Speakoo

Our project aims to help ESL and EFL learners by giving them crowdsourced feedback on their speaking accuracy.

Our motivation for this project is to improve communication, a significant social skill. English is the most widely accepted means of communication worldwide, and can give its speakers advantages in education and employment, and means to conversate with a large percentage of the world’s inhabitants, increasing opportunity to learn different perspectives from cultural differences. Learning a second language has also been shown to boost creativity.

This project will be based off of the technology of Roc Speak (available at rocspeak.com), the existing platform for users to get semi-automated feedback on their speaking skills. The users can record themselves speaking in Roc Speak, and get automated feedback on smile intensity, volume modulation, body gestures etc., along with peer feedback in the form of comments and ratings on those attributes. In Speakoo, we are interested in the accuracy and content quality of speech delivery, which leads to potential applications in the ESL domain.

In the bigger picture, an ESL learner will similarly be able to record their speeches, and get feedback on the accuracy of (1) grammar, (2) pronunciation, (3) vocabulary and (4) content development through the final system. The feedback will be semi-automated, as some of the objective feedback will be generated by the machine, while the peers in the learning community will chip in with subjective feedback. The ESL learners will be given customized prompts for recording their speeches on, and those prompts will focus on helping them improve the skills they are struggling with.

In this project we are exploring a grammatical feedback generation mechanism, and focusing only on the crowdworkers’ user interface, in order to engage them more effectively.