The companies combine for a ground-breaking application of NLP technology.
3/24/22 Update: Today, Summari earned the top spot on Product Hunt, demonstrating the power of the Primer Platform.
Summari is an AI-powered assistant that instantly summarizes things you read, simplifying the consumption of enormous amounts of publicly available information. The service is aimed at content teams, analysts, investors, and students who need to read and comprehend critical information in order to make recommendations and decisions that shape businesses and public institutions. Summari’s goal is to make this service available to anyone who needs a rich and relevant summary, instantly.
Reaching human limits and the search for an NLP partner
Before working with Primer, Summari hired college students from top universities to summarize articles manually. The summaries were high quality, but humans can only consume a finite amount of information at a time and hand writing summaries is time intensive. The latency diminished user experience. The upside is that this method resulted in a human-made dataset of the highest quality summaries with which to train computers to replicate.
Summari began investigating NLP as a solution to speed the production of human-quality summaries. CEO Ed Shrager attempted to optimize a large open source NLP library, but wasn’t able to fine tune the model to achieve results that Summari wanted. Summari was aiming for a precise format with a high-level abstract and several context-setting introductions for each section of an article, as well as bullet points explaining each section.
Primer enters the picture to build a summarization model
In November 2021, Summari selected Primer as its NLP partner of choice. Primer’s forward-deployed engineering team used Primer’s Platform to train and deploy a customized text-to-text summarization model, coupled with Primer’s data ingestion pipeline, that delivers human-level summarization capabilities for any long-form article on the internet. The Primer Platform makes it simple to ingest and normalize data from almost any data source – web data from any url, applications like Salesforce, databases and data stores, PDFs, and others – connect the data to a customized NLP model, and serve it at scale.
Primer made use use of the existing human written summaries that Summari had collected as training data for the custom summarization model. The Primer Platform was able to unify the human written summaries even though they were written by different Summari annotator.
The resulting custom summarization engine was configured to allow Summari to change key parameters, such as length, style and content. This essentially gave them control to “drive” the summarization engine for their own needs – resulting in better results for their users.
The results
Primer delivered a summarization model for Summari that achieves comprehensive, human-quality summaries of any long-form article instantly. Faster delivery times allow Summari to expand their offerings from a few dozen publications to the entire Internet. Customer experience has been dramatically improved. Their customers can take any article on the Internet, drop it into the Summari application — which is running a custom Primer text-to-text NLP model under the hood — and generate a quality summarization in moments.
The market’s reception has been strong, earning Summari the top spot on Product Hunt.
It’s not just a breakthrough for Summari, but for NLP technology itself. An instant summary, in human quality, of a massive amount of information is another display of the awesome power of NLP.