Throughout my career, I have been deeply interested in people, language, and computation. As a professor in the Department of Linguistics at the University of Texas at Austin, my research and teaching spanned linguistics, computer science and the digital humanities. Given this background, it was only a matter of time before the allure of social media became too great for me to ignore: it has language, it has people, and it has an unending set of interesting and important questions to explore by using computational ideas and means.
Shortly after I received tenure at UT Austin, Ken Cho reached out to me to discuss some of his experience working on social media management systems (SMMS’s) at his previous startup, Spredfast. An SMMS helps corporations identify and respond to compliments and complaints addressed to them on social media. Ken saw that corporate clients were doing their best to respond to their customers, but they were limited in how they could tailor their responses because they didn’t understand the wider context of their customer base and the unique individuals who comprise them.
One example of this comes from pro-bono work Ken did for an Edgar Allan Poe exhibit at the Harry Ransom Center at UT Austin. The Ransom Center had assumed the primary target audience would be women over the age of fifty; Ken instead saw that a lot of the general interest in Poe on social media came from young Goth women. By acting on these insights and expanding marketing efforts to additionally include fans of Goth bands, the Ransom Center was able to reach an interested segment that may not have otherwise known about the exhibit. This strategy drove a large increase in foot traffic to the exhibit.
Ken’s vision for a new company that eventually became People Pattern was driven by his desire to automate the observations and analysis that he and his team had done by hand for the Ransom Center. My own background in text analysis, anthropology, and machine learning were an excellent match to Ken’s core idea and his vision to create a company that provides such insights as a software solution. I was especially excited to use my experience in semi-supervised machine learning on both text and social data to build predictive models from scratch to solve this clear market need. So, in 2013, we teamed up to form People Pattern!
Over the past four years, our team has tackled a host of challenging problems in demographic and psychographic analysis, audience aggregation and clustering, record linkage, and more. We use a heterogenous mix of models including supervised techniques from logistic regression to deep learning and semi-supervised techniques like label propagation and our own in-house methods for learning from labeled features. My data science and development team have done a fantastic job of ensuring our data pipeline can retrieve the necessary data, analyze and store it with acceptable latency. In work over the past year, I’ve queued up a number of new ideas and analyses that will take their full care and attention to implement and surface for our clients.
With my work on People Pattern’s roadmap wrapped up and the company set on the path to success, I’m now moving on to my next challenge. On May 1, I’m joining Google as a research scientist in Mountain View and am taking a leadership position in the natural language processing efforts there! This will provide me with an opportunity to return to my research roots in deep NLP tasks in syntax, semantics and discourse. My time as a professor at UT Austin taught me a lot about managing projects, individuals, and small groups, and my time at People Pattern extended that to creating a software product and leading a larger team of developers as well as the company as a whole. I look forward to jumping into the bustle at Google and working with the tremendous people, data, tools, and computational resources that it has amassed. There’s no other organization better positioned to address the most important problems in natural language processing!
We have a fantastic team at People Pattern that works hard every day to build an amazing, innovative product in a new, important and challenging marketing space. I’m incredibly grateful for my team’s energy, dedication and hard work and for how much I’ve learned throughout the entire journey. Though I’m sad to be moving on from the day-to-day operations of the company, I will nonetheless continue to be involved with People Pattern as an advisor and will thus still be able to help the company maintain its edge and improve its offerings. I look forward to continuing to be involved as the company heads into its exciting next phase!
Recent Comments