WebFountain (WF) was IBM’s Web-scale mining and discovery platform that extracted trends, patterns, and relationships from massive amounts of unstructured and semi-structured text. During the project, over 1 petabyte (1000 terabyte) of content was in storage with over 3 billion pages indexed, and WF was capable of mining 20 million pages a day.
Our client, FBIS, was interested in creating a data mining application for national intelligence. With content collected from all corners of the Web, how would you design a search interface that could construct complex queries while remaining easy to use? How would you design an effective information communication tool through a combination of technical search applications? Most importantly, how would you design a tool to help our National Security analysts?
As the User Interface Lead on the project, I worked closely with an HCI expert to analyze the work flow of FBIS analysts. We studied how they conducted research online, from tools used to techniques on how they scanned for information. We conducted surveys, interviews, and shadowing of users to better understand what would really enhance their output.
What we discovered was FBIS analysts constantly skimmed web pages and search results for clues. They also preferred to save results in batches first and review them later. They looked for ways to quickly move from “search” to “data extraction” to “data analysis” to “intelligence report”. They constantly shared findings with each other and used multiple search applications at any given time.
In our solution, I first made sure the foundations of a good web UI was there. Besides having an overall consistent look and feel, I looked for ways to optimize screen real estate, decrease load times, and create “skimmable” content. Complex features could be toggled open and close, and we looked for ways to add effective visual tools, turning quantity and associations of data points into substantial visual clues.