Gary Ross, Stephen Henneberry, and Glenn Norris
JALT-CALL 2019, Aoyama Gakuin University, Tokyo, Japan
Session abstract:
The presenter will demonstrate a free online speaking system, open to all, where students can practice speaking with the voice recognition and synthesis systems built into their browser, and have the computer analyze speech patterns. Online speech represents a major development in how learners can practice language by allowing them to practice role-play, drills, and open-ended speech outside of the classroom while the computer can respond in varying dialects, speeds taking on varying roles and identities, and then grade both their accuracy and fluency.Furthermore, machine learning/AI can analyze hundreds of thousands of responses and discussions to discover patterns that are impossible to discern in a classroom or indeed in any setting. We can discover common clusters of words, patterns, tenses that Japanese students might use more than other learners. Unsupervised machine learning can even discover patterns that we aren’t looking for, and may enable us to assess speakers abilities with just a few questions. The system has been used extensively at Kanazawa University as part of the flipped classroom where students practice conversations or grammar points online in advance. Analysis of the data shows students who practiced in advance had a better understanding of the patterns when coming into the classroom thus allowing valuable class time to focus on the practice itself. (A separate paper presentation will explore the analysis of this data). The presentation will conclude with a discussion on the future of AI in the classroom particular in regards to speech.
