add all of Strand's data, and IPA (#9)

cldf/languages.csv

@@ -291,4 +291,5 @@ OHG,Old High German,oldh1241,10,52
 Toch,Tocharian,tokh1241,89.18,42.98
 deg,Pashai: Gorayk (Degano),sout2672,70.9008,34.6458
 Kam,Kamviri,kamv1242,71.34,35.41
-Kata,Katavari,kati1279,70.773611,35.301667
+Kata,Katavari,kati1279,70.773611,35.301667
+bhatr,Bhateri,bate1261,72.93,34.96

data/strand.py

@@ -3,6 +3,9 @@
 from urllib.error import HTTPError
 import csv
 import re
+from segments.tokenizer import Tokenizer, Profile
+
+t = Tokenizer('strand_profile.txt')
 
 chars = ['p', 'b', 'bAsp', 'f', 'v', 'w', 'm', 'uFrn', 'u', 'o', 'oFrn', 'uTns', 'oTns',
          'cDen', 'zDen', 't', 'd', 'dAsp', 's', 'z', 'l', 'lVls', 'n', 'cRet', 'jRet',
@@ -22,6 +25,35 @@
 
 codes = ['deg', 'Kam', 'Kata', 'Ash', 'Wg', 'Kho', 'Phal']
 
+with open('strand2.csv', 'w') as fout:
+    writer = csv.writer(fout)
+    link = f'http://nuristan.info/IndoAryan/SwatIndus/Bhatera/BhateraLanguage/Lexicon/lex.html'
+    try:
+        with urlopen(Request(link, headers={'User-Agent': 'Mozilla/5.0'})) as resp:
+            soup = BeautifulSoup(resp, 'html.parser')
+            for data in soup.find_all(class_='dic'):
+                word = data.find(class_='l')
+                if word:
+                    print(word)
+                    word = word.find(text=True, recursive=False)
+                    word2 = re.sub(r'ʹ(.)', r'\1ʹ', word)
+                    data = str(data).replace('\n', ' ')
+                    l = re.search(r'<b>]</b>\xa0 (.*?)\.\xa0 (.*?)\.', data)
+                    if not l:
+                        l = re.search(r'</span>[\xa0 ]+(.*?)\.\xa0\xa0([^\.]+)\.', data)
+                    print(l)
+                    if l:
+                        pos = l.group(1).lower()
+                        definition = l.group(2).lower()
+                        turner = re.search(r'T\..(\d+)', data)
+                        if turner:
+                            turner = turner.group(1)
+                            ipa = t(word2, column='IPA').replace(' ', '').replace('#', ' ')
+                            writer.writerow(['bhatr', turner, word, definition, '', ipa, '', 'strand'])
+
+    except HTTPError as e:
+        pass
+
 with open('strand.csv', 'w') as fout:
     writer = csv.writer(fout)
     for i, language in enumerate(languages):
@@ -35,6 +67,7 @@
                         word = data.find(class_='l')
                         if word:
                             word = word.find(text=True, recursive=False)
+                            word2 = re.sub(r'ʹ(.)', r'\1ʹ', word)
                             l = re.search(r'<b>]</b>\xa0 (.*?)\.\xa0 (.*?)\.', str(data))
                             if not l:
                                 l = re.search(r'</span>\xa0 (.*?)\.\xa0 (.*?)\.', str(data))
@@ -44,7 +77,8 @@
                                 turner = re.search(r'T\. (\d+)', str(data))
                                 if turner:
                                     turner = turner.group(1)
-                                    writer.writerow([codes[i], turner, word, definition, '', '', '', 'strand'])
+                                    ipa = t(word2, column='IPA').replace(' ', '').replace('#', ' ')
+                                    writer.writerow([codes[i], turner, word, definition, '', ipa, '', 'strand'])
 
