A Leap in Natural Language Processing
Glenn, or General Language Evaluation Neural Network, is a deep neural network that can perform summaries and analyses of large texts such as literature or scholarly articles. Glenn can detect in a text what is important and what is incidental information and reduce a text accordingly. Also using this understanding of relative importance, Glenn can produce new text that focuses on these important topics, based on a mixture of data from both the internet and the source text itself. Glenn displays a gigantic leap forward in Natural Language Processing by demonstrating high levels of reading comprehension; it will allow users to process huge amounts of text to gauge sentiments, recurring themes and common complaints, and then automatically produce reports based on that data. The applications of our AI are limitless: judging customer satisfaction, drafting meta-analyses of entire fields of study, and even just writing a book report, all these and other processes can either be partially or fully automated with Glenn.
How it Works
Previous NLP AIs have had multiple issues while still achieving impressive results; although they could, on paper, seem to be achieving on par or above humans in reading comprehension tests, various problems have been found with how they came to their answers, which meant that they were more influenced by how a certain answer was written than what the actual meaning of that answer was. In addition, they could only analyse one or two sentences in a row. Answering questions about a larger piece of text was not possible. Working from the findings of Google’s T5, we then created a model for the structure of larger texts, and then trained our neural network to compare these texts to summarised versions that were readily available (through study guides, articles on Wikipedia etc.). Over the course of several months, and a lot of supervised training, the neural network was able to predict how a similar process of summarisation and highlighting key text could occur with new material.
Our Implementation - The GlennStudy App
A Powerful Tool for Students
GlennStudy is an implementation of our neural network and it is an ideal application for the Glenn. The application scans, summarises and breaks down a text on the fly, allowing a student to quickly study a text they cannot find a guide for online, or a section of a text that is not mentioned elsewhere. It doesn’t matter if the text is scientific or prosaic in nature; the software adapts to both and can make appropriate learning content, identifying plot points in literature and key equations in physics books, for example. On top of this, GlennStudy also tracks each user’s progress through various topics, allowing the service to recommend related content to new users and auto-generating “courses” for them to indulge in their curiosities.
Main Features
Courses & Tracking Progress
As GlennStudy is used, it builds up a profile of each user’s learning habits. This then allows GlennStudy to recommend other topics within its own library of information that are relevant to what the user is already studying. The data generated, in turn, can tell us if a wide range of users are having trouble with the same text, or if people are having trouble understanding something that is supposed to be at their reading level. These are just a few examples about how GlennStudy, a single implementation of Glenn, can provide invaluable data to educators.
Markup Parsing Within NLP
Our AI also features a LaTeX parser, which allows it to reproduce equations in beautifully formatted LaTeX. For GlennStudy, this will increase readability for science and mathematics related texts, and many other NLP algorithms can only output in plain text. It would be possible to include other mark-up languages into Glenn, such as HTML or Markdown, for other projects.
Revenue Streams
We can offer companies the service to implement Glenn into their own systems, replacing the existing NLP algorithms they are using and fine tuning our solution to produce the exact results they need. We can also consult companies about NLP, whether they have implemented Glenn or not. Finally, as Glenn is used in various applications that are developed by us, such as GlennStudy, we will have a wealth of data that would not be available from anywhere else. Companies can share the data we are collecting for a fee, allowing for further development of our wholly owned applications of the Glenn AI.
© Copyright 2020, All Rights Reserved