Purpose: The Western University Department of Computer Science Data Validation research group will use this document as guidelines for themes to build out a research proposal for an article in line with Scientific Research for SR&ED. This group will be led by Professor Mostafa Milani, who specializes in data quality and data privacy. Research topics can be very broad and simply need to encompass our software’s purpose and relate in some way to gamification and operational activities.

Scientific Research needs to cover one or more of the following:

• Technological advancements
• Solutions to technological uncertainties
• Systematic investigation

Potential research topics:

• Data quality and accuracy correlation within a gamified interface
  • Does tagged data quality and accuracy improve when taggers are tagging in a gamified interface?
  • Are there specific gamification elements (e.g., points, leaderboards, badges) that are more effective in improving data quality compared to others?
  • Psycho-scientific approach validating engagement and accuracy levels when gamification is applied to data tagging, levering behavioral psychology department interest at Stanford
• Data validation and machine learning models to predict likelihood of correct tagging: development of algorithms that assist in predicting accuracy within a gamified interface
  • Unified data quality score based in metrics for overall assessment
  • Algorithms that train ML models to predict likelihood of tag being correct, improving automation and lowering feedback time - Neural networks, logistic regression, etc.
  • ID features that can influence tag accuracy (gamified features)
• Data privacy with large customer data sets in a gamified interface
• Breaking down complex data tagging needs into simple tasks, analyzing Sapien’s TagFlow system to increase the speed and accuracy by breaking down more complicated tagging needs into easy-to-execute tasks

Sapien’s TagFlow TLDR

The TagFlow system we are building is essentially a process that each data point follows on its "tagging journey." A data point starts at the top of the TagFlow and is sent to the first tagger, who performs the initial action on the data point. The results of that tagging are then stored and sent, along with the original data point, to the next step for further tagging. For example, if one were training a model to identify cracks in buildings and assess their seriousness, the TagFlow might start with the first task of identifying whether there is a crack. If a crack is identified, the next tagger would determine its size. If the crack is larger than a certain threshold, it could be sent to a structural engineer for more detailed analysis. This system is beneficial because the structural engineer would only review serious cracks, saving time and effort by not having to examine small or non-existent cracks, which would have been filtered out by taggers with less specialized skills.