Extract text (words) from or replace character-sequences in text

I am trying to extract words (mostly non-English) from a text string that contains unstructured combinations of other characters (letters, [-.:/]etc., and numbers). Each text string contains a different number of words. The shortest word has 2 letters. Words may contain the special letters:[ÁáÀàÂâÄäÉéÈèÊêËëÍíÌìÎîÏïIJijÓóÒòÔôÖöÚúÙùÛûÜüÝýŸÿ]
I want to extract words that only contain at least 2 letters (including special letters) and are surrounded by spaces. No periods, dash, slash etc. The icing on the cake: A valid word needs to have at least one vowel and at least one consonant.
I am trying to find a solution using the String Replacer, Regex Split, and String Manipulation nodes because all of these nodes should work. I can get somewhat close to a solution using 9 text processing nodes, but not with regular expressions. In regex101.com I can identify the words that I am looking for with the regex:
(?<=\s)([[:alpha:]])+(?=\s) but substition of the inverse is not working.
Failed attempts

  1. Regex Split with (?<=\s)([[:alpha:]])+(?=\s): Error message: did not match the pattern or contained more groups than expected. I presume it is the latter. Is there a way to add additional columns to the existing table even when the number of columns is different for each row? Is it somehow possible to attach a list of words to each row of the table? Can the “regex split” words be joined and attached?
    I am also unable to negate/inverse the regular expression. Any ideas?
  2. String Replacer: I can replace in regex101.com almost all text with that I don’t want, but not in Knime.
    I am getting single letters and extra spaces. I need a 2nd regex to clean up the remainder
    I am unable to negate the expression (?<=\s)([[:alpha:]])+(?=\s). I am not able to explicitly define the text that I do not want, because there is absolutely no pattern.
  3. String Manipulation: I am having the same problems with the negations.

Example text → desired output
RG-L-456 - oven material - lot 58363 mold.369 → oven material lot
RG-L-457 - at 1 m distance - lot 58363 mold.369 → at distance lot
1a -1035/aV1 level1 - AV1 cooling ventilator → cooling ventilator
1b - 1035/AV1 level 1 - AV1 -cooling ventilator → level ventilator
2 - 1035/AV1 level2 - AV1 cold-ventilator → (nothing should be found)
3 - 1035/AV1 work - AV1 line man → work line man
4 - 1035/AV1 insurance.1 - AV1 meeting → meeting
5 - 1035/bV1 insurance two - AV1 meeting room → insurance two meeting room
Six - 2035/aV1 LBK 8A.1 - UV1 LBK 8A → six
nr 1 - the SAS entrance → the SAS entrance

Any help would greatly be appreciated!

Hi @ssrsrb and welcome to KNIME Community Forum,

If you have the regex to extract the words you want, then the most convenient solution is to use the Regex Extractor node:


1 Like

Hi @armingrudd,
In the past I have not been successful to get anything out of the regex extractor node. Following your advice, I went back to it and found https://regexr.com/ and Pattern (Java Platform SE 8 ). It seems that the regex extractor node has not implemented the positive look behind feature (?<=X) - I am unable to get it to work. [[:alpha:]] seems to cause problems, too. I have changed my regex from (?<=\s)([[:alpha:]])+(?=\s) (which works in regex101, but not in regexr.com and not in KNIME) to
(\w+)(?:[^-.\d])(?:(?=\s)). The latter works almost perfectly in regexr.com
(1a,1b,8A should not match).
But it does not work in the extractor node in KNIME. It also matches substrings.
Output extractor node:

Configuration of regex extractor node:

Do you have any thoughts?

Also, I prepare the basic KNIME nodes to avoid dealing with constant updates. Any thoughts on how to get those basic nodes to work would be really helpful.

Thank you for any feedback.

hi @ssrsrb ,
In this workflow I’ve tried to solve the problem in two ways, involving

  • regex nodes
  • text processing nodes

The second one is better, I think, because it relies upon specialized nodes
Hope it will help you
KNIME_project4.knwf (35.2 KB)

1 Like

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.