maya / scripts perl /corpus_builder_praat.pl
de-francophones's picture
Upload 280 files
0836631 verified
## This programme analyses a Praat text-grid corpus of French interrogatives (caution: it does not filter sentence mood types -the corpus has to consist of questions only) according to their semantic and phonomorphosyntactic forms.
### call e.g. as...
### Corpus_builder_Praat.pl "Maya_l'Abeille.txt" "Maya"
### ...if you saved the questions of "Maya l'Abeille" under that file name.
##Preparations before using this programme:
### Praat writes by default to ".TextGrid", so you have to change the ending to ".txt" to make it readable.
### If you want to make sure that all inverted structures are found, the verb and subject of each question should be annotated as "_v" and "_s" whenever the subject is non-clitic (otherwise only pronominal inversion can be detected so far)
### To make sure that relative pronouns are not counted as question words, "qui/qu'/que" functioning as RP can be marked by "_rp"
### Then, you need to combine all of the files of your (sub-)corpus, e.g. by putting all text files into one folder and using the command "for %f in (*.txt) do type "%f" >> output.txt"(see http://www.wikihow.com/Merge-Text-(.Txt)-Files-in-Command-Prompt)
### Unfortunately this does not always work => replace the unwritten files manually.
### Make sure you did not use any question marks in your annotations since question marks are used as separators.
### Since special characters are not always recognized, you might want to use the following characters: A for à, AA for â, U for ù, E for é, EE for è, EEE for ê, C for ç, CC for Ç, I for î
########################################################################
# PART 1: Question Extracter & Annotater
########################################################################
#In the beginning
$filename = $ARGV[1]; #the second entry of the command line is the file name!
open(F, $ARGV[0]) or die("Can't open file $ARGV[0]!\n"); #opens the text file (=the first addition to the programme in the command line)
open(F1, ">$filename 1.sentences_with_intonation.txt") or die("can't write to file!\n"); #creates/overwrites the file in which the output is saved
#Go through the original text
while ($line = <F>) #for as long as there are lines, do the following for each line:
{
#set the nuclear configuration according to the last mentioned T*T%
if ($line =~ /Tones/) #if the line contains "Tones" (which is the tier name before indicating the tones)
{$TA = ""; $TB = ""} #(re)assign the empty string
if ($line =~ /\!H\*/) #if the current line contains "!H*"('a high tonal accent downstepped')
{$TA = "!H*"} #(re)assign the value "!H*" for $TA (the final accent tone)
elsif ($line =~ /\!H%/) #if the current line contains "!H%"('a high boundary tone downstepped')
{$TB = "!H%"} #(re)assign the value "!H%" for $TB (the final accent tone)
else
{
if ($line =~ /H\*/) #if the current line contains "H*"('a high tonal accent'),
{$TA = "H*"} #(re)assign the value "H*" for $TA (the final accent tone)
if ($line =~ /L\*/) #if the current line contains "L*"('a low tonal accent'),
{$TA = "L*"} #(re)assign the value "L*" for $TA (the final accent tone)
if ($line =~ /H%/) #if the current line contains "H%"('a high boundary tone'),
{$TB = "H%"} #(re)assign the value "H%" for $TB (the boundary tone)
if ($line =~ /L%/) #if the current line contains "L%"('a low boundary tone'),
{$TB = "L%"} #(re)assign the value "L%" for $TB (the boundary tone)
}
#replace the whole sentence line by the sentence only (i.e. delete everything else) and add its ID + its nuclear configuration
if ($TA eq "" or $TB eq "") #if the nuclear configuration is incomplete
{if ($line =~ /text = (.*)\?/) #leaves out all text that is not the sentence
{$line =~ s/(.*)text = (.*)\?/n.a. $2?/; #substitutes the string by nothing but the interrogative and adds the nuclear configuration to its front
print (F1 $line); #prints the adjusted line in the new file
}
}
elsif ($line =~ /text = (.*)\?/) #leaves out all text that is not the sentence
{$line =~ s/(.*)text = (.*)\?/$TA$TB $2?/; #substitutes the string by nothing but the interrogative and adds the nuclear configuration to its front
print (F1 $line); #prints the adjusted line in the new file
}
#set $id as "on"(1) when a line contains "ID" and then again as "off" (0) when "Situation" appears
if ($line =~ /ID/)
{$id = 1}
if ($line =~ /(S|s)ituation/)
{$id = 0}
if ($id == 1)
{
if($line =~ /text = "(.*)"/){$ID = $1} #when $id equals 1 and the current line contains "text = "..."", set $ID = ...
