The purpose of this work was to find phenolic compounds in plants thatcould act on ruminal fermentations to limit methane emissions by ruminants, in or-der to propose natural additives or food alternatives. We used a data mining methodto extract class association rules that would identify compounds likely to have asignificant effect. Such extraction usually generates a large number of rules. Ourproblem was to select the best rules, and thereby the most promising compounds.We carried out a new kind of extraction: mining forstrongly expressedrules, that isto say rules that govern whether compounds are abundant in the plants. We proposetwo new interesting measures to evaluate the intensity of expression rules, and anew type rule visualization. Among the1,075phenolic compounds found in the 208plants analysed,7promising compound and 5 useful associations of compounds were shortlisted.