I’m moderating the panel this afternoon at the Tech@State conference, convened by the State Department, taking place at George Washington University. We — David Broniatowski (Synexxus), Ravi Patel (Yahoo! Research), Noah Smith (Carnegie Mellon Univ), and V.S. Subrahmanian (Univ of Maryland) — will have 90 minutes of What Does It Tell Us?, looking at sense-making technologies that operate on social and online sources, within the context of the conference’s real-time awareness focus.
Here are my planned panel intro and starter discussion questions, shared in the hope that they, on their own, will provide insights into questions.
Our brief is to look at “Analyzing the vast amount of readily accessible data that flows constantly across the internet uncovers details, information and relationships that were unavailable a few years ago. This panel will examine methods and practices to glean sentiment from words and text, look at using this data to predict the future and discuss what information social networks can reveal – all accomplished with no limitation on language and on a real-time or near real-time basis.”
There’s a huge amount in that assignment. I count a dozen notions that are worth exploring. Start with “vast amount,” “readily accessible,” “data”,” “flows,” “constantly,” “analyzing… to uncover details, information, and relationships,” “unavailable a few years ago.” Then there’s “sentiment,” “predict the future,” “information social networks can reveal,” “no limitation on language,” “real-time or near real-time.”
That makes twelve notions (putting aside that some of them aren’t even atomic), or perhaps the count is muliplied given the interplay among individual notions. How do we detect “events” in “flows” and use them to “predict the future”? How is “sentiment” “data”? Is it truly “readily accessible,” and is there really “no limitation on language,” particularly when seeking to understand subjective information such as sentiment?
Let’s hear what our panelists have to say on the these points such as these, in particular as relates to today’s theme, Real Time Awareness.
… and Questions
It’s my job as moderator to prompt an interesting conversation. These questions will, I hope, serve the purpose. My expectation and hope, by the way, is that we’ll get through only a few of them. Here they are:
- Let’s start with sentiment. What role do sentiment, opinion, emotions, attitudes — various forms of subjectivity — play in analyses of the online and social worlds?
- Say you’re an analyst tasked with some business or research function (and I do include here study and formulation of policy, program analysis, intelligence, political strategy, and so on). There’s lots of information in text: “named entities” (people, place, organizations, and so on), geolocation, events, sentiment and opinion, identity clues, and so on. And then there are those imperatives: “real time,” prediction, flows. Where do you start, that is, what are the most important elements to understand, and the most important technical capabilities to have?
- We’re interested in social networks. Well, myself, I don’t view Facebook or Twitter as a social *network*. Instead, they’re platforms where networks consist of connected individuals and organizations whose links are rarely limited to any single platform. Certain technologies provide the ability to track individuals across platforms although they’re as-yet controversial. Anyway, to my question for you: How do analysis of content and of networks mesh up? Analytically, how do you match what people say to their actions and interactions? What can be learned from this sort or analysis?
- Has there been anything really cool, on the language technology front, that has emerged recently? IBM Watson, Siri, Wolfram Alpha? Something else? Only one rule for responses to this question: Please tell us about something other than what you’ve been working on yourself.
- How important, and how doable, are cross-lingual or multi-lingual analyses?
- There’s a temporal dimension to our analyses. Information sources capture, and are themselves, events. Patterns both simple and complex emerge from studying sources over time. Even the meaning of information evolves, both because today’s observer has different concerns from yesterday’s and because language changes over time. What’s your view on temporality and temporal analyses?
- In our discussion before today, some of you wanted to talk about the interplay between technical approaches and social-science techniques. Please tell us about that interplay.
- How is our field evolving? Where have we been and where are we heading? How do we make our tools more relevant now and more adaptable to emerging needs?