Artificial Intelligence: Lightning Talks
New industrial robotic systems that operate in the same physical space as people highlights the need for robots that can integrate seamlessly into human group dynamics. We present a framework that allows robots to automatically learn human preferences from demonstrations in human-robot collaborative tasks and adapt to human preferences in real time. Results from our human experiment indicates that our framework can support effective teaming in human-robot collaborative tasks.
Twitter faces many unique challenges in the Natural Language Processing field due to the fact that its user-generated content is often not composed of academically “correct” language, but rather of slang, nonstandard grammar, emoticons, etc. For the task of Language Identification, Twitter relies on a hybrid approach incorporating character encoding information and probabilistic models. We delineate difficulties encountered and solutions found when trying to process the idiosyncratic text of Twitter.
We present our system developed for analyzing and visualizing emotions. The system takes as input informal texts and presents their flow of emotion and enduring sentiment. We put forward two practical applications: visualizing emotions in students’ learning diaries and in tweets. The impact of our system is highlighted during a user-based evaluation, showing promising results for supporting feedback in educational settings, and for assisting decision making for business and marketing.
Increasing number of modestly equipped organizations need to search large datasets for various applications. Traditional search solutions which have high computational requirements cannot be employed by such organizations. We propose a new approach, selective search, that uses very few computational resources to search datasets with millions of documents. Selective search partitions the dataset such that only a few selected subsets need to be searched for a query.
With a plethora of choices in today’s world, recommendation systems play a key role in enhancing user experience with personalized suggestions. In this talk, i will explain how Spotify uses collaborative filtering for music recommendations and how user actions influence recommendations. We will also discuss how Spotify addresses the YOLO problem: "You Only Listen Once", before judging recommendations and the challenges arising from optimizing for user engagement.