Deep Learning for Chatbots - QC-090

Discipline nécessaire: Master (computing, data science)
# Stagiaires / # blocs : 1 / 1 bloc (6 months)
Langue :  Anglais
Lieu : Montreal, Québec
Début du projet : 01/02/2017
Date limite : jusqu’à l’identification du stagiaire
Préférences: N/A
Nom de la compagnie: Comfiz

À propos de la compagnie:

Most organizations now recognize the necessity to become more data-driven and that the potential of artificial intelligence (AI) is undoubtedly valuable.

But this is not a simple task as this transformation requires to build a strong data foundation, to identify what are the type of operations and services that will benefit from AI and to develop, train and maintain predictive models.

With Comfiz unified DDSM platform it is now easier to acquire a holistic view of data sources and business processes associated to data-driven operations and to rapidly deploy predictive models.

Founded by Martial Humblot, a French & Canadian serial entrepreneur with a broad international experience, Comfiz is a Canadian lean start-up with operations in Montreal and Madrid (Spain).

Description du ou des projet : 

Comfiz is currently receiving a high number of demands to implement Chatbots on top of its Data-Driven Service Management Platform. Specialized in artificial intelligence and natural language processing our team has made a number of investigations to analyze how Chatbots could be implemented in Comfiz DDSM. We have decided to move ahead and to launch an initial development plan which objective is to implement retrieval-based bots on closed domains and with a text-based interface.

Objectifs/sous-objectifs de recherche:

The tasks involve in this project are the following:

  • Analyze one or two chatbots business cases that could be implemented on top of Comfiz DDSM (Comfiz team will provide the initial inputs on these business cases).
  • Write Functional Design Specification
  • Identify and integrate data sources for model training and bots tests
  • Implement working prototype with test data

Méthodologie (pour chaque projet, si prédéterminée)

We plan to work on:

  • Neural Networks training (Recurrent Neural Networks could be a good fit, but other options are possible)
  • Natural Language Processing
  • APIs development (Json Restful APIs)

Compétences recherchées (pour chaque projet):

      • Data modelling / Database Management
      • Machine learning and deep learning (neural networks)
      • Backend programming (PHP)

      Pour plus d’information ou pour postuler sur ce projet de recherche appliqué, svp

      1. Vérifiez votre admissibilité et apprenez-en plus sur les projets ouverts.
      2. Remplissez ce formulaire web. Il vous sera demandé de télécharger votre CV. N’oubliez pas d’indiquer le(s) titre(s) de projet(s) pour lequel/lesquels vous êtes intéressé et d’obtenir l’accord de votre superviseur académique!
      3. Les étudiants intéressés doivent obtenir l’accord de leur superviseur et soumettre leur CV via le formulaire web ou à Jean-Philippe Valois (jpvalois@mitacs.ca), ainsi que le lien menant à la page web universitaire de leur superviseur

      Version anglaise disponible: Cliquez ici pour plus d’information sur les projets ouverts et le formulaire web en anglais.

       

      Programme: