Institute of Formal and Applied Linguistics Wiki

Hyderabad Dependency Treebank (HyDT-Hindi)

• ICON 2009
  • Shakti Standard Format (SSF; native)
  • CoNLL format
• ICON 2010
  • Shakti Standard Format (SSF; native)
  • CoNLL format
• HPST 2012 (MTPIL workshop COLING 2012; this version is also called HTB (Hindi Treebank) 0.5)
  • Shakti Standard Format (SSF; native)
  • CoNLL format

There has been no official release of the treebank yet. There have been three as-is sample releases for the purposes of the NLP tools contests in parsing Indian languages, attached to the ICON 2009 and 2010 conferences and the MTPIL workshop of COLING 2012.

There is no standard distribution channel for the treebank after the shared task evaluation period. Inquire at the LTRC (ltrc (at) iiit (dot) ac (dot) in) about the possibility of getting the data. The ICON 2010 and HPST 2012 license in short:

• non-commercial research usage
• no redistribution

HyDT-Hindi is being created by members of the Language Technologies Research Centre, International Institute of Information Technology, Gachibowli, Hyderabad, 500032, India.

• Website
  • no website dedicated to the treebank
  • ICON 2009 NLP Tools Contest On-Line Proceedings
  • ICON 2010 NLP Tools Contest On-Line Proceedings
• Data
  • no separate citation
• Principal publications
  • Samar Husain: Dependency Parsers for Indian Languages. In: Proceedings of ICON-2009 NLP Tools Contest, Hyderabad, India, 2009.
  • Samar Husain, Prashanth Mannem, Bharat Ambati, Phani Gadde: The ICON-2010 tools contest on Indian language dependency parsing. In: Proceedings of ICON-2010 NLP Tools Contest, pp. 1–8, Kharagpur, India, 2010.
• Documentation
  • http://ltrc.iiit.ac.in/nlptools2010/documentation.php

News domain corpus from ISI Kolkata.

HyDT-Hindi contains dependencies on two levels: between chunks and inside chunks. The ICON 2009 CoNLL-formatted version contained only dependencies between chunks, thus the node/tree ratio was much lower than in other treebanks. The ICON 2009 version came with a data split into three parts: training, development and test:

The ICON 2010 version came with a data split into three parts: training, development and test. The intra-chunk dependencies have been added:

I have counted the sentences and tokens (words) on the .conll files; there are slight differences from the statistics presented in (Husain et al., 2010).

The HTB 0.5 (2012) version came with a data split into three parts: training, development and test. The intra-chunk dependencies have been added:

HTB 0.5 is distributed in Devanagari UTF-8 and in the WX encoding (see below), both in SSF and CoNLL formats, each with gold-standard and automatic morphology.

The rest of this section applies to the ICON datasets. It may or may not still be valid for HTB 0.5.

The text uses the WX encoding of Indian letters. If we know what the original script is (Devanagari in this case) we can map the WX encoding to the original characters in UTF-8. WX uses English letters so if there was embedded English (or other string using Latin letters) it will probably get lost during the conversion. Note that there are (not infrequent) broken characters (\x{FFFD} REPLACEMENT CHARACTER) in the WX encoding and the correct characters cannot be recovered automatically.

Occasionally there are NULL nodes that do not correspond to any surface chunk or token. They represent ellided participants.

The syntactic tags (dependency relation labels) are karaka relations, i.e. deep syntactic roles according to the Pāṇinian grammar. There are separate versions of the treebank with fine-grained and coarse-grained syntactic tags.

According to (Husain et al., 2010), in the ICON 2010 version, the chunk tags, POS tags, lemma, morphosyntactic features and inter-chunk dependencies (topology + tags) were annotated manually. The rest (intra-chunk dependencies, headword of chunk) was marked automatically. The tool for intra-chunk dependency parsing achieves about 96% accuracy.

Note: There have been cycles in the Hindi part of HyDT.

The first two sentences of the ICON 2010 training data (with fine-grained syntactic tags) in the Shakti format:

<document docid="hi">
<head>
<title>  </title>			
<author>			
<firstname>  </firstname>			
<middlename>    </middlename>			
<lastname></lastname>			
</author>			
<availability format="electronic" />			
<bibl>			
</bibl>			
<bytecount>8.0K</bytecount>			
<domain name="general" />			
<creation creationdate="19/06/2007" institutename="IIIT Hyderabad">			
<creatorname>			
<lastname>Dipti</lastname>			
<middlename>			
</middlename>			
<firstname>Sharma</firstname>			
</creatorname>			
</creation>			
<distributor>CLIA Consortia, DIT</distributor>			
<edition number="1.0" />			
<encodingdesc>			
<newencoding>Unicode(UTF-8)</newencoding>			
<originalencoding>UTF-8</originalencoding>			
</encodingdesc>			
<sentencemarker marker=".">Specify Marker</sentencemarker>			
<language name="hi" writingsystem="LTR" script="Devanagari" />			
<normalization normalized="no">			
<utilityname>xxx.exe</utilityname>			
</normalization>			
<projectdesc name="ILMT" />			
<pubaddress addresstype="web">			
</pubaddress>			
<pubdate>			
<dateofpublication></dateofpublication>			
</pubdate>			
<publicationstmt type="copyrightfree">			
</publicationstmt>			
<publisher>			
<name></name>			
<url>xxx.com</url>			
</publisher>			
<pubplace place="books" />			
<wordcount>2  </wordcount>			
<caption>xuvryavahAra se biParIM bipASA Pilma mahowsava se vApasa lOta gaI bipASA govA. </caption>			
</caption>			
 