}
#set $print as "on"(1) when a line contains "prag" (or anything longer like "pragmatic question type") and then again as "off" (0) when "Tones" appears
if ($line =~ /\"(P|p)rag(.*)?\"/)
{$print = 1}
if ($line =~ /(C|c)omment/)
{$print = 0}
if ($print == 1)
{
if($line =~ /text = "(.*)"/){print (F1 "$ID $1\n")}
}
}
#Close the written file, reopen it to read it in, and open the next file to be written
close(F1); #close the corpus file
open(F1, "$filename 1.sentences_with_intonation.txt") or die("can't write to file!\n"); #reopens the corpus file to read it in
open(F1b, ">$filename 2.sentences_with_intonation_and_pragmatic_type.txt")or die("can't write to file!\n"); #creates/overwrites another file
open(F2, ">$filename 3.formally_annotated_corpus.txt") or die("can't write to file!\n"); #creates/overwrites another file in which the annotated output is saved
while ($line = <F1>)
{if ($line =~ /\w/)
{if ($line =~ /"(.*)"/)
{chomp $line}; #remove new line characters
print (F1b $line)}
}
close (F1b);
open (F1b, "$filename 2.sentences_with_intonation_and_pragmatic_type.txt")or die("can't open file!\n");
#Go through the corpus
while ($line = <F1b>)
{
#Annotate the morphosyntactic forms
## Elliptical constructions
if
($line =~ /, n'est-ce pas( )?\?/ or $line =~ /"(\(.*\) )?((Ou|Et|Mais|Rien de) )?(\S)*( (\S)*)?(, (\S)*)?( )?\?/ and $line !~ /((Q|q)u'est-ce)|(c|C)'est(.*)\?/) #if the line consists of two words only (or two words + 1 word, separated by a comma)
{
if ($line =~ /"(\(.*\))?(Tu|T'(a|e)s|Vous|C'est) (\S)*(, (\S)*)?( )?\?/ #if the line consists of "Tu, "Vous" or "C'est" + a second word (or two words separated by a comma)
or $line =~ /"((\S)+, )?(\S)+ (\S)+( (\S)+)*, n'est-ce pas( )?\?/)
{print (F2 "DEC_$line")} #annotate it as a declarative interrogative
elsif ($line =~ /(est-ce|(A|a)|(E|e))s-tu|((A|a)vez|(EEE|E|ê)tes)-vous/)
{print (F2 "INV_$line")} #annotate it as an inverted interrogative
else #otherwise
{print(F2 "ELL_$line")} #annotate it as an elliptical one
}
## "est-ce que" structures
elsif($line =~ /(\(.*\))?(E|e)st-ce (que|qu'|qui|tu|vous|nous)/) #if the current line contains either "est-ce que", "est-ce qu'" or "est-ce qui" (or "est-ce" + "tu/nous/vous")
{
print(F2 "ESK_$line") #print the annotated sentence in the file
}
## inverted structures
elsif ($line =~ /(e-j|es-t|(A|a)(i)?s-tu|t-i|t-e|ez-v|ons-n|(O|o)(U|ù) (es|sont)|_v(.*)_s)/) #if the current line contains either a verb (=>verbal ending) followed by a hyphen, or the verb is placed before the subject (=>)?
