Regular expression for recognizing in-text citations

I'm trying to create a regular expression to capture in-text citations.

Here's a few example sentences of in-text citations:

1. ... and the reported results in (Nivre et al., 2007) were not representative ...

2. ... two systems used a Markov chain approach (Sagae and开发者_运维技巧 Tsujii 2007).

3. Nivre (2007) showed that ...

4. ... for attaching and labeling dependencies (Chen et al., 2007; Dredze et al., 2007).

Currently, the regular expression I have is

\(\D*\d\d\d\d\)

Which matches examples 1-3, but not example 4. How can I modify this to capture example 4?

Thanks!

---

I’ve been using something like this for that purpose lately:

#!/usr/bin/env perl

use 5.010;
use utf8;
use strict;
use autodie;
use warnings qw< FATAL all >;
use open qw< :std IO :utf8 >;

my $citation_rx = qr{
    \( (?:
        \s*

        # optional author list
        (?: 
            # has to start capitalized
            \p{Uppercase_Letter}        

            # then have a lower case letter, or maybe an apostrophe
            (?=  [\p{Lowercase_Letter}\p{Quotation_Mark}] )

            # before a run of letters and admissible punctuation
            [\p{Alphabetic}\p{Dash_Punctuation}\p{Quotation_Mark}\s,.] +

        ) ?  # hook if and only if you want the authors to be optional!!

        # a reasonable year
        \b (18|19|20) \d\d 

        # citation series suffix, up to a six-parter
        [a-f] ?         \b                 

        # trailing semicolon to separate multiple citations
        ; ?  
        \s*
    ) +
    \)
}x;

while (<DATA>) {
    while (/$citation_rx/gp) {
        say ${^MATCH};
    } 
} 

__END__
... and the reported results in (Nivré et al., 2007) were not representative ...
... two systems used a Markov chain approach (Sagae and Tsujii 2007).
Nivre (2007) showed that ...
... for attaching and labelling dependencies (Chen et al., 2007; Dredze et al., 2007).

When run, it produces:

(Nivré et al., 2007)
(Sagae and Tsujii 2007)
(2007)
(Chen et al., 2007; Dredze et al., 2007)

---

Building on Tex's answer, I've written a very simple Python script called Overcite to do this for a friend (end of semester, lazy referencing you know how it is). It's open source and MIT licensed on Bitbucket.

It covers a few more cases than Tex's which might be helpful (see the test file), including ampersands and references with page numbers. The whole script is basically:

author = "(?:[A-Z][A-Za-z'`-]+)"
etal = "(?:et al.?)"
additional = "(?:,? (?:(?:and |& )?" + author + "|" + etal + "))"
year_num = "(?:19|20)[0-9][0-9]"
page_num = "(?:, p.? [0-9]+)?"  # Always optional
year = "(?:, *"+year_num+page_num+"| *\("+year_num+page_num+"\))"
regex = "(" + author + additional+"*" + year + ")"

matches = re.findall(regex, text)

---

\((.+?)\) should capture all of them

---

/\(\D*\d\d\d\d(?:;\D*\d\d\d\d)*\)/

---

All you need is to insert pattern that matches zero or more occurrences of your pattern for a citation, preceded by a semicolon. Conceptually, it's: \(cite(; cite)*\).

The pattern is: \(\D*\d{4}(;\D*\d{4})*\).

---

This is my solution, in C++ with boost regex. Hope it helps someone :-)

#include <string>
#include <boost/algorithm/string.hpp>
#include <boost/algorithm/string_regex.hpp>
#include <boost/regex.h>

using namespace std;
using namespace boost;

int Section::countCitations() {
    string Personsname = "([A-Z][a-z'`-]+)"; // Apostrophes like in "D'Alembert" and hyphens like in "Flycht-Eriksson".
    string YearPattern = "(, *(19|20)[0-9][0-9]| ?\( *(19|20)[0-9][0-9]\))"; // Either Descartes, 1990 or Descartes (1990) are accepted.
    string etal = "( et al.?)"; // You may find this
    string andconj = Personsname + " and " + Personsname;
    string commaconj = Personsname + ", " + "(" + Personsname + "|"+"et al.?"+")"; // Some authors write citations like "A, B, et al. (1995)". The comma before the "et al" pattern is not rare.

    string totcit = Personsname+"?"+etal+"?"+"("+andconj+"|"+commaconj+")*"+etal+"?"+YearPattern; 
    // Matches the following cases:
    // Xig et al. (2004); 
    // D'Alembert, Rutherford et al (2008);
    // Gino, Nino and Rino, Pino (2007)
    // (2009)
    // Gino, et al. (2005)
    cout << totcit << endl;
    regex citationform(totcit);

    int count = 0;
    string_range citation;
    string running_text(text.begin(), text.end() );
    while ((citation = find_regex(running_text, citationform)) ) { // Getting the last one
        ++count;
        string temp(running_text.begin(), citation.end() );
        running_text = running_text.substr( temp.length()-1 );
    }
    return count;
}

---

So far, this one works for me:

\s\([^(]*?\d{4}.*?\)

Depending on what you're trying to achieve, you may want to remove the leading white space (\s). I have it there because I want to remove the captured citations and if I don't include the white space, I will end up with a space between the word before the citation and the punctuation after it.

It captures all the examples mentioned in the question (see https://regex101.com/r/BwBVif/1).