<annotated-resource name="HyDT-Hindi" version="2.0" type="dep-words" layers="morph,pos,chunk,dep-word" language="hin" date-of-release="20100823">
    <annotation-standard>
        <morph-standard name="Anncorra-morph" version="1.31" date="20080920" />
        <pos-standard name="Anncorra-pos" version="" date="20061215" />
        <chunk-standard name="Anncorra-chunk" version="" date="20061215" />
        <intrachunk-dependency-standard name="Anncorra-intrachunk-dep" version="1.0" date="" dep-tagset-granularity="5" />
        <dependency-standard name="Anncorra-dep" version="2.0" date="" dep-tagset-granularity="6" />
    </annotation-standard>
</annotated-resource>
</head>
<body>
<tb number="1" segment="no" bullet="no">
<foreign language="select" writingsystem="LTR"></foreign>
<text>
<Sentence id="1">
1	bAwa	NN	<fs af='bAwa,n,f,sg,3,d,0,0' drel='k1:ho' posn='10' name='bAwa' chunkId='NP' chunkType='head:NP'>
2	galawa	JJ	<fs af='galawa,adj,any,any,,any,,' drel='k1s:ho' posn='20' name='galawa' chunkId='JJP' chunkType='head:JJP'>
3	ho	VM	<fs af='ho,v,any,any,any,,0,0' drel='vmod:hE' stype='declarative' posn='30' voicetype='active' name='ho' chunkId='VGF' chunkType='head:VGF'>
4	wo	CC	<fs af='wo,avy,,,,,,' posn='40' name='wo' chunkId='CCP' chunkType='head:CCP'>
5	gussA	NN	<fs af='gussA,n,m,sg,3,d,0,0' drel='pof:AnA' posn='50' name='gussA' chunkId='NP2' chunkType='head:NP2'>
6	selebritija	NN	<fs af='selebritija,unk,,,,,0_ko,' drel='k4a:AnA' posn='60' vpos='vib_2_RP' name='selebritija' chunkId='NP3' chunkType='head:NP3'>
7	ko	PSP	<fs af='ko,psp,,,,,,' posn='70' drel='lwg__psp:selebritija' chunkType='child:NP3' name='ko'>
8	BI	RP	<fs af='BI,avy,,,,,,' posn='80' drel='lwg__rp:selebritija' chunkType='child:NP3' name='BI'>
9	AnA	VM	<fs af='A,v,any,any,any,d,nA,nA' drel='k1:hE' posn='90' name='AnA' chunkId='VGNN' chunkType='head:VGNN'>
10	lAjamI	JJ	<fs af='lAjamI,adj,any,any,,,,' drel='pof:hE' posn='100' name='lAjamI' chunkId='JJP2' chunkType='head:JJP2'>
11	hE	VM	<fs af='hE,v,any,sg,3,,hE,hE' drel='ccof:wo' stype='declarative' posn='110' voicetype='active' name='hE' chunkId='VGF2' chunkType='head:VGF2'>
12	.	SYM	<fs af='.,punc,,,,,,' posn='120' drel='rsym:hE' chunkType='child:VGF2' name='.'>
</Sentence>
 
 
<Sentence id="2">
1	bqhaspawivAra	NNP	<fs af='bqhaspawivAra,n,m,sg,3,o,0_ko,0' drel='k7t:hue' posn='10' vpos='vib_2' name='bqhaspawivAra' chunkId='NP' chunkType='head:NP'>
2	ko	PSP	<fs af='ko,psp,,,,,,' posn='20' drel='lwg__psp:bqhaspawivAra' chunkType='child:NP' name='ko'>
3	jZI	NNP	<fs af='jI,n,m,sg,3,o,0_meM,0' drel='k7:hue' posn='30' vpos='vib_2' name='jZI' chunkId='NP2' chunkType='head:NP2'>
4	meM	PSP	<fs af='meM,psp,,,,,,' posn='40' drel='lwg__psp:jZI' chunkType='child:NP2' name='meM'>
5	SurU	NN	<fs af='SurU,n,m,sg,3,d,0,0' drel='pof:hue' posn='50' name='SurU' chunkId='NP3' chunkType='head:NP3'>
6	hue	VM	<fs af='ho,v,m,sg,any,,eM,eM' drel='nmod__k1inv:mahowsava' posn='60' name='hue' chunkId='VGNF' chunkType='head:VGNF'>
7	��veM	XC	<fs af='��veM,n,m,sg,3,d,0,0' posn='70' drel='mod:mahowsava' chunkType='child:NP4' name='��veM'>
8	aMwarrARtrIya	XC	<fs af='aMwarrARtrIya,n,m,sg,3,d,0,0' posn='80' drel='mod:mahowsava' chunkType='child:NP4' name='aMwarrARtrIya'>
9	Pilma	XC	<fs af='Pilma,n,f,sg,3,d,0,0' posn='90' drel='mod:mahowsava' chunkType='child:NP4' name='Pilma'>
10	mahowsava	NNP	<fs af='mahowsava,n,m,sg,,o,0_kA,0' drel='r6:raMga' posn='100' vpos='vib_5' name='mahowsava' chunkId='NP4' chunkType='head:NP4'>
11	ke	PSP	<fs af='kA,psp,m,sg,,o,,' posn='110' drel='lwg__psp:mahowsava' chunkType='child:NP4' name='ke'>
12	raMga	NN	<fs af='raMga,n,m,sg,3,o,0_meM,0' drel='k7:padZA' posn='120' vpos='vib_2' name='raMga' chunkId='NP5' chunkType='head:NP5'>
13	meM	PSP	<fs af='meM,psp,,,,,,' posn='130' drel='lwg__psp:raMga' chunkType='child:NP5' name='meM2'>
14	BaMga	JJ	<fs af='BaMga,adj,any,any,,any,,' drel='pof:padZA' posn='140' name='BaMga' chunkId='JJP' chunkType='head:JJP'>
15	usa	DEM	<fs af='vaha,pn,any,sg,3,o,,' posn='150' drel='nmod__adj:samaya' chunkType='child:NP6' name='usa'>
16	samaya	NN	<fs af='samaya,n,any,sg,3,d,0,0' drel='k7t:padZA' posn='160' name='samaya' chunkId='NP6' chunkType='head:NP6'>
17	padZA	VM	<fs af='pada,v,any,any,any,,yA,yA' stype='declarative' posn='170' voicetype='active' name='padZA' chunkId='VGF' chunkType='head:VGF'>
18	jaba	PRP	<fs af='jaba,pn,,,,,,' drel='k7t:kiyA' posn='180' coref='samaya' name='jaba' chunkId='NP7' chunkType='head:NP7'>
19	vahAM	PRP	<fs af='vahAz,pn,,,,,0_para,' drel='jjmod:wEnAwa' posn='190' vpos='vib_2' name='vahAM' chunkId='NP8' chunkType='head:NP8'>
20	para	PSP	<fs af='para,psp,,,,,,' posn='200' drel='lwg__psp:vahAM' chunkType='child:NP8' name='para'>
21	wEnAwa	JJ	<fs af='wEnAwa,adj,any,any,,o,,' drel='nmod:surakRAkarmiyoM' posn='210' name='wEnAwa' chunkId='JJP2' chunkType='head:JJP2'>
22	surakRAkarmiyoM	NN	<fs af='surakRAkarmI,n,m,pl,3,o,0_ne,0' drel='k1:kiyA' posn='220' vpos='vib_2' name='surakRAkarmiyoM' chunkId='NP9' chunkType='head:NP9'>
23	ne	PSP	<fs af='ne,psp,,,,,,' posn='230' drel='lwg__psp:surakRAkarmiyoM' chunkType='child:NP9' name='ne'>
24	bOYlIvuda	NN	<fs af='bOYlIvuda,n,m,sg,3,o,0_kA,0' drel='r6:basu' posn='240' vpos='vib_2' name='bOYlIvuda' chunkId='NP10' chunkType='head:NP10'>
25	kI	PSP	<fs af='kA,psp,f,sg,,o,,' posn='250' drel='lwg__psp:bOYlIvuda' chunkType='child:NP10' name='kI'>
26	aBinewrI	NN	<fs af='aBinewrI,n,f,sg,3,o,0,0' posn='260' drel='nmod:bipASA' chunkType='child:NP11' name='aBinewrI'>
27	bipASA	NN	<fs af='bipASA,n,f,sg,3,d,0,0' posn='270' drel='nmod:basu' chunkType='child:NP11' name='bipASA'>
28	basu	NNP	<fs af='basu,n,f,sg,3,o,0_ke_sAWa,0' drel='k2:kiyA' posn='280' vpos='vib_vib_vib_4_5' name='basu' chunkId='NP11' chunkType='head:NP11'>
29	ke	PSP	<fs af='ke,psp,,,,,,' posn='290' drel='lwg__psp:basu' chunkType='child:NP11' name='ke2'>
30	sAWa	NST	<fs af='sAWa,nst,m,sg,3,d,,' posn='300' drel='lwg__psp:basu' chunkType='child:NP11' name='sAWa'>
31	xuvyarvahAra	NN	<fs af='xuvyarvahAra,n,m,sg,3,d,0,0' drel='pof:kiyA' posn='310' name='xuvyarvahAra' chunkId='NP12' chunkType='head:NP12'>
32	kiyA	VM	<fs af='kara,v,m,sg,any,,yA,yA' drel='nmod__relc:samaya' stype='declarative' posn='320' voicetype='active' name='kiyA' chunkId='VGF2' chunkType='head:VGF2'>
33	.	SYM	<fs af='.,punc,,,,,,' posn='330' drel='rsym:kiyA' chunkType='child:VGF2' name='.'>
</Sentence>

