In general, this kind of text analytics problem is best solved by pre-processing in advance of visualization. Word Counts are the "Hello World" example for Hadoop, for example.
However, if you have a small number of words that you care about, and you want to count how many records contain at least one reference to each of your words, you can write a calculated field similar to the following:
Mentions_Tableau "if contains([DESCRIPTION], "Tableau") then 1 end"
and display COUNT([Mentions_Tableau])
Repeat for a few other important keywords, or make the search string a parameter. Probably forces a table scan in most databases, so not usually very efficient.
I would like to add that if you want to count the number of all occurrences, you can try to create something like this:
( len( [Description] )-len( replace([Description],"YourWord","" )) ) / len("YourWord")
2 of 2 people found this helpful
Another solution that would work if this is not something you have to do over live data is
Split the data Split a Field into Multiple Fields
View data, copy all and paste into Tableau as a new data source
Pivot the columns resulting from Split Pivot Data (from Columns to Rows)
Now you have each word as a separate row in the same column which allows you to calculate what you need without writing any calculations.