{
print(F2 "INV_$line") #print the annotaded sentence in the file
}
## morphosyntactically unmarked structures
else { #if the current line is neither inverted nor does it contain "est-ce que/qu'/qui"
print(F2 "DEC_$line") #print the annotated sentence in the file
}
}
#Close the written file, reopen it to read it in, and open the next two files to be written
close(F2); #closes the written annotated corpus file
open(F2, "$filename 3.formally_annotated_corpus.txt") or die("can't open the file!\n"); #reopens the annotated corpus file to read it in
open(F3, ">$filename 4.semantically_annotated_corpus.txt")or die("can't write to file!\n"); #opens/create a new file to edit the corpus
open(F4, ">$filename 5.fully_annotated_corpus.txt")or die("can't write to file!\n"); #opens/create a new file to edit the corpus
#categorize the sentences as yes-no, disjunctive or wh-interrogatives
##Set the counters to zero (otherwise printing the counters will result in printing no number if there is no token of one type)
$num_yni = 0;
$num_whi = 0;
$num_dsi = 0;
##Go through the formally annotated corpus
while ($line = <F2>)
{
#"ou quoi" => disjunctive interrogatives (DSI)
if ($line =~ / ou quoi( )?\?/) #if the current line contains "ou quoi" before the question mark
{
print(F3 "DSI_$line") #print DSI together with the sentence in the file
}
## => if the interrogative contains 'or what', it is a disjunctive one
#tag-only YNI
elsif ($line =~ /N'est-ce pas|Hein/)
{
$line =~ s/DEC/ELL/; print(F3 "YNI_$line")
}
#wh-interrogatives (WHI)
elsif ($line =~ /(.*)((Q|q)u'|Qui |quoi )est-ce/) #if the current line contains "Qu'est-ce", "Qui est-ce que" or (à, de, ....)"quoi est-ce que"
{$line =~ s/DEC/ESK/; print(F3 "WHI_$line")} #annotate it as WHI
elsif ($line =~ /(.*)(( q|Q)(uand|u'|ue|uoi|ui|uel)|(C|c)(omment|ombien)|(P|p)ourquoi|(O|o)(ù|U))( |\?|,)/) #if the current line contains a question word
{
if ($line =~ /INV/){print(F3 "WHI_$line")} #if the line contains "INV", print the whole sentence in the file and add "WHI" in front of it
elsif ($line =~ /ELL/){print(F3 "WHI_$line")} #if the line contains "ELL", print the whole sentence in the file and add "WHI" in front of it
elsif ($line =~ /(.*)(( q|Q)(uand|u'|uoi|ui|uel)|(C|c)(omment|ombien)|(P|p)ourquoi|(O|o)(ù|U)) est-ce qu/){print(F3 "WHI_$line")}
elsif (($line !~ /(.*)Est-ce qu| ce que |On peut savoir|(C|c)'est(.*)qu(e|i)|((V|v)ous|(T|t)u)(.*)(que|qui|quel(le)?)/) or ($line =~ /(P|p)ourquoi (.*) que/)) #and the line does not contain the interrogative particle "est-ce que" nor the relative pronoun "ce que" nor the presentative "c'est... que" (=>"que" is really a question word)
{
if ($line =~ /DEC(.*)(Q(uand|u'|ue|ui|uel|uoi)|Lequel|C(omment|ombien)|(P|p)ourquoi|Pour qu|O(U|ù)) / #if the current line is marked as declarative and starts with a question word
or $line =~ /DEC(.*)((((Et|(H(ey,|é,|E,) m|M)ais)( )?((à|A)|de|par|pour)?)|(À|A|De|Par|Pour)) (q(uand|u'|ue|ui|uel|uoi)|lequel|c(omment|ombien)|pourquoi|pour qu|o(U|ù))) /) #or a conjunction/preposition + question word
{$line =~ s/DEC/FRO/; print(F3 "WHI_$line")} #replace "DEC" by "FRO" (since the interrogative element is fronted)
elsif ($line =~ /DEC/) #if that is not the case, but the line is marked as declarative
{$line =~ s/DEC/SIT/; print(F3 "WHI_$line")}
}
else {print(F3 "YNI_$line")}
} #if the question word presumed is actually part of "este-ce que" or "c'est ... que" print "YNI" in front of the whole sentence
## => if the interrogative contains a question word, it is a constituent one
#disjunctive interrogatives (DSI)
elsif ($line =~ / ou / and $line =~ /L%/){ #if the current line contains "ou" AND is uttered with falling intonation
print(F3 "DSI_$line"); #print the number of occurrences and the whole sentence in the file
}
## => if the interrogative contains "or" AND is uttered with falling intonation, it is a disjunctive one
#yes-no interrogative (YNI)
else { #otherwise...