The same two sentences converted to the CoNLL format, WX characters decoded back to Devanagari in UTF-8:

The first sentence of the ICON 2010 development data (with fine-grained syntactic tags) in the Shakti format:

<document docid="fullnews_id_2489467">
<head>
	<caption>jela meM svasWa hE sarabajIwa xo BArawIya aXikAriyoM ne mulAkAwa kI pre isalAmAbAxa.</caption>
	<language>Hindi </language>
	<domain_name>News Articles </domain_name>
	<word_count>524</word_count>
	<byte_count>64554</byte_count>
	<availability>
		<format>CML/SSF</format>
	<sentence_marker>.</sentence_marker>
	<normalization>No</normalization>
	</availability>
	<encoding_description>
		<original_encoding>ISO 8859</format>
		<new_encoding>Unicode UTF8</new_encoding>
	</encoding_description>
	<distributor>LTRC, IIIT Hyderabad</distributor>
	<project_description>NSF Hindi/Urdu Dependency Treebanking Project</place>
	<creation>
		</raw_corpus creation_date="" institute_name="IIIT Hyderabad">
		</annotated_corpus creation_date="06/01/2009" institute_name="IIIT Hyderabad">
	<edition_number>1.0</edition_number>
	</creation>
	<publication>
		<place>New Delhi</place>
		<date>30/5/2004</date>
		<type>Newspaper</type>
		<publisher>
			<name>Amar Ujala</name>
			<url>http://www.amarujala.com</url>
		</publisher>
	</publication>
 
