Updating and Improving Classification Models for askpolly - ON-1246

Genre de projet: Innovation
Discipline(s) souhaitée(s): Informatique, Sciences mathématiques, Mathématiques, Statistiques / études actuarielles
Entreprise: Anonymous
Durée du projet: Flexible
Date souhaitée de début: Dès que possible
Langue exigée: Flexible
Emplacement(s): Ottawa, ON, Canada
Nombre de postes: 1
Niveau de scolarité désiré: DoctoratRecherche postdoctorale
Ouvert aux candidatures de personnes inscrites à un établissement à l’extérieur du Canada: No

Au sujet de l’entreprise: 

askpolly is an AI-powered market research platform that analyzes large-scale, real-world social media conversations to understand public attitudes, behaviours and audience characteristics in real time.

Veuillez décrire le projet.: 

The project will focus on:

Evaluating, retraining and improving attitude classifiers that identify beliefs, opinions and behavioural signals.
Improving demographic classifiers, including age, gender and income.
Improving geographic classifiers that identify or estimate user and conversation location.
Developing and validating additional classification models for market research and audience analysis.
Evaluating newer machine-learning and large-language-model approaches for accuracy, robustness and scalability.
Developing validation methods and performance benchmarks.

The researcher will work with askpolly’s existing datasets, models and technical team. The objective is to modernize the classification system, identify where newer AI/ML techniques can improve performance, and develop production-ready approaches for integration into the askpolly platform.

Expertise ou compétences exigées: 

The ideal candidate would have experience in some combination of:
• Machine learning and statistical classification
• Natural language processing
• Large language models
• Social media or computational social science
• Demographic inference
• Geolocation or geographic inference
• Model validation and performance measurement
• Bias, sampling and classification accuracy