print(F3 "YNI_$line") #print the number of occurrences and the whole sentence in the file
}
## => in any other case, it is a polar interrogative
}
close(F3);
open(F3, "$filename 4.semantically_annotated_corpus.txt");
while ($line = <F3>)
{
#copy the short form of the pragmatic type in front of the sentence
if($line =~ /[^\(](((G|g)enuine)|(info(.*)?-seeking)|precision) (Q|q)/)
{$line =~ s/(.*)%/$1% GEN:/; $line =~ s/(.*)n\.\a./$1n.a. GEN:/;print(F4 "$line")}
elsif($line =~ /[^\(](C|c)onfirmation-seeking (Q|q)/)
{$line =~ s/(.*)%/$1% CON:/; $line =~ s/(.*)n\.a\./$1n.a. CON:/;print(F4 "$line")}
elsif($line =~ /[^\(](E|e)cho (Q|q)/)
{$line =~ s/(.*)%/$1% ECH:/; $line =~ s/(.*)n\.a\./$1n.a. ECH:/; print(F4 "$line")}
elsif($line =~ /[^\(](P|p)rompting (Q|q)/)
{$line =~ s/(.*)%/$1% PRO:/; $line =~ s/(.*)n\.a\./$1n.a. PRO:/; print(F4 "$line")}
elsif($line =~ /[^\(](S|s)elf-(add?ress?|direct)ed (Q|q)/)
{$line =~ s/(.*)%/$1% SEL:/; $line =~ s/(.*)n\.a\./$1n.a. SEL:/; print(F4 "$line")}
elsif($line =~ /[^\(](E|e)xam|(P|p)lay (Q|q)/)
{$line =~ s/(.*)%/$1% EXA:/; $line =~ s/(.*)n\.a\./$1n.a. EXA:/; print(F4 "$line")}
elsif($line =~ /[^\(](I|i)ntroductory|(T|t)hought-guiding|(N|n)ew topic (Q|q)/)
{$line =~ s/(.*)%/$1% NEW:/; $line =~ s/(.*)n\.a\./$1n.a. NEW:/; print(F4 "$line")}
elsif($line =~ /[^\(]((S|s)uggest(ing|ive)|offering) (Q|q)/)
{$line =~ s/(.*)%/$1% OFF:/; $line =~ s/(.*)n\.a\./$1n.a. OFF:/; print(F4 "$line")}
elsif($line =~ /[^\(](T|t)easing (Q|q)/)
{$line =~ s/(.*)%/$1% TEA:/; $line =~ s/(.*)n\.a\./$1n.a. TEA:/; print(F4 "$line")}
elsif($line =~ /[^\(](A|a)nswer-offering (Q|q)/)
{$line =~ s/(.*)%/$1% ANS:/; $line =~ s/(.*)n\.a\./$1n.a. ANS:/; print(F4 "$line")}
elsif($line =~ /[^\(](R|r)(eproach|eprimanding) (Q|q)/)
{$line =~ s/(.*)%/$1% REP:/; $line =~ s/(.*)n\.a\./$1n.a. REP:/; print(F4 "$line")}
elsif($line =~ /[^\(](R|r)eminding (Q|q)/)
{$line =~ s/(.*)%/$1% REM:/; $line =~ s/(.*)n\.a\./$1n.a. REM:/; print(F4 "$line")}
elsif($line =~ /[^\(](R|r)hetorical (Q|q)/)
{$line =~ s/(.*)%/$1% RHE:/; $line =~ s/(.*)n\.a\./$1n.a. RHE:/; print(F4 "$line")}
elsif($line =~ /[^\(]((I|i)mperative|(R|r)equest) (Q|q)/)
{$line =~ s/(.*)%/$1% REQ:/; $line =~ s/(.*)n\.a\./$1n.a. REQ:/; print(F4 "$line")}
elsif($line =~ /[^\(](P|p)ermission (Q|q)/)
{$line =~ s/(.*)%/$1% PER:/; $line =~ s/(.*)n\.a\./$1n.a. PER:/; print(F4 "$line")}
elsif($line =~ /[^\(](S|s)urprise (Q|q)/)
{$line =~ s/(.*)%/$1% SUR:/; $line =~ s/(.*)n\.a\./$1n.a. SUR:/; print(F4 "$line")}
elsif($line =~ /[^\(](I|i)nterview (Q|q)/)
{$line =~ s/(.*)%/$1% ITV:/; $line =~ s/(.*)n\.a\./$1n.a. ITV:/; print(F4 "$line")}
else {print(F4 "OTHER $line")}
}
close(F4);