             except HTTPError as e:
                 pass

data/strand2.csv

@@ -0,0 +1,156 @@
+bhatr,9691,mʹâ,i,,mɑ́,,strand
+bhatr,9691,mʹio˜,mine,,míõ,,strand
+bhatr,11302,bʹe,we,,bé,,strand
+bhatr,986,âsʹo˜,us,,ɑsó̃,,strand
+bhatr,5889,tʹu,thou,,t̪ú,,strand
+bhatr,5889,tʹi,thee,,t̪í,,strand
+bhatr,10511,tʹus,you (pl,,t̪ús,,strand
+bhatr,12815,sa,this; that,,sə,,strand
+bhatr,6495,dʹûr,far,,d̪ú̬ɾ,,strand
+bhatr,2462,yʹâk,one,,j̣ɑ́k,,strand
+bhatr,6648,dʹû,two,,d̪ú̬,,strand
+bhatr,5994,c̣ʹâ,three,,ʈ͡ʂɑ́,,strand
+bhatr,4655,ćʹôr,four,,t͡só̬ɾ,,strand
+bhatr,7655,pʹâ˜z,five,,pɑ́̃z,,strand
+bhatr,12803,ṣʹo,six,,ʂó,,strand
+bhatr,13139,sʹât,seven,,sɑ́t̪,,strand
+bhatr,941,ʹâṭ,eight,,ɑ́ʈ,,strand
+bhatr,6984,nʹu˜,nine,,nú̃,,strand
+bhatr,6227,dʹaš,ten,,d̪ə́ʃ,,strand
+bhatr,2485,yʹeš,eleven,,j̣éʃ,,strand
+bhatr,6658,duʹäš,twelve,,d̪uä́ʃ,,strand
+bhatr,6001,čʹîš,thirteen,,t͡ʃí̬ʃ,,strand
+bhatr,4605,ćʹodaš,fourteen,,t͡sód̪əʃ,,strand
+bhatr,7662,pa˜ǰʹiš,fifteen,,pə̃d͡ʒíʃ,,strand
+bhatr,12812,ṣuʹeš,sixteen,,ʂuéʃ,,strand
+bhatr,13146,satʹäš,seventeen,,sət̪ä́ʃ,,strand
+bhatr,946,aṭʹäš,eighteen,,əʈä́ʃ,,strand
+bhatr,2411,ʹu˜biš,nineteen,,ú̃biʃ,,strand
+bhatr,11616,bʹîš,twenty,,bí̬ʃ,,strand
+bhatr,6004,c̣ʹobiš,sixty,,ʈ͡ʂóbiʃ,,strand
+bhatr,4623,ćʹorbiš,eighty,,t͡sóɾbiʃ,,strand
+bhatr,12278,šʹal,hundred,,ʃə́l,,strand
+bhatr,9568,bʹuṭo˜,all,,búʈõ,,strand
+bhatr,4424,ghʹu˜,big,,gʱú̃,,strand
+bhatr,6368,ʹig,tall,,íg,,strand
+bhatr,1670,uzʹal,white,,uzə́l,,strand
+bhatr,3083,kʹâl,black,,kɑ́l,,strand
+bhatr,10539,râtʹu,red,,ɾɑt̪ú,,strand
+bhatr,8233,pʹil,yellow,,píl,,strand
+bhatr,7563,nʹil,blue,,níl,,strand
+bhatr,10930,lʹâc̣h,bad,,lɑ́ʈ͡ʂh,,strand
+bhatr,13768,thʹu,is,,t̪hú,,strand
+bhatr,6906,nʹâ,not,,nɑ́,,strand
+bhatr,9402,bhaṭʹera,bhaṭera,,bʱəʈéɾə,,strand
+bhatr,11396,bʹâṣ,rain,,bɑ́ʂ,,strand
+bhatr,14096,hiʹu˜,snow,,hiú̃,,strand
+bhatr,13493,sʹûm,earth; ground,,sú̬m,,strand
+bhatr,11348,bʹâṭh,rock,,bɑ́ʈh,,strand
+bhatr,3018,khʹûr,boulder,,khú̬ɾ,,strand
+bhatr,13386,sigʹal,sand,,sigə́l,,strand
+bhatr,13627,khʹân,mountain,,khɑ́n,,strand
+bhatr,3790,khʹâṛ,valley; stream,,khɑ́ɽ,,strand
+bhatr,8082,puʹe˜i,water,,pué̃i,,strand
+bhatr,13415,sʹîn,river,,sí̬n,,strand
+bhatr,1869,ʹûć,spring,,ú̬t͡s,,strand
+bhatr,125,ʹâ˜r,fire,,ɑ́̃ɾ,,strand
+bhatr,6849,dhʹû˜,smoke,,d̪ʱú̬̃,,strand
+bhatr,5177,zâ˜gʹâl,forest,,zɑ̃gɑ́l,,strand
+bhatr,12067,bʹîc̣h,tree,,bí̬ʈ͡ʂh,,strand
+bhatr,11120,lʹô,cedar,,ló̬,,strand
+bhatr,11209,bʹân,holly oak,,bɑ́n,,strand
+bhatr,48,âc̣hʹô,walnut,,ɑʈ͡ʂhó̬,,strand
+bhatr,3331,kʹul,half walnut,,kúl,,strand
+bhatr,22,âc̣hʹâ,apricot,,ɑʈ͡ʂhɑ́,,strand
+bhatr,1103,ʹâru,peach,,ɑ́ɾu,,strand
+bhatr,2747,kʹâu,olive,,kɑ́u,,strand
+bhatr,4287,gʹû˜,wheat,,gú̬̃,,strand
+bhatr,10431,ǰʹô,barley,,d͡ʒó̬,,strand
+bhatr,12415,šʹâl,rice (in field),,ʃɑ́l,,strand
+bhatr,9827,mʹoṣ,man,,móʂ,,strand
+bhatr,4889,ćʹun,children,,t͡sún,,strand
+bhatr,4147,gʹâ,cow,,gɑ́,,strand
+bhatr,4255,gʹu,bull,,gú,,strand
+bhatr,11239,bâćʹu,calf,,bɑt͡sú,,strand
+bhatr,4963,ćhʹel,goat (f,,t͡shél,,strand
+bhatr,10264,mʹugur,billy goat,,múguɾ,,strand
+bhatr,4973,ćhâtʹu,kid,,t͡shɑt̪ú,,strand
+bhatr,4973,ćhâtʹai,kid (f,,t͡shɑt̪ə́i,,strand
+bhatr,9606,bhʹeḍe˜,ewe,,bʱéɖẽ,,strand
+bhatr,9606,bhʹiḍ,ram,,bʱíɖ,,strand
