Because your data is not big, the simplest thing that might work is printing out the data on paper and reading it. Yes this might take eight hours. That's alm

Because your data is not big, the simplest thing that might work is printing out the data on paper and reading it. Yes this might take eight hours. That's almost certainly less time any alternative approach. And most importantly, time invested scales linearly. Printing out and reading twenty contacts right now, would be useful elephant-eating-one-bite-at-a-time progress in five minutes. Among the useful outcomes might be discovering that your database is less useful than you thought. Or it might let you understand what data is useful. Most importantly, it might identify holes in your data. A machine learning model cannot do that. Good luck.