<annotated-resource name="HyDT-Hindi" version="2.0" type="dep-words" layers="morph,pos,chunk,dep-word" language="hin" date-of-release="20100831">
    <annotation-standard>
        <morph-standard name="Anncorra-morph" version="1.31" date="20080920" />
        <pos-standard name="Anncorra-pos" version="" date="20061215" />
        <chunk-standard name="Anncorra-chunk" version="" date="20061215" />
        <intrachunk-dependency-standard name="Anncorra-intrachunk-dep" version="1.0" date="" dep-tagset-granularity="5" />
        <dependency-standard name="Anncorra-dep" version="2.0" date="" dep-tagset-granularity="6" />
    </annotation-standard>
</annotated-resource>
</head>
<body>
<tb number="1" segment="no" bullet="no">
<foreign language="select" writingsystem="LTR"></foreign>
<text>
<Sentence id="1">
1	kota	XC	<fs af='kota,n,m,sg,3,d,0,0' posn='10' drel='mod:lAhOra' chunkType='child:NP' name='kota'>
2	laKapawa	XC	<fs af='laKapawa,n,m,sg,3,d,0,0' posn='20' drel='mod:lAhOra' chunkType='child:NP' name='laKapawa'>
3	jela	XC	<fs af='jela,n,m,sg,3,d,0,0' posn='30' drel='mod:lAhOra' chunkType='child:NP' name='jela'>
4	lAhOra	NNP	<fs af='lAhOra,n,m,sg,3,o,0_meM,0' drel='jjmod:baMxa' posn='40' vpos='vib_5' name='lAhOra' chunkId='NP' chunkType='head:NP'>
5	meM	PSP	<fs af='meM,psp,,,,,,' posn='50' drel='lwg__psp:lAhOra' chunkType='child:NP' name='meM'>
6	baMxa	JJ	<fs af='baMxa,adj,any,any,,o,,' drel='nmod:siMha' posn='60' name='baMxa' chunkId='JJP' chunkType='head:JJP'>
7	sarabajIwa	XC	<fs af='sarabajIwa,n,m,sg,3,d,0,0' posn='70' drel='mod:siMha' chunkType='child:NP2' name='sarabajIwa'>
8	siMha	NNP	<fs af='siMha,n,m,sg,3,o,0_ne,0' drel='k1:xIM' posn='80' vpos='vib_3' name='siMha' chunkId='NP2' chunkType='head:NP2'>
9	ne	PSP	<fs af='ne,psp,,,,,,' posn='90' drel='lwg__psp:siMha' chunkType='child:NP2' name='ne'>
10	maMgalavAra	NNP	<fs af='maMgalavAra,n,m,sg,3,o,0_ko,0' drel='k7t:xIM' posn='100' vpos='vib_2' name='maMgalavAra' chunkId='NP3' chunkType='head:NP3'>
11	ko	PSP	<fs af='ko,psp,,,,,,' posn='110' drel='lwg__psp:maMgalavAra' chunkType='child:NP3' name='ko'>
12	BArawIya	JJ	<fs af='BArawIya,adj,any,any,,o,,' posn='120' drel='nmod__adj:xUwAvAsa' chunkType='child:NP4' name='BArawIya'>
13	xUwAvAsa	NN	<fs af='xUwAvAsa,n,m,sg,3,o,0_kA,0' drel='r6:aXikAriyoM' posn='130' vpos='vib_3' name='xUwAvAsa' chunkId='NP4' chunkType='head:NP4'>
14	ke	PSP	<fs af='kA,psp,m,pl,,o,,' posn='140' drel='lwg__psp:xUwAvAsa' chunkType='child:NP4' name='ke'>
15	xo	QC	<fs af='xo,num,any,pl,,o,,' posn='150' drel='nmod__adj:aXikAriyoM' chunkType='child:NP5' name='xo'>
16	aXikAriyoM	NN	<fs af='aXikArI,n,m,pl,3,o,0_ko,0' drel='k4:xIM' posn='160' vpos='vib_3' name='aXikAriyoM' chunkId='NP5' chunkType='head:NP5'>
17	ko	PSP	<fs af='ko,psp,,,,,,' posn='170' drel='lwg__psp:aXikAriyoM' chunkType='child:NP5' name='ko2'>
18	apane	PRP	<fs af='apanA,pn,any,sg,1,o,0_bAre_meM,0' drel='k7:xIM' posn='180' vpos='vib_2_3' name='apane' chunkId='NP6' chunkType='head:NP6'>
19	bAre	PSP	<fs af='bAre,psp,,,,,,' posn='190' drel='lwg__psp:apane' chunkType='child:NP6' name='bAre'>
20	meM	PSP	<fs af='meM,psp,,,,,,' posn='200' drel='lwg__psp:apane' chunkType='child:NP6' name='meM2'>
21	wamAma	JJ	<fs af='wamAma,adj,any,any,,d,,' posn='210' drel='nmod__adj:jAnakAriyAM' chunkType='child:NP7' name='wamAma'>
22	vyakwigawa	JJ	<fs af='vyakwigawa,adj,any,any,,d,,' posn='220' drel='nmod__adj:jAnakAriyAM' chunkType='child:NP7' name='vyakwigawa'>
23	jAnakAriyAM	NN	<fs af='jAnakAriyAM,n,f,pl,3,d,0,0' drel='k2:xIM' posn='230' name='jAnakAriyAM' chunkId='NP7' chunkType='head:NP7'>
24	xIM	VM	<fs af='xe,v,f,pl,3,,yA,yA' stype='declarative' posn='240' voicetype='active' name='xIM' chunkId='VGF' chunkType='head:VGF'>
25	ki	CC	<fs af='ki,avy,,,,,,' drel='rs:jAnakAriyAM' posn='250' name='ki' chunkId='CCP' chunkType='head:CCP'>
26	kina	WQ	<fs af='kOna,pn,any,pl,3,o,,' posn='260' drel='mod__wq:parisWiwiyoM' chunkType='child:NP8' name='kina'>
27	parisWiwiyoM	NN	<fs af='parisWiwi,n,f,pl,3,o,0_meM,0' drel='k7:kiyA' posn='270' vpos='vib_3' name='parisWiwiyoM' chunkId='NP8' chunkType='head:NP8'>
28	meM	PSP	<fs af='meM,psp,,,,,,' posn='280' drel='lwg__psp:parisWiwiyoM' chunkType='child:NP8' name='meM3'>
29	use	PRP	<fs af='vaha,pn,any,sg,3,o,ko,ko' drel='k2:kiyA' posn='290' name='use' chunkId='NP9' chunkType='head:NP9'>
30	giraPwAra	JJ	<fs af='giraPwAra,adj,any,any,,,,' drel='pof:kiyA' posn='300' name='giraPwAra' chunkId='JJP2' chunkType='head:JJP2'>
31	kiyA	VM	<fs af='kara,v,m,sg,3,,yA_jA+yA�,yA' drel='ccof:Ora' stype='declarative' posn='310' voicetype='passive' vpos='tam_2' name='kiyA' chunkId='VGF2' chunkType='head:VGF2'>
32	gayA	VAUX	<fs af='jA,v,m,sg,3,,yA�,yA1' posn='320' drel='lwg__vaux:kiyA' chunkType='child:VGF2' name='gayA'>
33	,	SYM	<fs af=',s,punc,,,,,' posn='330' drel='rsym:kiyA' chunkType='child:VGF2' name=','>
34	mukaxamA	NN	<fs af='mukaxamA,n,m,sg,3,d,0,0' drel='k1:calA' posn='340' name='mukaxamA' chunkId='NP10' chunkType='head:NP10'>
35	calA	VM	<fs af='cala,v,m,sg,3,,yA,yA' hlt='true' drel='ccof:Ora' stype='declarative' posn='350' voicetype='active' name='calA' chunkId='VGF3' chunkType='head:VGF3'>
36	Ora	CC	<fs af='Ora,avy,,,,,,' drel='ccof:ki' posn='360' name='Ora' chunkId='CCP2' chunkType='head:CCP2'>
37	sajA	NN	<fs af='sajA,n,f,sg,3,d,0,0' drel='k1:huI' posn='370' name='sajA' chunkId='NP11' chunkType='head:NP11'>
38	huI	VM	<fs af='ho,v,f,sg,3,,yA,yA' drel='ccof:Ora' stype='declarative' posn='380' voicetype='active' name='huI' chunkId='VGF4' chunkType='head:VGF4'>
39	.	SYM	<fs af='.,punc,,,,,,' posn='390' drel='rsym:huI' chunkType='child:VGF4' name='.'>
</Sentence>