+bhatr,10310,mâmtʹai,lamb (f,,mɑmt̪ə́i,,strand
+bhatr,10310,mâmtʹu,lamb,,mɑmt̪ú,,strand
+bhatr,4516,ghʹô,horse,,gʱó̬,,strand
+bhatr,4516,ghuʹai,mare,,gʱuə́i,,strand
+bhatr,13331,sʹâṇ,buffalo (male),,sɑ́ɳ,,strand
+bhatr,9964,mhʹeṣ,buffalo (f,,mhéʂ,,strand
+bhatr,3219,kućʹur,dog,,kut͡súɾ,,strand
+bhatr,3219,kućʹir,bitch,,kut͡síɾ,,strand
+bhatr,5177,zâ˜glʹei,bird,,zɑ̃gléi,,strand
+bhatr,9758,mʹâć,fish,,mɑ́t͡s,,strand
+bhatr,12497,ṣʹiṣ,head,,ʂíʂ,,strand
+bhatr,11572,bʹâl,hair,,bɑ́l,,strand
+bhatr,5803,tʹâl,forehead,,t̪ɑ́l,,strand
+bhatr,43,ʹâ˜c̣h,eye,,ɑ́̃ʈ͡ʂh,,strand
+bhatr,43,ʹe˜c̣h,eyes,,é̃ʈ͡ʂh,,strand
+bhatr,7031,natʹûr,nose,,nət̪ú̬ɾ,,strand
+bhatr,1533,ʹâ˜,mouth,,ɑ́̃,,strand
+bhatr,5853,dʹut,lip,,d̪út̪,,strand
+bhatr,5853,dʹot,lips,,d̪ót̪,,strand
+bhatr,6152,dʹân,tooth,,d̪ɑ́n,,strand
+bhatr,6152,dʹan,teeth,,d̪ə́n,,strand
+bhatr,5228,zʹîb,tongue; language,,zí̬b,,strand
+bhatr,9083,pho˜g,mustache,,phõg,,strand
+bhatr,6250,dʹâi,beard,,d̪ɑ́i,,strand
+bhatr,13640,kʹum,shoulder,,kúm,,strand
+bhatr,13640,kʹom,shoulders,,kóm,,strand
+bhatr,9229,bakuʹi˜,arm,,bəkuí̃,,strand
+bhatr,14024,hʹât,hand,,hɑ́t̪,,strand
+bhatr,14024,hʹat,hands,,hə́t̪,,strand
+bhatr,135,â˜guʹi,finger,,ɑ̃guí,,strand
+bhatr,3243,kʹuṭ,knee,,kúʈ,,strand
+bhatr,3906,khʹur,foot,,khúɾ,,strand
+bhatr,3906,khʹor,feet,,khóɾ,,strand
+bhatr,137,â˜gṭʹo,toe,,ɑ̃gʈó,,strand
+bhatr,14152,hiʹû,heart,,hiú̬,,strand
+bhatr,5589,dhʹîr,stomach,,d̪ʱí̬ɾ,,strand
+bhatr,3696,c̣hʹîr,milk,,ʈ͡ʂhí̬ɾ,,strand
+bhatr,8139,pʹâṣ,cow dung,,pɑ́ʂ,,strand
+bhatr,3696,c̣hʹîr,milk,,ʈ͡ʂhí̬ɾ,,strand
+bhatr,6148,dʹed,yoghurt,,d̪éd̪,,strand
+bhatr,4501,ghiʹû,ghee,,gʱiú̬,,strand
+bhatr,4368,gʹâ˜,village,,gɑ́̃,,strand
+bhatr,4336,gʹûṭ,house,,gú̬ʈ,,strand
+bhatr,12323,šʹû˜,bed,,ʃú̬̃,,strand
+bhatr,5905,tulʹei,bedding,,t̪uléi,,strand
+bhatr,5481,ṭôpʹai,hat,,ʈo̬pə́i,,strand
+bhatr,4483,gʹu˜ḍ,button,,gú̃ɖ,,strand
+bhatr,138,â˜gṭhʹui,ring,,ɑ̃gʈhúi,,strand
+bhatr,9402,bhaṭʹera wâla,man from bhaṭera,,bʱəʈéɾə wɑlə,,strand
+bhatr,1135,âpʹâ˜,kinsman,,ɑpɑ́̃,,strand
+bhatr,9935,mhʹal,father,,mhə́l,,strand
+bhatr,9935,mhʹel,mother,,mhél,,strand
+bhatr,8188,pic̣ʹu,father's brother,,piʈ͡ʂú,,strand
+bhatr,2988,kâkʹai,father's older brother,,kɑkə́i,,strand
+bhatr,9089,pʹâi,father's sister,,pɑ́i,,strand
+bhatr,10055,mâmʹai,mother's brother,,mɑmə́i,,strand
+bhatr,10001,mâsʹei,mother's sister,,mɑséi,,strand
+bhatr,9661,hʹo˜,brother,,hó̃,,strand
+bhatr,9349,bhiʹo˜,sister,,bʱió̃,,strand
+bhatr,8265,pûc̣h,son,,pu̬ʈ͡ʂh,,strand
+bhatr,6481,dhʹî,daughter,,d̪ʱí̬,,strand
+bhatr,9664,hʹoc̣,brother's son,,hóʈ͡ʂ,,strand
+bhatr,13918,sazuʹi,sister's daughter,,səzuí,,strand
+bhatr,8416,pʹu,grandson,,pú,,strand
+bhatr,8417,pʹui,granddaughter,,púi,,strand
+bhatr,9467,bâriʹu,husband,,bɑɾiú,,strand
+bhatr,12753,šʹûr,spouse's father,,ʃú̬ɾ,,strand
+bhatr,12759,ṣʹâṣ,spouse's mother,,ʂɑ́ʂ,,strand
+bhatr,9660,hâzʹe,brother's wife,,hɑzé,,strand
+bhatr,10124,mʹic̣,wife's brother,,míʈ͡ʂ,,strand
+bhatr,13871,sârʹâ˜,wife's sister,,sɑɾɑ́̃,,strand
+bhatr,2718,kʹe˜ṭ,husband's brother (younger?),,ké̃ʈ,,strand
+bhatr,5200,zʹâ˜il,husband's sister,,zɑ́̃il,,strand
+bhatr,11251,bhuʹâiṭ,son's wife,,bʱuɑ́iʈ,,strand
+bhatr,5228,zʹîb,tongue; language,,zí̬b,,strand
+bhatr,9402,bhaṭʹe sa zîb,bhaṭera language,,bʱəʈé sə zi̬b,,strand

