'-' from tweet input shows after LDA topic detection with high weight

I hope you can help me, I’m playing around with tweets (Topic extraction, Parallel LDA), the tweets are about RFID implants. unfortunately ‘-’ weighs-in reasonably well. I have tried to ‘eliminate’ this ‘-’ with String Replacer, I’m also using Punctuation Erasure.
I hope you have good ideas!
Many thanks for all your suggestions!

Hi Outi,
Would you like to replace the whole string in a cell (for example row 4 in the second column) or just a substring which contains one or more ‘-’?
In either case, the String Replacer node or the String Manipulation node could be used to replace ‘-’ with an empty string, for example.
Could you give an example of what would be the desired outcome?
Best regards,