There are three major activities or class of problems that we are interested in for tinyDATA: Evaluation, Prediction, and Recommendation.


This involves looking at a process - often dynamic in nature - and figuring out different characteristics about it. For example, we look at an information seeker's search process in online environments and find out its goodness in real-time.


Through evaluation, we find how things have been so far. Taking this further, we may want to find out how they could be in the future. This is the prediction problem. Example include predicting search performance for an exploratory search process.


If evaluation and prediction steps inform us of problems in the ongoing process, can we suggest a better route to the user? This is the recommendation problem. Example include suggesting new queries, different sources, or a person to collaborate with.

Data. As the name suggests, tinyDATA data is small in scale. The scale could mean size, availability, scope, or time.

Task/Problem. The three major problems that we are interested in are already listed above. Other related problems include personalization, extraction, and enhancement.

Algorithms/Approaches. tinyDATA algorithms or approaches stem from machine learning and/or data mining. Specific methods include time-series analysis, interpolation, extrapolation, approximation, smoothing, and probabilistic/stochastic approaches.

Outcomes. Not surprisingly, typical outcomes of addressing tinyDATA problems include prediction/forecast, enhancement, and suggestion.

Applications. The work on tinyDATA has applications in many fields and areas, including educational intervention, error detection, information reconstruction, and health (epidemic).