Sample data for the DISRPT2019 shared task
The main shared task data will be provided in two formats: with dependency syntax trees / other annotations in CoNLL-U format, and in a tokenization-only 'plain' format.
Files are given in the 10 column, tab delimited CoNLL-U format, with three kinds of lines:
- Token lines (10 columns, word form in column 2) - the last column (CoNLL-U MISC field) may include an annotation
BeginSeg=Yesto indicate the beginning of a discourse unit - this is the annotation systems should predict. Note: Other annotations from the original treebanks may also be present in this field, in which case they will be pipe separated, e.g.BeginSeg=Yes|SpaceAfter=No. - Blank lines between sentences
- Comment lines (some optional) - begin with
#and contain one of three key-value pairs, e.g.:# newdoc id = GUM_academic_art[This is the identifier indicating a new corpus document]# sent_id = GUM_academic_art-1[If the treebank has native sentence IDs, there are included in this way for reference]# text = Aesthetic Appreciation and Spanish Art:[Original, untokenized sentence text - not available for all treebanks]
Example:
# newdoc id = GUM_fiction_veronique
# sent_id = GUM_fiction_veronique-1
# text = The Cost to Be Wise
1 The the DET DT Definite=Def|PronType=Art 2 det _ BeginSeg=Yes
2 Cost cost PROPN NNP Number=Sing 0 root _ _
3 to to PART TO _ 5 mark _ _
4 Be be AUX NNP Number=Sing 5 cop _ _
5 Wise Wise PROPN NNP Number=Sing 2 xcomp _ _
# sent_id = GUM_fiction_veronique-2
# text = Veronique stayed with me that night, lying next to me in my blankets and furs.
1 Veronique Veronique PROPN NNP Number=Sing 2 nsubj _ BeginSeg=Yes
2 stayed stay VERB VBD Mood=Ind|Tense=Past|VerbForm=Fin 0 root _ _
3 with with ADP IN _ 4 case _ _
4 me me PRON PRP Case=Acc|Number=Sing|Person=1|PronType=Prs 2 obl _ _
5 that that DET DT Number=Sing|PronType=Dem 6 det _ _
6 night night NOUN NN Number=Sing 2 obl:tmod _ SpaceAfter=No
7 , , PUNCT , _ 8 punct _ _
8 lying lie VERB VBG VerbForm=Ger 2 advcl _ BeginSeg=Yes
9 next next ADJ JJ Degree=Pos 8 advmod _ _
10 to to ADP TO _ 11 case _ _
11 me me PRON PRP Case=Acc|Number=Sing|Person=1|PronType=Prs 9 obl _ _
12 in in ADP IN _ 14 case _ _
13 my my PRON PRP$ Number=Sing|Person=1|Poss=Yes|PronType=Prs 14 nmod:poss _ _
14 blankets blanket NOUN NNS Number=Plur 8 obl _ _
15 and and CCONJ CC _ 16 cc _ _
16 furs fur NOUN NNS Number=Plur 14 conj _ SpaceAfter=No
17 . . PUNCT . _ 2 punct _ _
# sent_id = GUM_fiction_veronique-3
# text = She didn't sleep, I don't think.
1 She she PRON PRP Case=Nom|Gender=Fem|Number=Sing|Person=3|PronType=Prs 4 nsubj _ BeginSeg=Yes
2 did do AUX VBD Mood=Ind|Tense=Past|VerbForm=Fin 4 aux _ SpaceAfter=No
3 n't n't PART RB Polarity=Neg 4 advmod _ _
...
This format contains one token per line, with a running ID and word form in the first two tab delimited columns, and the possible annotation BeginSeg=Yes in the tenth and last column (if a new segment boundary is present). Document boundaries are indicated by comments, but other annotations are not used:
# newdoc id = GUM_fiction_veronique
1 The _ _ _ _ _ _ _ BeginSeg=Yes
2 Cost _ _ _ _ _ _ _ _
3 to _ _ _ _ _ _ _ _
4 Be _ _ _ _ _ _ _ _
5 Wise _ _ _ _ _ _ _ _
6 Veronique _ _ _ _ _ _ _ BeginSeg=Yes
7 stayed _ _ _ _ _ _ _ _
8 with _ _ _ _ _ _ _ _
9 me _ _ _ _ _ _ _ _
10 that _ _ _ _ _ _ _ _
11 night _ _ _ _ _ _ _ _
12 , _ _ _ _ _ _ _ _
13 lying _ _ _ _ _ _ _ BeginSeg=Yes
14 next _ _ _ _ _ _ _ _
15 to _ _ _ _ _ _ _ _
16 me _ _ _ _ _ _ _ _
17 in _ _ _ _ _ _ _ _
18 my _ _ _ _ _ _ _ _
19 blankets _ _ _ _ _ _ _ _
20 and _ _ _ _ _ _ _ _
21 furs _ _ _ _ _ _ _ _
22 . _ _ _ _ _ _ _ _
23 She _ _ _ _ _ _ _ BeginSeg=Yes
24 did _ _ _ _ _ _ _ _
25 n't _ _ _ _ _ _ _ _
...