Task 2 - Value Tracking

Problem Definition

For each event, a system will traverse the input stream of documents from the event onset time, t₀, until some fixed period afterward, t_T. Throughout this simulation, the system will emit short timestamped value estimates whenever appropriate. At the end of the simulation, the system will have produced a list of tuples,

where

is the timestamp of the i^th value estimate update, v_i. The content of v_i can be either extracted from documents in the stream or generated by the system.

Evaluation

Systems will be evaluated based on how quickly they converge to a retrospectively-established value. This value will be extracted from the wikipedia infoboxes. We will be specifying as well as possible the exact piece of information we want, to reduce ambiguity, and will try to pick similar facts across multiple events to aid the learning process.