October 7, 2013
In the month leading up to CrowdConf, we will be speaking with our presenters to get a glimpse of their background and motivations behind their talks. Matt Lease, assistant professor at University of Texas at Austin, helped CrowdFlower find the most interesting research that applies to commercial applications – TM
CF: How did you get involved in crowdsourcing?
ML: I come from the search engine community, and the challenge we have here is that we need experiment results – labels – on the scale at which people are actually searching. So I need labels on a whole bunch of queries. Crowdsourcing was something I became aware of as a solution to the problems I was having getting accurate label data at scale. But the more I dug into it, the more applicability I saw for research across different fields.
CF: What would you say about the state of research in crowdsourcing? And are there particular drivers, like machine learning?
Research into crowdsourcing is still in its crucible, or infancy, right now. Maybe just five or so years old. And there is now research in all kinds of areas like volunteer-based work, open innovation, prediction markets, and others. There are still so many areas to explore, like: How do you do economics well in this market? How do I design task interfaces, examples, clear questions – the whole area of human interaction and the design space? You are talking about the intersection of humans and machines, so you have researchers in the field of psychology, sociology, marketing all finding different areas to push on, which makes for a very vibrant research community.
CF: Tell me about the selection process for this year’s conference? What were you looking for?
CrowdConf is special because it is an industry-led conference. When we put together the program committee, we made sure we had half from academia and half from industry research. My co-chair, Paul Bennett, is from Microsoft, for example. We made sure to pick talks that were not only advancing best practices but were interesting and applicable to commercial applications.
CF: Can you highlight a few of the more interesting talks you are looking forward to hearing?
When I take a look at the speakers, papers, and posters we accepted, what I’m most excited about is the diversity. We have talks on natural language processing, machine learning, visualization, psychology, assessing microloans using a distributed judgment pool. We are looking at how crowdsourcing can be used on higher-level, more complex skilled tasks. We have a talk on collaborative programming with the crowd. Even one looking at potential security applications. So we hope there is something there for everyone, and people will walk away with something they had not thought of before.
CF: What can people expect from the poster sessions?
When you have a talk, you don’t always have time for in-depth interaction. With the poster session, the researchers will be in the room, with their poster, and you can have a detailed discussion with the researcher, one on one.
CF: Tell me about the courses you teach at the University of Texas.
I teach a range of classes. One is on how search engines work. and how we can improve their design and innovate. I am teaching a class on crowdsourcing – one of the first in the country, I believe. I teach about big data computing called “Massive Text Analysis with MapReduce.” Plus a few in our professional masters program on information design among organizations, people, and technology.
CF: What do you do in your free time, when not lecturing or researching?
I love SXSW, of course. But I love the art scene and the outdoor life you get to live in Austin.