Web-based Corpus Collection
In order to extend the lexical corpus of our robot we deign and create a web-based corpus collector, http://www.dis.uniroma1.it/~s4r. The system collects the recorded audio file of the spoken command and the text counterpart of the voice command. The recording back-end of the system is supported by a modified version of [WAMI, 2008].
The user will be asked to enter some anonymous data about himself, as seen in the below picture.
The web-based corpus collection consists of two parts. In the first part we provide randomly to the users a high level task that the robot should accomplish. Tasks such as reach a place, carry objects, follow people and search and check. Shown in the bellow picture is an example of a random task.
In the second part of the web-based corpus collector, the user is presented with four sentences already recorded by someone else which he will have to repeat.
Bibliography:
[WAMI, 2008] - Alexander Gruenstein, Ian McGraw and Ibrahim Badr. The WAMI Toolkit for Developing, Deploying, and Evaluation Web-Accessible Multimodal Interfaces. 10th International Conference On Multimodal Interfaces (ICMI), 2008, URL: link