regex - Combining negative look-aheads with not-statements - r

Question

Dear People of Stack Overflow,
While I would say that I get the job done with regular expressions most of the time, now I have a problem I cannot seem to grasp:
I have text files I need to parse (language being R, but that doesn’t seem to matter). Essentially these files are protocols of speakers and I want to extract some information. The speakers generally follow this pattern:
Mr. Paul (speaks in English): Text.
Mr. Hernandez Gabriel (speaks in Spanish): Text.
Mr. Jenchewkow (speaks in Russian, translation provided): Text.
The regex I use for these speakers is: ^(Mr\.)\s*([^\(]*?)\s*(|\(speaks.*?\)):\s*(.*)$
The problem occurs when these speakers quote somebody else or reference something like:
Mr. Puk once said: ‚Hello‘ and I want to second that.
Here, sometimes a mismatch occurs as the regex captures everything between „Mr.“ and the colon, parsing the second capture group as: „Puk once said“ and messing up the parsed document. Thus, I tried to exclude these matches with a negative lookahead, guessing the words that could occur between Mr. and the colon like „said“, „expressed“, etc.
However, a) I seem to be unable to combine the negative lookahead with the second capture group’s ([^\(]*?) and b) this approach doesn’t seem to be that universal, given that there are other mismatches like:
Mr. Peter thought it acceptable that: Some text.
So my question is twofold: How would I exclude matches that have a „said“, „expressed“, etc. after the ‚name‘? And secondly: Is there a better, more universal way of achieving this? I thought about limiting the number of words between „Mr.“ and the colon, but that doesn’t seem to solve the problem.
Thanks in advance!
Edit:
As a reaction to the very helpful answers up to this point, I should emphasize that
a) there are indeed people with multiple names in the data
and
b) there are speakers which are not followed by a "speaks in...". Thus, Mr\.\s*([^\(]*)\s\(speaks in [^\)]*\): doesn't match them. An example would be:
Mr. Paul: Hello!
The last one was an oversight on my part when giving the initial examples. Sorry!

Answer 1

I suggest the following more flexible but still anchored pattern:
Mr\.\s*([^\(]*)\s\(speaks in [^\)]*\):
Demo
Mr. acts as start anchor, and \s\(speaks in ... ): is used for the second part. The single \s is not absolutely required but the output becomes nicer.
Your updated requirement makes it hard to come up with a watertight solution. If there is only a limited number of speakers of the loose type you could add them as separate cases like: Mr. (Paul|Peter|Matt)(?=:)) and then wrap-up everything with:
(?|Mr\.\s*([^\(]*)\s\(speaks in [^\)]*\):|Mr. (Paul|Peter|Mary)(?=:))
If this is not enough you could add alternations for the cases where there is just a name (including a second first name):
(?|Mr\.\s*([^\(]*)\s\(speaks in [^\)]*\):|Mr. ([A-Z]\w+)(?=:)|Mr. ([A-Z]\w+ [A-Z]\w+)(?=:))
Demo2

Answer 2

This more generic regex would catch the name in each case and then any text after the colon:
^Mr\.?\s*([^\s]*)[^:]*:\s*(.+)$
Note I've put a question mark after the first period in case you occasionally have Mr without a . Remove the question mark if you always want the period matched. Also, you might consider setting case insensitivity again in case occasionally you have mr. And are there no women potentially speaking?
Forgot to say: this regex assumes there is only one surname. If you have something like "Mr. García Hernández said:" then the regex will need to be more complicated to find the name. This one will only match García in that case.
EDIT: In response to further info, I'd now write the Regex like this (in R syntax):
grepl("Mr\\.?\\s*([A-Z](?:[^\\s:]|\\s(?=[A-Z]))+)[^:]*:\\s*(.+)", subject, perl=TRUE);
The conditions for this to work are that Mr is always with a capital letter, and that names always begin with a capital letter in the ASCII range [A-Z] (otherwise how is the Regex going to know it's a name?). As a plain regex it looks like this (without the R syntax):
Mr\.?\s*([A-Z](?:[^\s:]|\s(?=[A-Z]))+)[^:]*:\s*(.+)
Note that I've removed the start-of-string ^ and end-of-string $ because it seems matching ^ and $ at the end of lines in a long string is not supported in R (3.1-3.4). Change that if you're dealing with single strings. It seems that the dot doesn't work multi-line either in R, so the last (.+) matches to the end of the line. You could get some false positives if there is a speaker who addresses "As Mr. Hernández said...", though if there are no colons after that to the end of the line it should still work. This is where $ at the start could help, so add it back if necessary.
This will match any number of surnames before the colon so long as they begin with [A-Z]. This also has to be run in case-sensitive mode. If you want an explanation of how it works, just ask, but maybe you follow anyway.
Output of above regex by numbered capturing groups:
Mr. Paul (speaks in English): Text. -> 1. Paul -> 2. Text.
Mr. Hernandez Gabriel Theodor (speaks in Spanish): Text. -> 1. Hernandez Gabriel Theodor -> 2. Text.
Mr. Jenchewkow (speaks in Russian, translation provided): Text. -> 1. Jenchewkow -> 2. Text.
Mr. Puk once said: ‚Hello‘ and I want to second that. -> 1. Puk -> 2. ‚Hello‘ and I want to second that.
Mr. Peter thought it acceptable that: Some text. -> 1. Peter -> 2. Some text.
Mr Paul: Hello! -> 1. Paul -> 2. Hello!
FURTHER EDIT:
OK, so to exclude anything that has text other than something in parentheses before the colon, you can do this:
Mr\.?\s*([A-Z](?:[^\s:]|\s(?=[A-Z]))+)(?=[\s]*[(:])[^:]*:\s*(.+)
You can try it out and change options here: https://regex101.com/r/YzHPa0/1 - have a look at the Match Information on right-hand side of that screen to see what the capture groups match.
Note that this needs to be case sensitive. If you want to specify the text that goes in the parentheses for even more selectivity you'll have to change [^:]* to (?:\s\(speaks\sin[^:]+)?.