And in the CoNLL format:

And after conversion of the WX encoding to the Devanagari script in UTF-8:

The first sentence of the ICON 2010 test data (with fine-grained syntactic tags) in the Shakti format:

<document docid="fullnews_id_2484368">
<head>
	<caption>elaosI Kolane para hogI bAwacIwa pre isalAmAbAxa.</caption>
	<language>Hindi </language>
	<domain_name>News Articles </domain_name>
	<word_count>313</word_count>
	<byte_count>37563</byte_count>
	<availability>
		<format>CML/SSF</format>
	<sentence_marker>.</sentence_marker>
	<normalization>No</normalization>
	</availability>
	<encoding_description>
		<original_encoding>ISO 8859</format>
		<new_encoding>Unicode UTF8</new_encoding>
	</encoding_description>
	<distributor>LTRC, IIIT Hyderabad</distributor>
	<project_description>NSF Hindi/Urdu Dependency Treebanking Project</place>
	<creation>
		</raw_corpus creation_date="" institute_name="IIIT Hyderabad">
		</annotated_corpus creation_date="06/01/2009" institute_name="IIIT Hyderabad">
	<edition_number>1.0</edition_number>
	</creation>
	<publication>
		<place>New Delhi</place>
		<date>28/5/2004</date>
		<type>Newspaper</type>
		<publisher>
			<name>Amar Ujala</name>
			<url>http://www.amarujala.com</url>
		</publisher>
	</publication>
 
