LING 575 - Spoken Dialog
Spring 2016
Possible Final Project Topics
Possible project types include:
- Implement a spoken dialog system for a particular task building on a
spoken dialogue toolkit.
- Perform a rigorous, corpus-based analysis of some conversational phenomenon in
conversational dialog. You may use any of spoken corpora available on the
CL cluster.
- NOTE: Required for Linguistics elective credit.
- Implement and evaluate an algorithm that performs recognition or
interpretation of a conversational dialogue behavior.
- Augment a basic spoken dialogue system with improved handling of
some dialog behavior.
Project resources
A variety of system, software, and corpus resources may be used for the
project. Some examples include:
- VoiceXML platforms, such as Voxeo, support XML-based declarative specification of dialog behavior and grammar, while supporting calls to external javascript or other code. Use of external
service APIs is strongly encouraged to expand the scope of functionality.
- Commersial API (free for research purposes) that provides ASR/NLU capabilities: Wit.ai.
- Research Dialog Framework that provides Dialog management capabilities: OpenDial. It is a Java-based
open-source development toolkit by University of Norway.
- Speech Recognition Virtual Kitchen:
Collection of Virtual Machines for speech recognition, built around open
source tools, including Kaldi speech recognition. Includes a virtual
worlds interaction environment that can be expanded, modified, etc.
- Projectoxford.ai: Microsoft's
REST APIs to speech recognition, TTS, and LUIS (Language Understanding system).
Free use up to some limit then charged.
- IBM's Watson Cloud Developer APIs, Amazon Alexa Voice Services,Google Cloud Speech API all provide similar
cloud-based speech tools
Possible APIs:
Some previous final topics:
- Analysis of turn-taking in conversational human-human dialog
- Phone-based Dialog interface to the King County Library System
- Dialog for game-playing
- Feature selection for detection of hotspots in multi-party meetings
- Recognition of emotions in human-computer dialog
- Dialog act recognition using Support Vector Machines and Hidden Markov Models
- Spoken Dialog Interface for car search
- A Web-based Annotation Tool for Spoken Dialog Research
- Chinese tutoring dialog
- Spoken Dialog Interface for Restaurant Order-taking
- Spoken Dialog search for New York Times articles
- Speech interface for tourist information
- Spoken "One Bus Away"
- Speech-based Multi-player RPG
- Interactive turn-by-turn information
- Speech-enabling "Yelp!"
- Talking to FitBit
- Recipes by Voice
- Various pre-recorded dialog systems demos can be found here