CNR-ITD

FRAMEWORK

To substantially upscale the TRIS model for inclusive hybrid classes, the TRIS.2 project developed and deployed an I-MOOC (Interactive - Massive Open Online Course), namely a MOOC with an optional mode allowing learners to follow a tutored version the course.

The intrinsically open nature of MOOCs means that the size of the learner cohort is always unpredictable, and they can sometimes attract hundreds of users at peak periods.

Running a tutored version of a MOOC with an unforeseeable number of participants poses enormous challenges, so the project team has responded by seeking to develop a conversational assistant (tutor-bot) that could assist tutors in handling routine tasks and automated activities in support of learners. The inherent complexity of course content queries means that these will necessarily be assigned to human tutors, as such matters are beyond the capacities of an automatic agent.

OBJECTIVES

The HEY-TUTOR! project aims to develop a multimodal / multi-surface conversational assistant capable of assisting an online tutor in performing routine support tasks within an I-MOOC based on the TRIS model. This conversational assistant is to operate in both pull and push modes: in "pull" mode, it will answer user questions by drawing on a repertoire of FAQs; in "push" mode, it will provide learners with prompts such as reminders and notifications, as well as proactive stimuli that specifically target any participants running the apparent risk of drop-out.

WORKPLAN

• Phase 1– Developing the knowledge base for the multimodal conversional assistant

  • Compiling an extensive repertoire of user support queries commonly put to CNR-ITD personnel during initial runs of the tutored I-MOOC, together with typical learner behaviour patterns related to interaction with course contents.

  • Design and implementation of the tutor-bot knowledge base underpinning development of the tutor module to be integrated in the I-MOOC.

• Phase 2- Design and implementation of a tutor-bot prototype

  • Design and implementation of an initial prototype interaction engine to manage conversational interaction between the tutor-bot and course participants, drawing on Phase 1 outputs

  • Controlled testing of the prototype tutor-bot with a sample user group to measure its effectiveness / efficiency as an online tutor system

  • Updating of the knowledge base and conversational engine

• Phase 3 – Prototype operationalisation for full multimodal/multi-surface deployment
  • Production of the release version tutor-bot, i.e. adaptation of the Phase 2 prototype for multimodal (voice, text, images, video) and multi-surface operation.

EXPECTED RESULTS

At the conclusion of the HEY-TUTOR! project, the I-MOOC based on the TRIS model for inclusive hybrid classes will be endowed with a semi-automatic tutoring system, namely a conversational assistant (tutor-bot). This integrated system will be capable of executing routine support activities, providing course participants with automated assistance, while referring any content issues, which are intrinsically complex in nature, to human tutors.