===== Telugu (te) ===== Hyderabad Dependency Treebank (HyDT-Telugu) ==== Versions ==== * ICON 2009 * [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/SSF.pdf|Shakti Standard Format]] (SSF; native) * CoNLL format * ICON 2010 * Shakti Standard Format (SSF; native) * CoNLL format There has been no official release of the treebank yet. There have been two as-is sample releases for the purposes of the NLP tools contests in parsing Indian languages, attached to the [[http://ltrc.iiit.ac.in/nlptools2009/|ICON 2009]] and [[http://ltrc.iiit.ac.in/nlptools2010/|2010]] conferences. ==== Obtaining and License ==== There is no standard distribution channel for the treebank after the ICON 2010 evaluation period. Inquire at the LTRC (ltrc (at) iiit (dot) ac (dot) in) about the possibility of getting the data. The ICON 2010 license in short: * non-commercial research usage * no redistribution HyDT-Telugu is being created by members of the [[http://ltrc.iiit.ac.in/|Language Technologies Research Centre]], International Institute of Information Technology, Gachibowli, Hyderabad, 500032, India. ==== References ==== * Website * //no website dedicated to the treebank// * [[http://ltrc.iiit.ac.in/nlptools2009/papers.php|ICON 2009 NLP Tools Contest On-Line Proceedings]] * [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/toolscontest10-workshoppaper-final.pdf|ICON 2010 NLP Tools Contest On-Line Proceedings]] * Data * //no separate citation// * Principal publications * Samar Husain: [[http://ltrc.iiit.ac.in/nlptools2009/CR/intro-husain.pdf|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: [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/toolscontest10-workshoppaper-final.pdf|Proceedings of ICON-2010 NLP Tools Contest]], pp. 1–8, Kharagpur, India, 2010. * Documentation * http://ltrc.iiit.ac.in/nlptools2010/documentation.php ==== Domain ==== Unknown. ==== Size ==== HyDT-Telugu shows dependencies between chunks, not words. The node/tree ratio is thus much lower than in other treebanks. The ICON 2009 version came with a data split into three parts: training, development and test: ^ Part ^ Sentences ^ Chunks ^ Ratio ^ | Training | 1456 | 5494 | 3.77 | | Development | 150 | 675 | 4.50 | | Test | 150 | 583 | 3.89 | | TOTAL | 1756 | 6752 | 3.85 | As for ICON 2010, the data description in [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/toolscontest10-workshoppaper-final.pdf|(Husain et al., 2010)]] does not match the data that we downloaded during the contest. They indicate the number of words, while we can count the number of nodes, i.e. chunks. Anyway, the number of training sentences should match and it does not. Also note that they give the same number of words as they gave for ICON 2009 in [[http://ltrc.iiit.ac.in/nlptools2009/CR/intro-husain.pdf|(Husain et al., 2009)]]. In any case, the training data shrank during the year (clean up?) In the following table, we give both the published and real numbers of sentences, the published number of words and the counted number of chunks (nodes). ^ Part ^ Sentences ^ Chunks ^ Ratio ^ PSentences ^ Words ^ Ratio ^ | Training | 1300 | 5125 | 3.94 | 1400 | 7602 | 5.43 | | Development | 150 | 597 | 3.98 | 150 | 839 | 5.59 | | Test | 150 | 599 | 3.99 | 150 | 836 | 5.57 | | TOTAL | 1600 | 6321 | 3.95 | 1700 | 9277 | 5.46 | ==== Inside ==== The text uses the [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/map.pdf|WX encoding]] of Indian letters. If we know what the original script is (Telugu 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. The CoNLL format contains only the chunk heads. The native SSF format shows the other words in the chunk, too, but it does not capture intra-chunk dependency relations. This is an example of a multi-word chunk: 3 (( NP 3.1 932 QC 3.2 maMxi CL 3.3 AdavAlYle NN )) In the CoNLL format, the CPOS column contains the [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/Chunk-Tag-List.pdf|chunk