4.4. Evaluating model performance

There are several metrics that we could use to evaluate the performance of each iteration of the analysis pipeline, each of which is uniquely defined by 1) the query generation method, 2) our choice of topic model, and 3) the specification of the relevance function.

We considered several options, which are listed below.

	Evaluation metric	Definition	Benefits	Other considerations
1	Precision	xx	xx	xx
2	Recall	xx	xx	xx
3	Percent unique	xx	xx	xx

Our primary metric will be XXX based on the work of YYY.