data/strand_profile.txt

@@ -0,0 +1,83 @@
+Grapheme	IPA
+p	p
+b	b
+bh	bʱ
+f	f
+v	v
+w	w
+m	m
+ü	ü
+ü	ü
+-	
+u	u
+ö	ö
+o	o
+û	u̬
+ô	o̬
+ć	t͡s
+ź	d͡z
+źh	d͡zʱ
+t	t̪
+d	d̪
+dh	d̪ʱ
+s	s
+z	z
+l	l
+ɬ	ɬ
+n	n
+c̣	ʈ͡ʂ
+j̣	ɖ͡ʐ
+J̣	ɖ͡ʐ
+j̣h	ɖ͡ʐʱ
+ṭ	ʈ
+ḍ	ɖ
+ḍh	ɖʱ
+ṣ	ʂ
+ẓ	ʐ
+r	ɾ
+ṛ	ɽ
+ḷ	ɭ
+ř	ɻ
+ň	ɻ̃
+ṇ	ɳ
+r̥	ɻ̩
+č	t͡ʃ
+ǰ	d͡ʒ
+ǰh	d͡ʒʱ
+š	ʃ
+ž	ʒ
+y	j̣
+i	i
+e	e
+ä	ä
+î	i̬
+ê	e̬
+k̂	kʲ
+ĝ	gʲ
+ĝh	gʲʱ
+k	k
+g	g
+gh	gʱ
+x	x
+ǧ	ɣ
+ʱ	ʱ
+ʷ	ʷ
+ŋ	ŋ
+ï	ï
+a	ə
+â	ɑ
+å	a
+q	q
+ḥ	ħ
+ʕ	ʕ
+ʔ	ʔ
+h	h
+ɦ	ɦ
+ɦ̊	ɦ̊
+ʹ	́
+˜	̃
+_ 
+=	
+·	
++	
+,