label]] (e.g. ''NP'' = //noun phrase//) and the POS column contains the [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/POS-Tag-List.pdf|part of speech]] of the chunk head. Occasionally there are ''NULL'' nodes that do not correspond to any surface chunk or token. They represent ellided participants. The [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/dep-tagset.pdf|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 [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/mapping_fine-to-coarse.pdf|fine-grained and coarse-grained]] syntactic tags. According to [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/toolscontest10-workshoppaper-final.pdf|(Husain et al., 2010)]], in the ICON 2010 version, the chunk tags, POS tags and inter-chunk dependencies (topology + tags) were annotated manually. The rest (lemma, morphosyntactic features, headword of chunk) was marked automatically. Note: There have been cycles in the Hindi part of HyDT but no such problem occurs in the Telugu part. ==== Sample ==== The first sentence of the ICON 2010 training data (with fine-grained syntactic tags) in the Shakti format: 1 (( NP 1.1 maro QF 1.2 saMgawi NN )) 2 (( NP 2.1 mIku PRP )) 3 (( VGF 3.1 weVlusA VM 3.2 ? SYM )) And in the CoNLL format: | 1 | saMgawi | saMgawi | NP | NN | lex-saMgawi|cat-n|gend-|num-sg|pers-|case-d|vib-0|tam-0|head-saMgawi | 3 | k1 | _ | _ | | 2 | mIku | mIru | NP | PRP | lex-mIru|cat-pn|gend-any|num-pl|pers-2|case-|vib-ki|tam-ki|head-mIku|name-NP2 | 3 | k4 | _ | _ | | 3 | weVlusA | weVlusA | VGF | VM | lex-weVlusA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-weVlusA|name-VGF | 0 | main | _ | _ | And after conversion of the WX encoding to the Telugu script in UTF-8: | 1 | సంగతి | సంగతి | NP | NN | lex-saMgawi|cat-n|gend-|num-sg|pers-|case-d|vib-0|tam-0|head-saMgawi | 3 | k1 | _ | _ | | 2 | మీకు | మీరు | NP | PRP | lex-mIru|cat-pn|gend-any|num-pl|pers-2|case-|vib-ki|tam-ki|head-mIku|name-NP2 | 3 | k4 | _ | _ | | 3 | తెలుసా | తెలుసా | VGF | VM | lex-weVlusA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-weVlusA|name-VGF | 0 | main | _ | _ | The first sentence of the ICON 2010 development data (with fine-grained syntactic tags) in the Shakti format: 1 (( RBP 1.1 eVMwa WQ )) 2 (( NP 2.1 bAXEnA NN )) 3 (( NP 3.1 xIni PRP )) 4 (( RBP 4.1 eVlA WQ )) 5 (( NP 5.1 bayata NST )) 6 (( VGNF 6.1 peVttukoVnexi VM )) 7 (( RBP 7.1 sarigA RB )) 8 (( NP 8.1 viRayaM NN )) 9 (( VGNN 9.1 weVliyakapovadaM VM 9.2 . SYM )) And in the CoNLL format: | 1 | eVMwa | eVMwa | RBP | WQ | lex-eVMwa|cat-pn|gend-|num-sg|pers-|case-d|vib-0|tam-0|head-eVMwa | 2 | adv | _ | _ | | 2 | bAXEnA | bAXEnA | NP | NN | lex-bAXEnA|cat-unk|gend-|num-|pers-|case-|vib-|tam-|head-bAXEnA|name-NP | 6 | k2s | _ | _ | | 3 | xIni | ixi | NP | PRP | lex-ixi|cat-pn|gend-fn|num-sg|pers-3|case-o|vib-ti|tam-ti|head-xIni|name-NP2 | 6 | k2 | _ | _ | | 4 | eVlA | eVlA | RBP | WQ | lex-eVlA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-eVlA|name-RBP2 | 6 | adv | _ | _ | | 5 | bayata | bayata | NP | NST | lex-bayata|cat-n|gend-|num-sg|pers-|case-d|vib-0|tam-0|head-bayata|name-NP3 | 6 | pof | _ | _ | | 6 | peVttukoVnexi | peVttuko | VGNF | VM | lex-peVttuko|cat-pn|gend-|num-sg|pers-|case-|vib-e_axi|tam-e_axi_0|head-peVttukoVnexi|name-VGNF | 9 | k1s | _ | _ | | 7 | sarigA | sarigA | RBP | RB | lex-sarigA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-sarigA|name-RBP3 | 9 | adv | _ | _ | | 8 | viRayaM | viRayaM | NP | NN | lex-viRayaM|cat-n|gend-|num-sg|pers-|case-d|vib-0|tam-0|head-viRayaM|name-NP4 | 9 | k1 | _ | _ | | 9 | weVliyakapovadaM | weVliyu | VGNN | VM | lex-weVliyu|cat-v|gend-any|num-any|pers-any|case-|vib-aka_po_adaM|tam-aka_po_adaM|head-weVliyakapovadaM|name-VGNN | 0 | main | _ | _ | And after conversion of the WX encoding to the Telugu script in UTF-8: | 1 | ఎంత | ఎంత | RBP | WQ | lex-eVMwa|cat-pn|gend-|num-sg|pers-|case-d|vib-0|tam-0|head-eVMwa | 2 | adv | _ | _ | | 2 | బాధైనా | బాధైనా | NP | NN | lex-bAXEnA|cat-unk|gend-|num-|pers-|case-|vib-|tam-|head-bAXEnA|name-NP | 6 | k2s | _ | _ | | 3 | దీని | ఇది | NP | PRP | lex-ixi|cat-pn|gend-fn|num-sg|pers-3|case-o|vib-ti|tam-ti|head-xIni|name-NP2 | 6 | k2 | _ | _ | | 4 | ఎలా | ఎలా | RBP | WQ | lex-eVlA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-eVlA|name-RBP2 | 6 | adv | _ | _ | | 5 | బయట | బయట | NP | NST | lex-bayata|cat-n|gend-|num-sg|pers-|case-d|vib-0|tam-0|head-bayata|name-NP3 | 6 | pof | _ | _ | | 6 | పెట్టుకొనేది | పెట్టుకొ | VGNF | VM | lex-peVttuko|cat-pn|gend-|num-sg|pers-|case-|vib-e_axi|tam-e_axi_0|head-peVttukoVnexi|name-VGNF | 9 | k1s | _ | _ | | 7 | సరిగా | సరిగా | RBP | RB | lex-sarigA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-sarigA|name-RBP3 | 9 | adv | _ | _ | | 8 | విషయం | విషయం | NP | NN | lex-viRayaM|cat-n|gend-|num-sg|pers-|case-d|vib-0|tam-0|head-viRayaM|name-NP4 | 9 | k1 | _ | _ | | 9 | తెలియకపొవడం | తెలియు | VGNN | VM | lex-weVliyu|cat-v|gend-any|num-any|pers-any|case-|vib-aka_po_adaM|tam-aka_po_adaM|head-weVliyakapovadaM|name-VGNN | 0 | main | _ | _ | The first sentence of the ICON 2010 test data (with fine-grained syntactic tags) in the Shakti format: 1 (( NP 1.1 iMkA PRP )) 2 (( RBP 2.1 warawarAlugA RB )) 3 (( VGNF 3.1 nAtukupoyina VM )) 4 (( NP 4.1 aBiprAyAlu NN )) 5 (( NULL_VGF 5.1 NULL VM 5.2 . SYM )) And in the CoNLL format: | 1 | iMkA | iMkA | NP | PRP | lex-iMkA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-iMkA|name-NP|poslcat-NM | 5 | vmod | _ | _ | | 2 | warawarAlugA | warawarAlugA | RBP | RB | lex-warawarAlugA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-warawarAlugA|name-RBP|poslcat-NM | 3 | adv | _ | _ | | 3 | nAtukupoyina | nAtuko | VGNF | VM | lex-nAtuko|cat-v|gend-any|num-any|pers-any|case-|vib-i_po_ina|tam-i_po_ina|head-nAtukupoyina|name-VGNF | 4 | nmod | _ | _ | | 4 | aBiprAyAlu | aBiprAyaM | NP | NN | lex-aBiprAyaM|cat-n|gend-|num-pl|pers-|case-d|vib-0|tam-0|head-aBiprAyAlu|name-NP2 | 5 | k1 | _ | _ | | 5 | NULL | NULL | NULL_VGF | VM | name-NULL_VGF | 0 | main | _ | _ | And after conversion of the WX encoding to the Telugu script in UTF-8: | 1 | ఇంకా | ఇంకా | NP | PRP | lex-iMkA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-iMkA|name-NP|poslcat-NM | 5 | vmod | _ | _ | | 2 | తరతరాలుగా | తరతరాలుగా | RBP | RB | lex-warawarAlugA|cat-avy|gend-|num-|pers-|case-|vib-0|tam-0_avy|head-warawarAlugA|name-RBP|poslcat-NM | 3 | adv | _ | _ | | 3 | నాటుకుపొయిన | నాటుకొ | VGNF | VM | lex-nAtuko|cat-v|gend-any|num-any|pers-any|case-|vib-i_po_ina|tam-i_po_ina|head-nAtukupoyina|name-VGNF | 4 | nmod | _ | _ | | 4 | అభిప్రాయాలు | అభిప్రాయం | NP | NN | lex-aBiprAyaM|cat-n|gend-|num-pl|pers-|case-d|vib-0|tam-0|head-aBiprAyAlu|name-NP2 | 5 | k1 | _ | _ | | 5 | NULL | NULL | NULL_VGF | VM | name-NULL_VGF | 0 | main | _ | _ | ==== Parsing ==== Nonprojectivities in HyDT-Telugu are very rare. Only 13 of the 5722 chunks in the training+development ICON 2010 version are attached nonprojectively (0.23%). The results of the ICON 2009 NLP tools contest have been published in [[http://ltrc.iiit.ac.in/nlptools2009/CR/intro-husain.pdf|(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 Telugu/coarse-grained results of the four officially ranked systems. ^ Parser (Authors) ^ LAS ^ UAS ^ | Malt (Nivre) | 62.44 | 86.28 | | Mannem | 65.01 | 85.76 | | Hyderabad (Ambati et al.) | 65.01 | 85.25 | | Malt+MST (Zeman) | 56.43 | 81.30 | The results of the ICON 2010 NLP tools contest have been published in [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/toolscontest10-workshoppaper-final.pdf|(Husain et al., 2010)]], page 6. These are the best results for Telugu with fine-grained syntactic tags: ^ Parser (Authors) ^ LAS ^ UAS ^ | Kosaraju et al. | 70.12 | 91.82 | | Attardi et al. | 65.61 | 90.48 | | Kolachina et al. | 68.11 | 90.15 |