Differences

This shows you the differences between two versions of the page.

--- user:zeman:treebanks:eu [2011/11/29 10:42]
zeman Sample.
+++ user:zeman:treebanks:eu [2011/12/13 10:11] (current)
zeman Typo.
@@ Line 5: / Line 5: @@
 ==== Versions ====
-  * CoNLL 2007
+  * CoNLL 2007 (BDT-I)
   * BDT-II (obtained per e-mail in 2011)
@@ Line 29: / Line 29: @@
     * Description of tags and feature values is hard to find; the ''doc/README'' file in the CoNLL 2007 data distribution is not very informative. See below for information obtained per e-mail communication.
     * María Jesús Aranzabe, José Mari Arriola, Aitziber Atutxa, Irene Balza, Larraitz Uria: [[http://ixa.si.ehu.es/Ixa/Argitalpenak/Barne_txostenak/1068549887/publikoak/guia.pdf|Guía para la anotación sintáctica manual de Eus3LB (corpus del euskera anotado a nivel sintáctico, semántico y pragmático)]]. UPV/EHU/LSI/TR 13-2003, Donostia, Spain, 2003.
+    * [[http://www.google.cz/url?sa=t&rct=j&q=adlativo%20direccional%20norantz&source=web&cd=1&ved=0CB0QFjAA&url=http%3A%2F%2Flenguaesp.usal.es%2Fhtml%2Fes%2Fdbfs%2Fdownload.html%3FfileId%3D1118%26_key_%3D248d9f4b64589181dfabafad22b8e483&ei=Qg3VTpKCFpDNswaarJyNDg&usg=AFQjCNEA86oRVR_7sNixk1EKvDFCoSrSsg&sig2=yTsTylb19CsOqsdu-wOtwA&cad=rja|Here]] at the University of Salamanca is a Microsoft Word document in Spanish describing the Basque morphology. It does not mention the treebank but it could help understand some of the tags.
+    * José Ignacio Hualde, Jon Ortiz de Urbina: [[http://books.google.cz/books?id=Kss999lxKm0C&printsec=frontcover&dq=grammar+of+basque&cd=1&redir_esc=y#v=onepage&q&f=false|A Grammar of Basque]]. Mouton de Gruyter, Berlin, 2003. ISBN 3-11-017683-1.
 ==== Domain ====
@@ Line 96: / Line 98: @@
 Main features:
-  * KAS = case (ERG = ergative, ABS = absolutive, DAT = dative...)
+  * KAS = case. Various descriptions of Basque grammar list different numbers of cases and it is not easy to match all of the BDT case tags with them. Some but not all of them are described in the Annex 3 of the technical report mentioned above. The following list gives all case tags occurring in BDT with their frequencies in brackets.
+    * KAS:ABL (984) = ablativo = ablative
+    * KAS:ABS (22805) = absolutivo = absolutive
+    * KAS:ABU (32) = adlativo terminal ("-raino") = "until, as far as" = terminative
+    * KAS:ABZ (27) = adlativo direccional ("-rantz") = "towards" ~ lative?
+    * KAS:ALA (1093) = adlativo = allative
+    * KAS:BNK (13) =? special case of the locative genitive ("-ko", "-eko")
+    * KAS:DAT (1451) = dativo = dative
+    * KAS:DES (181) = destinativo = benefactive ("-entzat")
+    * KAS:DESK (223) =? descriptive locative genitive ("-ko", "-eko"), also frequently used for counted noun after numeral
+    * KAS:EM (705) = multiword token with postposition (e.g. "_gabe", "_arabera", "_batera", "_bezala"...)
+    * KAS:ERG (6059) = ergativo = ergative
+    * KAS:GEL (6259) = genitivo locativo = locative genitive
+    * KAS:GEN (4307) = genitivo de posesión = possessive genitive
+    * KAS:INE (7690) = inesivo = inessive
+    * KAS:INS (1370) = instrumental
+    * KAS:MOT (165) = motivativo = causative
+    * KAS:PAR (930) = partitivo = partitive
+    * KAS:PRO (89) = prolativo = essive
+    * KAS:SOZ (928) = asociativo = comitative
   * ASP = aspect
   * ERL = relation (relative sentence, completive sentence, indirect question...)
+List of all 286 features found in the corpus with frequencies:
+  * ADM:ADIZE	3612
+  * ADM:ADOIN	2919
+  * ADM:PART	14711
+  * ASP:BURU	7491
+  * ASP:EZBU	2421
+  * ASP:GERO	2166
+  * ASP:PNT	6631
+  * BIZ:+	2303
+  * BIZ:-	22116
+  * ENT:???	35
+  * ENT:Erakundea	3499
+  * ENT:Pertsona	4401
+  * ENT:Tokia	3949
+  * ERL:AURK	1264
+  * ERL:BALD	332
+  * ERL:DENB	390
+  * ERL:EMEN	5969
+  * ERL:ERLT	1531
+  * ERL:ESPL	129
+  * ERL:HAUT	408
+  * ERL:HELB	925
+  * ERL:KAUS	864
+  * ERL:KONPL	2614
+  * ERL:KONT	215
+  * ERL:MOD	1152
+  * ERL:MOD/DENB	244
+  * ERL:MOS	146
+  * ERL:ONDO	160
+  * ERL:ZHG	232
+  * HIT:NO	50
+  * HIT:TO	38
+  * IZAUR:+	1499
+  * IZAUR:-	5930
+  * KAS:ABL	984
+  * KAS:ABS	22807
+  * KAS:ABU	32
+  * KAS:ABZ	27
+  * KAS:ALA	1094
+  * KAS:BNK	13
+  * KAS:DAT	1451
+  * KAS:DES	181
+  * KAS:DESK	223
+  * KAS:EM	707
+  * KAS:ERG	6059
+  * KAS:GEL	6266
+  * KAS:GEN	4307
+  * KAS:INE	7693
+  * KAS:INS	1370
+  * KAS:MOT	165
+  * KAS:PAR	930
+  * KAS:PRO	89
+  * KAS:SOZ	928
+  * KLM:AM	80
+  * KLM:HAS	2
+  * MAI:GEHI	38
+  * MAI:IND	36
+  * MAI:KONP	244
+  * MAI:SUP	406
+  * MDN:A1	11766
+  * MDN:A3	107
+  * MDN:A4	1
+  * MDN:A5	282
+  * MDN:B1	6666
+  * MDN:B2	185
+  * MDN:B3	11
+  * MDN:B4	59
+  * MDN:B5A	1
+  * MDN:B5B	27
+  * MDN:B6	1
+  * MDN:B7	79
+  * MDN:B8	38
+  * MDN:C	52
+  * MOD:EGI	2244
+  * MOD:ZIU	126
+  * MTKAT:LAB	16
+  * MTKAT:SIG	696
+  * MTKAT:SNB	22
+  * MUG:M	42116
+  * MUG:MG	8449
+  * MW:B	3615
+  * NEUR:-	193
+  * NMG:MG	1055
+  * NMG:P	2690
+  * NMG:S	2156
+  * NOR:GU	223
+  * NOR:HAIEK	4248
+  * NOR:HI	20
+  * NOR:HURA	14342
+  * NOR:NI	337
+  * NOR:ZU	93
+  * NOR:ZUEK	12
+  * NORI:GURI	124
+  * NORI:HAIEI	306
+  * NORI:HARI	1085
+  * NORI:HIRI-NO	2
+  * NORI:HIRI-TO	5
+  * NORI:NIRI	152
+  * NORI:ZUEI	12
+  * NORI:ZURI	39
+  * NORK:GUK	721
+  * NORK:HAIEK-K	2618
+  * NORK:HARK	5981
+  * NORK:HIK	6
+  * NORK:HIK-NO	10
+  * NORK:HIK-TO	8
+  * NORK:NIK	662
+  * NORK:ZUEK-K	46
+  * NORK:ZUK	208
+  * NUM:P	9347
+  * NUM:PH	172
+  * NUM:S	32570
+  * PER:GU	242
+  * PER:HAIEK	93
+  * PER:HI	14
+  * PER:HURA	1
+  * PER:NI	290
+  * PER:ZU	60
+  * PER:ZUEK	29
+  * PLU:+	149
+  * PLU:-	10257
+  * POS:+	2353
+  * POS:POSAldeko	2
+  * POS:POSAurkako	1
+  * POS:POSGabeko	1
+  * POS:POSInguruko	1
+  * POS:POSKontrako	2
+  * POS:POSaintzinean	1
+  * POS:POSaitzina	2
+  * POS:POSaitzinean	5
+  * POS:POSaitzineko	2
+  * POS:POSaitzinetik	3
+  * POS:POSalboan	2
+  * POS:POSaldamenetik	1
+  * POS:POSalde	38
+  * POS:POSaldean	11
+  * POS:POSaldeaz	1
+  * POS:POSaldeko	37
+  * POS:POSaldera	20
+  * POS:POSalderat	1
+  * POS:POSaldetik	25
+  * POS:POSantzean	1
+  * POS:POSantzeko	9
+  * POS:POSantzekoa	2
+  * POS:POSantzera	3
+  * POS:POSarabera	135
+  * POS:POSaraberako	1
+  * POS:POSarte	82
+  * POS:POSartean	158
+  * POS:POSarteetik	1
+  * POS:POSarteko	108
+  * POS:POSartekoak	1
+  * POS:POSat	6
+  * POS:POSatzean	15
+  * POS:POSatzeko	6
+  * POS:POSatzera	1
+  * POS:POSatzetik	12
+  * POS:POSaurka	103
+  * POS:POSaurkaa	1