<annotated-resource name="HyDT-Hindi" version="2.0" type="dep-words" layers="morph,pos,chunk,dep-word" language="hin" date-of-release="20101013">
    <annotation-standard>
        <morph-standard name="Anncorra-morph" version="1.31" date="20080920" />
        <pos-standard name="Anncorra-pos" version="" date="20061215" />
        <chunk-standard name="Anncorra-chunk" version="" date="20061215" />
        <intrachunk-dependency-standard name="Anncorra-intrachunk-dep" version="1.0" date="" dep-tagset-granularity="5" />
        <dependency-standard name="Anncorra-dep" version="2.0" date="" dep-tagset-granularity="6" />
    </annotation-standard>
</annotated-resource>
</head>
<body>
<tb number="1" segment="no" bullet="no">
<foreign language="select" writingsystem="LTR"></foreign>
<text>
<Sentence id="1">
1	pAkiswAna	XC	<fs af='pAkiswAna,n,m,sg,3,d,0,0' posn='10' drel='mod:kaSmIra' chunkType='child:NP' name='pAkiswAna'>
2	aXikqwa	XC	<fs af='aXikqwa,adj,any,any,,o,,' posn='20' drel='mod:kaSmIra' chunkType='child:NP' name='aXikqwa'>
3	kaSmIra	NNP	<fs af='kaSmIra,n,m,sg,3,o,0_meM,0' drel='k7p:Ae' posn='30' vpos='vib_4' name='kaSmIra' chunkId='NP' chunkType='head:NP'>
4	meM	PSP	<fs af='meM,psp,,,,,,' posn='40' drel='lwg__psp:kaSmIra' chunkType='child:NP' name='meM'>
5	�	XC	<fs af='�,num,m,sg,3,d,,' posn='50' drel='mod:akwUbara' chunkType='child:NP2' name='�'>
6	akwUbara	NNP	<fs af='akwUbara,n,m,sg,3,o,0_ko,0' drel='k7t:Ae' posn='60' vpos='vib_3' name='akwUbara' chunkId='NP2' chunkType='head:NP2'>
7	ko	PSP	<fs af='ko,psp,,,,,,' posn='70' drel='lwg__psp:akwUbara' chunkType='child:NP2' name='ko'>
8	Ae	VM	<fs af='A,v,m,sg,any,,yA,yA' drel='nmod__k1inv:BUkaMpa' posn='80' name='Ae' chunkId='VGNF' chunkType='head:VGNF'>
9	BUkaMpa	NN	<fs af='BUkaMpa,n,m,sg,3,o,0_se,0' drel='rh:macI' posn='90' vpos='vib_2' name='BUkaMpa' chunkId='NP3' chunkType='head:NP3'>
10	se	PSP	<fs af='se,psp,,,,,,' posn='100' drel='lwg__psp:BUkaMpa' chunkType='child:NP3' name='se'>
11	macI	VM	<fs af='maca,v,f,sg,any,,yA,yA' drel='nmod__k1inv:wabAhI' posn='110' name='macI' chunkId='VGNF2' chunkType='head:VGNF2'>
12	wabAhI	NN	<fs af='wabAhI,n,f,sg,3,o,0_kA_bAxa,0' drel='k7t:kareMge' posn='120' vpos='vib_2_3' name='wabAhI' chunkId='NP4' chunkType='head:NP4'>
13	ke	PSP	<fs af='kA,psp,m,sg,3,o,,' posn='130' drel='lwg__psp:wabAhI' chunkType='child:NP4' name='ke'>
14	bAxa	NST	<fs af='bAxa,n,,,,,,' posn='140' drel='lwg__psp:wabAhI' chunkType='child:NP4' name='bAxa'>
15	BArawa	NNP	<fs af='BArawa,n,m,sg,3,d,0,0' drel='ccof:Ora' posn='150' name='BArawa' chunkId='NP5' chunkType='head:NP5'>
16	Ora	CC	<fs af='Ora,avy,,,,,,' drel='k1:kareMge' posn='160' name='Ora' chunkId='CCP' chunkType='head:CCP'>
17	pAkiswAna	NNP	<fs af='pAkiswAna,n,m,sg,3,d,0,0' drel='ccof:Ora' posn='170' name='pAkiswAna2' chunkId='NP6' chunkType='head:NP6'>
18	mAnavIya	JJ	<fs af='mAnavIya,adj,any,any,,o,,' posn='180' drel='nmod__adj:xqRtikoNa' chunkType='child:NP7' name='mAnavIya'>
19	xqRtikoNa	NN	<fs af='xqRtikoNa,n,m,sg,3,d,0,0' drel='k2:apanAwe' posn='190' name='xqRtikoNa' chunkId='NP7' chunkType='head:NP7'>
20	apanAwe	VM	<fs af='apanA,v,m,pl,any,,wA_ho+yA,wA' drel='vmod:kareMge' posn='200' vpos='tam_2' name='apanAwe' chunkId='VGNF3' chunkType='head:VGNF3'>
21	hue	VAUX	<fs af='ho,v,m,pl,any,,yA,yA' posn='210' drel='lwg__vaux:apanAwe' chunkType='child:VGNF3' name='hue'>
22	SanivAra	NNP	<fs af='SanivAra,n,m,sg,3,o,0_ko,0' drel='k7t:kareMge' posn='220' vpos='vib_2' name='SanivAra' chunkId='NP8' chunkType='head:NP8'>
23	ko	PSP	<fs af='ko,psp,,,,,,' posn='230' drel='lwg__psp:SanivAra' chunkType='child:NP8' name='ko2'>
24	islAmAbAxa	NNP	<fs af='isalAmAbAxa,n,m,sg,3,d,0_meM,0' drel='k7p:kareMge' posn='240' vpos='vib_2' name='islAmAbAxa' chunkId='NP9' chunkType='head:NP9'>
25	meM	PSP	<fs af='meM,psp,,,,,,' posn='250' drel='lwg__psp:islAmAbAxa' chunkType='child:NP9' name='meM2'>
26	niyaMwraNa	XC	<fs af='niyaMwraNa,n,m,sg,3,d,0,0' posn='260' drel='mod:reKA' chunkType='child:NP10' name='niyaMwraNa'>
27	reKA	NN	<fs af='reKA,n,f,sg,3,d,0,0' drel='k2:Kolane' posn='270' name='reKA' chunkId='NP10' chunkType='head:NP10'>
28	(	SYM	<fs af=',punc,,,,,,' posn='280' drel='rsym:elaosI' chunkType='child:NP11' name='('>
29	elaosI	NN	<fs af='elaosI,n,m,sg,3,d,0,0' drel='nmod:reKA' posn='290' name='elaosI' chunkId='NP11' chunkType='head:NP11'>
30	)	SYM	<fs af=',punc,,,,,,' posn='300' drel='rsym:elaosI' chunkType='child:NP11' name=')'>
31	Kolane	VM	<fs af='Kola,v,any,sg,any,o,nA_kA,nA' drel='r6:masale' posn='310' vpos='tam_2' name='Kolane' chunkId='VGNN' chunkType='head:VGNN'>
32	ke	PSP	<fs af='kA,psp,m,sg,,o,,' posn='320' drel='lwg__psp:Kolane' chunkType='child:VGNN' name='ke2'>
33	masale	NN	<fs af='masalA,n,m,sg,3,o,0_para,0' drel='k7:kareMge' posn='330' vpos='vib_2' name='masale' chunkId='NP12' chunkType='head:NP12'>
34	para	PSP	<fs af='para,psp,,,,,,' posn='340' drel='lwg__psp:masale' chunkType='child:NP12' name='para'>
35	bAwacIwa	NN	<fs af='bAwacIwa,n,f,sg,3,d,0,0' drel='pof:kareMge' posn='350' name='bAwacIwa' chunkId='NP13' chunkType='head:NP13'>
36	kareMge	VM	<fs af='kara,v,m,pl,3,,gA,gA' posn='360' name='kareMge' chunkId='VGF' chunkType='head:VGF'>
37	.	SYM	<fs af='.,punc,,,,,,' posn='370' drel='rsym:kareMge' chunkType='child:VGF' name='.'>
</Sentence>

And in the CoNLL format:

And after conversion of the WX encoding to the Devanagari script in UTF-8:

The first sentence of the HPST 2012 training data in UTF8 SSF format with gold-standard morphology:

