Choosing the correct test

#1
I have to do an automated search of specific keywords that belongs to certain products over a referenced content. The result could be referring to more than one product group's keywords. Say, one group has 14 keywords and another has 7 and another has 11 keywords in total. My search could result in picking a specific number of keywords from the each group - say my results could be 7 of 14 keywords, 3 of the 7, and 2 from 11. Every keyword has the same weightage. However, the frequency of their distribution could be different. For instance, my results could be of the 7 out of 14 from the first group could be 10, 20, 25, 30, 5, 2, 50 etc and similarly for the next groups too. In this scenario, how do I provide a scoring system to evaluate a suggestion that which of the three should be valued. What is the statistical significance I should use. and what should be the confidence intervals to prescribe