+  * POS:POSaurkako	48
+  * POS:POSaurrean	74
+  * POS:POSaurreko	10
+  * POS:POSaurrera	36
+  * POS:POSaurrerako	2
+  * POS:POSaurretik	26
+  * POS:POSazpian	9
+  * POS:POSazpitik	6
+  * POS:POSbaitan	12
+  * POS:POSbarik	2
+  * POS:POSbarna	1
+  * POS:POSbarnean	11
+  * POS:POSbarneko	2
+  * POS:POSbarnera	1
+  * POS:POSbarrena	4
+  * POS:POSbarrenean	1
+  * POS:POSbarru	7
+  * POS:POSbarruan	37
+  * POS:POSbarruetatik	1
+  * POS:POSbarruko	3
+  * POS:POSbarrura	1
+  * POS:POSbarrutik	2
+  * POS:POSbatera	42
+  * POS:POSbatera	1
+  * POS:POSbegira	31
+  * POS:POSbehera	11
+  * POS:POSbestaldean	1
+  * POS:POSbezala	75
+  * POS:POSbezalako	15
+  * POS:POSbezalakoa	1
+  * POS:POSbezalakoen	1
+  * POS:POSbidez	45
+  * POS:POSbila	20
+  * POS:POSbitarte	2
+  * POS:POSbitartean	18
+  * POS:POSbitarteko	5
+  * POS:POSbitarterako	1
+  * POS:POSbitartez	13
+  * POS:POSburuan	7
+  * POS:POSburuz	47
+  * POS:POSburuzko	36
+  * POS:POSeran	1
+  * POS:POSerdian	11
+  * POS:POSerdiko	1
+  * POS:POSerdira	3
+  * POS:POSerditan	1
+  * POS:POSeske	2
+  * POS:POSesker	30
+  * POS:POSesku	12
+  * POS:POSeskuetan	5
+  * POS:POSeskuko	1
+  * POS:POSeskutik	6
+  * POS:POSezean	4
+  * POS:POSgabe	74
+  * POS:POSgabeko	17
+  * POS:POSgain	36
+  * POS:POSgaindi	1
+  * POS:POSgaindiko	1
+  * POS:POSgainean	33
+  * POS:POSgaineko	12
+  * POS:POSgainera	9
+  * POS:POSgainerat	1
+  * POS:POSgainetik	16
+  * POS:POSgero	1
+  * POS:POSgeroztik	18
+  * POS:POSgertu	4
+  * POS:POSgibeleko	1
+  * POS:POSgibeletik	2
+  * POS:POSgisa	34
+  * POS:POSgisako	1
+  * POS:POSgisan	2
+  * POS:POSgisara	1
+  * POS:POSgoiko	1
+  * POS:POSgoitik	1
+  * POS:POSgora	30
+  * POS:POSgorago	1
+  * POS:POSgorako	7
+  * POS:POSgorakoen	1
+  * POS:POShurbil	8
+  * POS:POShurrean	1
+  * POS:POSinguru	16
+  * POS:POSingurua	1
+  * POS:POSinguruan	77
+  * POS:POSinguruetako	1
+  * POS:POSinguruetan	2
+  * POS:POSinguruetara	1
+  * POS:POSinguruko	27
+  * POS:POSingurura	5
+  * POS:POSingururako	1
+  * POS:POSirian	1
+  * POS:POSkanpo	28
+  * POS:POSkanpoko	12
+  * POS:POSkanpora	4
+  * POS:POSkontra	72
+  * POS:POSkontrako	38
+  * POS:POSlanda	7
+  * POS:POSlandara	2
+  * POS:POSlegez	1
+  * POS:POSlekuan	4
+  * POS:POSlepora	1
+  * POS:POSmendean	1
+  * POS:POSmenpe	8
+  * POS:POSmenpera	1
+  * POS:POSmoduan	1
+  * POS:POSmodura	1
+  * POS:POSondoan	19
+  * POS:POSondoko	1
+  * POS:POSondora	1
+  * POS:POSondoren	32
+  * POS:POSondorengo	2
+  * POS:POSondotik	14
+  * POS:POSordez	9
+  * POS:POSostean	17
+  * POS:POSosteko	1
+  * POS:POSpare	1
+  * POS:POSparean	5
+  * POS:POSpareko	2
+  * POS:POSpartean	3
+  * POS:POSpartez	1
+  * POS:POSpean	1
+  * POS:POStruke	9
+  * POS:POSurrun	3
+  * POS:POSurruti	3
+  * POS:POSzai	2
+  * POS:POSzain	12
+  * POS:POSzehar	42
+  * ZENB:-	192
+  * _	36940
 The syntactic guidelines (structure and labels) are described in Spanish in this [[http://ixa.si.ehu.es/Ixa/Argitalpenak/Barne_txostenak/1068549887/publikoak/guia.pdf|technical report]]. See Appendix 3 for some lists of tags.
