.
Future Plans for the NL API
After successfully leveraging the NL API in their Peer Reviewers App, Peter and Timothy have plans to experiment with another NLU capability in a separate, upcoming project. They plan to use the API’s sentiment analysis feature to analyze user posts and comments in hopes of gaining deeper user insight and using it to facilitate the decision-making process for businesses.
Given their impressive use of the NL API their first time around, we are very excited to see what Peter and Timothy come up with next! Until then, congratulations on your inspiring work with the NL API. You clearly have what it takes to become a natural language expert.
