How efficiently generate unique pairs of values from a Pandas data frame with different indexes?

Question

I have a Pandas dataframe as:

+--------+------+
|numbers1|words1|
+--------+------+
|       1| word1|
|       1| word2|
|       2| word3|
|       3| word4|
|