+Multi-word expressions have been collapsed into one token, using underscore as the joining character (e.g. Espainia_Poliziak, iduri_zait).
 ==== Sample ====
@@ Line 205: / Line 516: @@
 ==== Parsing ====
-Nonprojectivities in GDT are not frequent. Only 823 of the 70223 tokens in the CoNLL 2007 version are attached nonprojectively (1.17%).
+BDT is a mildly nonprojective treebank. 1925 of the 151,604 tokens of combined BDT-II training and test sets are attached nonprojectively (1.27%).
-The results of the CoNLL 2007 shared task are [[http://nextens.uvt.nl/depparse-wiki/AllScores|available online]]. They have been published in [[http://aclweb.org/anthology-new/D/D07/D07-1096.pdf|(Nivre et al., 2007)]]. The evaluation procedure was changed to include punctuation tokens. These are the best results for Greek:
+The results of the CoNLL 2007 shared task are [[http://nextens.uvt.nl/depparse-wiki/AllScores|available online]]. They have been published in [[http://aclweb.org/anthology-new/D/D07/D07-1096.pdf|(Nivre et al., 2007)]]. The evaluation procedure was changed to include punctuation tokens. These are the best results for Basque:
 ^ Parser (Authors) ^ LAS ^ UAS ^
-| Nakagawa | 76.31 | 84.08 |
+| Malt (Nilsson et al.) | 76.94 | 82.84 |
-| Keith Hall et al. | 74.21 | 82.04 |
+| Titov et al. | 75.49 | 81.93 |
-| Carreras | 73.56 | 81.37 |
+| Sagae | 74.64 | 81.19 |
-| Malt (Nilsson et al.) | 74.65 | 81.22 |
+| Carreras | 75.75 | 81.11 |
-| Titov et al. | 73.52 | 81.20 |
+| Nakagawa | 72.56 | 81.04 |
-| Chen | 74.42 | 81.16 |
+| Malt (J. Hall et al.) | 74.99 | 80.61 |
-| Duan | 74.29 | 80.77 |
+| Johansson et al. | 75.08 | 80.43 |
-| Attardi et al. | 73.92 | 80.75 |
-| Malt (J. Hall et al.) | 74.21 | 80.66 |
 The two Malt parser results of 2007 (single malt and blended) are described in [[http://aclweb.org/anthology-new/D/D07/D07-1097.pdf|(Hall et al., 2007)]] and the details about the parser configuration are described [[http://w3.msi.vxu.se/users/jha/conll07/|here]].
+Parsing results on BDT-II have been published in Kepa Bengoetxea, Koldo Gojenola: [[http://aclweb.org/anthology-new/W/W10/W10-1404.pdf|Application of Different Techniques to Dependency Parsing of Basque]]. In: Proceedings of the First Workshop on Statistical Parsing of Morphologically Rich Languages (SPMRL 2010), NAACL Workshop, Los Angeles, California, USA, 2010. They report only Labeled Attachment Score (LAS) and their best system achieved LAS = 78.98%.

[ Back to the navigation ] [ Back to the content ]

Institute of Formal and Applied Linguistics Wiki

Differences