<Sentence id='1'>
1	गुजरात	NNP	<fs af='गुजरात,n,m,sg,3,o,0_का,0' name='गुजरात' posn='10' chunkId='NP' drel='r6:मुख्यमंत्री' vpos='vib_2' chunkType='head:NP'>
2	के	PSP	<fs af='का,psp,m,sg,,o,,' name='के' posn='20' drel='lwg__psp:गुजरात' chunkType='child:NP'>
3	मुख्यमंत्री	NNP	<fs af='मुख्यमंत्री,n,m,sg,3,o,0,0' name='मुख्यमंत्री' posn='30' chunkId='NP2' drel='nmod:मोदी' chunkType='head:NP2'>
4	नरेंद्र	NNPC	<fs af='नरेंद्र,n,m,sg,3,d,0,0' name='नरेंद्र' posn='40' drel='pof__cn:मोदी' chunkType='child:NP3'>
5	मोदी	NNP	<fs af='मोदी,n,m,sg,3,o,0_ने,0' name='मोदी' posn='50' chunkId='NP3' drel='k1:किया' vpos='vib_3' chunkType='head:NP3'>
6	ने	PSP	<fs af='ने,psp,,,,,,' name='ने' posn='60' drel='lwg__psp:मोदी' chunkType='child:NP3'>
7	मंगलवार	NNP	<fs af='मंगलवार,n,m,sg,3,o,0_को,0' name='मंगलवार' posn='70' chunkId='NP4' drel='k7t:किया' vpos='vib_2' chunkType='head:NP4'>
8	को	PSP	<fs af='को,psp,,,,,,' name='को' posn='80' drel='lwg__psp:मंगलवार' chunkType='child:NP4'>
9	गृह	NNPC	<fs af='गृह,n,m,sg,3,d,0,0' name='गृह' posn='90' drel='pof__cn:मंत्री' chunkType='child:NP5'>
10	मंत्री	NNP	<fs af='मंत्री,n,m,sg,3,d,0,0' name='मंत्री' posn='100' drel='nmod__adj:पाटिल' chunkType='child:NP5'>
11	शिवराज	NNPC	<fs af='शिवराज,n,m,sg,3,d,0,0' name='शिवराज' posn='110' drel='pof__cn:पाटिल' chunkType='child:NP5'>
12	पाटिल	NNP	<fs af='पाटिल,n,m,sg,3,o,0_से,0' name='पाटिल' posn='120' chunkId='NP5' drel='k4:किया' vpos='vib_vib_5' chunkType='head:NP5'>
13	से	PSP	<fs af='से,psp,,,,,,' name='से' posn='130' drel='lwg__psp:पाटिल' chunkType='child:NP5'>
14	मुलाकात	NN	<fs af='मुलाकात,n,f,sg,3,d,0,0' name='मुलाकात' posn='140' chunkId='NP6' drel='pof:कर' chunkType='head:NP6'>
15	कर	VM	<fs af='कर,v,any,any,any,,0,0' name='कर' posn='150' chunkId='VGNF' drel='vmod:किया' chunkType='head:VGNF'>
16	आईएएस	NNP	<fs af='आईएएस,n,m,sg,3,o,0,0' name='आईएएस' posn='160' chunkId='NP7' drel='ccof:और' chunkType='head:NP7'>
17	और	CC	<fs af='और,avy,,,,,,' name='और' posn='170' chunkId='CCP' drel='r6:तर्ज' chunkType='head:CCP'>
18	आईपीएस	NNP	<fs af='आईपीएस,n,m,sg,3,o,0_का,0' name='आईपीएस' posn='180' chunkId='NP8' drel='ccof:और' vpos='vib_2' chunkType='head:NP8'>
19	की	PSP	<fs af='का,psp,f,sg,,o,,' name='की' posn='190' drel='lwg__psp:आईपीएस' chunkType='child:NP8'>
20	तर्ज	NN	<fs af='तर्ज,n,f,sg,3,o,0_पर,0' name='तर्ज' posn='200' chunkId='NP9' drel='k7:किया' vpos='vib_2' chunkType='head:NP9'>
21	पर	PSP	<fs af='पर,psp,,,,,,' name='पर' posn='210' drel='lwg__psp:तर्ज' chunkType='child:NP9'>
22	राष्ट्रीय	JJ	<fs af='राष्ट्रीय,adj,any,any,,o,,' name='राष्ट्रीय' posn='220' drel='nmod__adj:स्तर' chunkType='child:NP10'>
23	स्तर	NN	<fs af='स्तर,n,m,sg,3,o,0_पर,0' name='स्तर' posn='230' chunkId='NP10' drel='k7:किया' vpos='vib_3' chunkType='head:NP10'>
24	पर	PSP	<fs af='पर,psp,,,,,,' name='पर2' posn='240' drel='lwg__psp:स्तर' chunkType='child:NP10'>
25	एक	QC	<fs af='एक,num,any,any,,any,,' name='एक' posn='250' drel='nmod__adj:सेवा' chunkType='child:NP11'>
26	खुफिया	JJ	<fs af='खुफिया,adj,any,any,,d,,' name='खुफिया' posn='260' drel='nmod__adj:सेवा' chunkType='child:NP11'>
27	सेवा	NN	<fs af='सेवा,n,f,sg,3,d,0,0' name='सेवा' posn='270' chunkId='NP11' drel='k2:करने' chunkType='head:NP11'>
28	शुरू	NN	<fs af='शुरू,n,m,sg,3,d,0,0' name='शुरू' posn='280' chunkId='NP12' drel='pof:करने' chunkType='head:NP12'>
29	करने	VM	<fs af='कर,v,any,sg,any,o,ना_का,nA' name='करने' posn='290' chunkId='VGNN' drel='r6-k2:अनुरोध' vpos='tam_2' chunkType='head:VGNN'>
30	का	PSP	<fs af='का,psp,m,sg,,d,,' name='का' posn='300' drel='lwg__psp:करने' chunkType='child:VGNN'>
31	अनुरोध	NN	<fs af='अनुरोध,n,m,sg,3,d,0,0' name='अनुरोध' posn='310' chunkId='NP13' drel='pof:किया' chunkType='head:NP13'>
32	किया	VM	<fs af='कर,v,m,sg,any,,या,yA' name='किया' posn='320' chunkId='VGF' chunkType='head:VGF' voicetype='active' stype='declarative'>
33	।	SYM	<fs af='।,punc,,,,,,' name='।' posn='330' chunkId='BLK' drel='rsym:किया' chunkType='head:BLK'>
</Sentence>

And the same in CoNLL format:

The same sentence with “automatically tagged” morphology. Apparently it means no morphology at all, and the contestants should probably use their own taggers to tag it.

