# Interaction terms for dummy variables in OLS

I'm running an OLS model to predict unpaid work hours. I am interested specifically, in the effect of education. So, I include two dummy variables for education (I have education coded as low, medium, and high, and I use low as a reference category). The dataset that I am working with is pooled cross-sections from 1994, 2002 and 2012. So, I also include two year dummies in my OLS to check the effect of time (I use 1994 as a reference category). So, overall I have four dummies.

Having run regressions separately for 1994, 2002 and 2012, I found that the effect of education varies significantly over the years. So, I want to introduce an interaction term $\text{education}\times\text{year}$ to account for this. I'm not clear how many dummies do I actually need to add in this case? And what would be my interaction reference category, if I still keep low education and 1994 as reference categories for my main effects?