Situated dialogue systems (aka conversational interfaces)
Dialogue systems are systems that can communicated using natural language. Situated dialogue systems are aware of their environments and be used in different applications. They can take the role of the instructor or the role of the assistant. Virtual instructors can be used in applications ranging from guides or trainers in simulated worlds to non player characters for virtual games. In this project we develop virtual instructors that are able to give context-dependent instructions and to interpret the user reactions in a virtual world. Virtual assistants are much more popular nowadays due to IPhone Siri.
CL: A virtual instructor that learns from human behavior
The CL system implements a novel algorithm for rapidly prototyping virtual instruction-giving agents using human-human interactions. This system does not generate utterances using classical Natural Language Generation methods but selects them from human interactions in which a similar context occurred. In a nutshell, this systems says what people said and made the task advance. CL uses automated planning techniques to group instructions into paraphrases (those instructions that caused similar reactions).
LISI: A virtual assistant for indoors environments
LISI uses CL paraphrases sets to interpret new instructions from new instruction givers. The interpretation is done by using paraphrases unigrams, visible objects and Support Vector Machines to interpret the intention of the instructor giver. Bellow you can see a video of how the system would behave on the GIVE corpus.
CS: A virtual instructor that explain action effects
The CS system uses dialogue grounding techniques to make the interaction on a virtual environment more engaging. This system was the top performer at the GIVE Challenge 2.5.
CHATBOT: dialogue systems for teenagers on Facebook
CHATBOT is a platform designed for teenagers. Users of this platform can create their own chatbots for Facebook. To see its basic features watch this video (in Spanish). This platform is used in the Challenge Dale Aceptar, a competition organized by the Fundación Sadosky where high school students can win hundreds of prizes. The platform is intended to teach basic concepts in Computer Science such as variables, regular expressions, conditionals, recursion, among others while playing a game called Coartada. For more details go here.