The first sentence of the development data in the UTF8 SSF format with gold-standard morphology:

<Sentence id='1'>
1	भाजपा	NNP	<fs af='भाजपा,n,f,sg,3,o,0_ने,0' name='भाजपा' posn='10' chunkId='NP' drel='k1:लगाया' vpos='vib_2' chunkType='head:NP'>
2	ने	PSP	<fs af='ने,psp,,,,,,' name='ने' posn='20' drel='lwg__psp:भाजपा' chunkType='child:NP'>
3	केंद्र	NNPC	<fs name='केंद्र' chunkId='FRAGP' chunkType='head:'FRAGP' drel='ccof:और'>
4	और	CC	<fs af='और,avy,,,,,,' name='और' posn='40' chunkId='CCP' drel='nmod:सरकार' chunkType='head:CCP'>
5	केरल	NNPC	<fs name='केरल' chunkId='FRAGP2' chunkType='head:'FRAGP2' drel='ccof:और'>
6	सरकार	NNP	<fs af='सरकार,n,f,sg,3,o,0_पर,0' name='सरकार' posn='60' chunkId='NP2' drel='k7:लगाया' vpos='vib_2' chunkType='head:NP2'>
7	पर	PSP	<fs af='पर,psp,,,,,,' name='पर' posn='70' drel='lwg__psp:सरकार' chunkType='child:NP2'>
8	भारतीय	JJ	<fs af='भारतीय,adj,any,any,,o,,' name='भारतीय' posn='80' drel='nmod__adj:ड्राइवर' chunkType='child:NP3'>
9	ड्राइवर	NN	<fs af='ड्राइवर,n,m,sg,3,o,0,0' name='ड्राइवर' posn='90' chunkId='NP3' drel='nmod:कुट्टी' chunkType='head:NP3'>
10	एम.	NNPC	<fs af='एम.,n,m,sg,3,d,0,0' name='एम.' posn='100' drel='pof__cn:कुट्टी' chunkType='child:NP4'>
11	आर.	NNPC	<fs af='आर.,n,m,sg,3,d,0,0' name='आर.' posn='110' drel='pof__cn:कुट्टी' chunkType='child:NP4'>
12	कुट्टी	NNP	<fs af='कुट्टी,n,m,sg,3,o,0_का,0' name='कुट्टी' posn='120' chunkId='NP4' drel='r6:हत्या' vpos='vib_4' chunkType='head:NP4'>
13	की	PSP	<fs af='का,psp,f,sg,,o,,' name='की' posn='130' drel='lwg__psp:कुट्टी' chunkType='child:NP4'>
14	हत्या	NN	<fs af='हत्या,n,f,sg,3,o,0_के_लिए,0' name='हत्या' posn='140' chunkId='NP5' drel='jjmod:जिम्मेदार' vpos='vib_2_3' chunkType='head:NP5'>
15	के	PSP	<fs af='के,psp,,,,,,' name='के' posn='150' drel='lwg__psp:हत्या' chunkType='child:NP5'>
16	लिए	PSP	<fs af='लिए,psp,,,,,,' name='लिए' posn='160' drel='lwg__cont:हत्या' chunkType='child:NP5'>
17	जिम्मेदार	JJ	<fs af='जिम्मेदार,adj,any,any,,o,,' name='जिम्मेदार' posn='170' chunkId='JJP' drel='nmod:तालिबान' chunkType='head:JJP'>
18	तालिबान	NNP	<fs af='तालिबान,n,m,sg,3,o,0_के_साथ,0' name='तालिबान' posn='180' chunkId='NP6' drel='ras-k1:लगाया' vpos='vib_2_3' chunkType='head:NP6'>
19	के	PSP	<fs af='के,psp,,,,,,' name='के2' posn='190' drel='lwg__psp:तालिबान' chunkType='child:NP6'>
20	साथ	NST	<fs af='साथ,nst,m,sg,3,d,,' name='साथ' posn='200' drel='lwg__cont:तालिबान' chunkType='child:NP6'>
21	निपटने	VM	<fs af='निपट,v,any,any,any,o,ना_में,nA' name='निपटने' posn='210' chunkId='VGNN' drel='k7:लगाया' vpos='tam_2' chunkType='head:VGNN'>
22	में	PSP	<fs af='में,psp,,,,,,' name='में' posn='220' drel='lwg__psp:निपटने' chunkType='child:VGNN'>
23	ढिलाई	NN	<fs af='ढिलाई,n,f,sg,3,d,0,0' name='ढिलाई' posn='230' chunkId='NP7' drel='k2:बरतने' chunkType='head:NP7'>
24	बरतने	VM	<fs af='बरत,v,any,sg,any,o,ना_का,nA' name='बरतने' posn='240' chunkId='VGNN2' drel='r6:आरोप' vpos='tam_2' chunkType='head:VGNN2'>
25	का	PSP	<fs af='का,psp,m,sg,,d,,' name='का' posn='250' drel='lwg__psp:बरतने' chunkType='child:VGNN2'>
26	आरोप	NN	<fs af='आरोप,n,m,sg,3,d,0,0' name='आरोप' posn='260' chunkId='NP8' drel='k2:लगाया' chunkType='head:NP8'>
27	लगाया	VM	<fs af='लगा,v,m,sg,3,,या_है,yA' name='लगाया' posn='270' chunkId='VGF' chunkType='head:VGF' voicetype='active' vpos='tam_2' stype='declarative'>
28	है	VAUX	<fs af='है,v,any,sg,3,,है,hE' name='है' posn='280' drel='lwg__vaux:लगाया' chunkType='child:VGF'>
29	।	SYM	<fs af='।,punc,,,,,,' name='।' posn='290' chunkId='BLK' drel='rsym:लगाया' chunkType='head:BLK'>
</Sentence>

Nonprojectivities in HyDT-Hindi are not frequent. Only 862 of the 77068 chunks in the training+development ICON 2010 version are attached nonprojectively (1.12%).

The results of the ICON 2009 NLP tools contest have been published in (Husain, 2009). There were two evaluation rounds, the first with the coarse-grained syntactic tags, the second with the fine-grained syntactic tags. To reward language independence, only systems that parsed all three languages were officially ranked. The following table presents the Hindi/coarse-grained results of the four officially ranked systems.

The results of the ICON 2010 NLP tools contest have been published in (Husain et al., 2010), page 6. These are the best results for Hindi with fine-grained syntactic tags: