At Blueprint, we build AI that empowers the work force. We focus on eliminating blockers in current workflows and empowering analysts to tap into the potential of Big Data and Big Compute. We design systems that allow computers and people to do what they do best. To those ends, we created the Analytics Based Intelligence (A.B.I.) application. But it wasn’t enough for us to simply revolutionize the world of data analysis; we wanted to accomplish this in three weeks.
In this article I will be covering two aspects of the A.B.I. application: how it helps analysts work effectively with Big Data, and how Blueprint leveraged its strong partnerships to deliver this state-of-art technology so quickly.
With Big Data, analysts have the potential to sift through much more data and therefore discover more impactful, granular insights. But working with Big Data requires more than the traditional analyst’s skillset.
In the era of Big Data, many occupations have seen massive increases in their potential to affect business decisions and the skills required to reach that potential, and data analysts have been no exception. An analyst’s job is to look at some type of data, whether it is columnal, text, imagery, video, or any combination thereof, and produce insightful information from it. With Big Data, analysts have the potential to sift through much more data and therefore discover more impactful, granular insights. But working with Big Data requires more than the traditional analyst’s skillset.
As part of the A.B.I. initiative, we built a product that enables an analyst to sift through large amounts of news articles and blog posts related to things, people, companies, or “locations of interest.” We designed A.B.I. to present the most recent, non-redundant, and important articles to analysts without having to read each article individually. It was also imperative that analysts be able to navigate the large amount of data in an intuitive fashion to ease the process of organizing and processing that data into insights.
So what does A.B.I. do? The backend of A.B.I. is a Natural Language Processing (NLP) engine, with various functions that come together in the UI to produce a fluent research workflow. These functions include the gathering of data, summarization of articles, topic extraction, document similarity determination, information ranking, and topic grouping. Let’s briefly cover each function and then we’ll discuss how they come together to produce a fluent research workflow. Based on a user-defined list of topics, thousands of news articles and blog posts are gathered and processed within seconds. A.B.I. then calculates the most important sentences in each document and aggregates them to form the document summarization. From each article A.B.I. then extracts all topics and ranks them in terms of importance in each document and scores how similar each document is to the others. The information ranking function then scores each document by how much different information is present in each. Finally, the topic grouping function clusters documents that are about similar topics. That’s a lot of NLP and it all happens in seconds.
How does A.B.I. pull all this data together so quickly to produce non-redundant and important results? It is crucial that analysts be able to quickly determine a document’s meaning by reading the A.B.I.-derived summarization instead of having to read an entire article. A.B.I. presents the analyst with a queue of summarizations ranked by recency and information ranking, while giving them a diverse selection of topics based on similarity. A.B.I. also presents the analyst with a list of document groupings that gives a quick view of all the information from the documents in an organized manner. In addition, A.B.I. alerts the user to the main topics of each group. It’s all about using the analyst’s time efficiently. Now the analyst can read the summaries of only a few documents to fully understand their research objective using our sorted queue and document groupings features.
With the A.B.I. project, Blueprint not only wanted to change the way analysts work with Big Data, we also challenged ourselves to build it in a short amount of time, going from idea conception to working product in only three weeks.
How did we do that? By working with our partners Microsoft and Agolo, an NLP service firm. To build a state-of-the-art product, you must have matured building blocks in place. To solve that problem, we built A.B.I. using components of Microsoft Azure, Microsoft Teams, and Agolo’s text processing services. Microsoft Azure allowed us to enable fast and stable processing, Microsoft Teams provided a collaborative UI hub for A.B.I., and by using Agolo services, A.B.I. was able to ingest highly accurate and fast document similarity values, as well as text summarization. By partnering with these two companies, Blueprint was able to focus on idea generation and component integration while avoiding the need to build components from scratch, which would require months of trial and error to perfect.
I’m excited to share the A.B.I. story with everyone, and not just because of the cool tech. I love this story because it also shows what Blueprint is all about. We don’t just take on a project from a client and deliver on their requirements, any number of companies can do that. At Blueprint we challenge ourselves to always outperform those requirements, whether they be technical or time-based, and we’re known for thinking outside the box and using our extensive partner network to deliver outstanding results.
