Datasets:
cjvt
/

Modalities:
Text
Languages:
Slovenian
Libraries:
Datasets
License:
Matej Klemen commited on
Commit
a522dc2
1 Parent(s): 43ed7c5

Fix issue with parsing inconsistently formatted normalized words

Browse files
Files changed (2) hide show
  1. README.md +2 -2
  2. janes_tag.py +21 -9
README.md CHANGED
@@ -14,10 +14,10 @@ dataset_info:
14
  sequence: string
15
  splits:
16
  - name: train
17
- num_bytes: 2652674
18
  num_examples: 2957
19
  download_size: 2871765
20
- dataset_size: 2652674
21
  task_categories:
22
  - token-classification
23
  language:
 
14
  sequence: string
15
  splits:
16
  - name: train
17
+ num_bytes: 2653609
18
  num_examples: 2957
19
  download_size: 2871765
20
+ dataset_size: 2653609
21
  task_categories:
22
  - token-classification
23
  language:
janes_tag.py CHANGED
@@ -46,18 +46,30 @@ def word_info(wordlike_tag, _namespace):
46
 
47
  if wordlike_tag.tag in {f"{_namespace}w", f"{_namespace}pc"}:
48
  nes = None
49
- if "lemma" in wordlike_tag.attrib:
50
- words = [wordlike_tag.text.strip()]
51
- lemmas = [wordlike_tag.attrib["lemma"].strip()]
52
- msds = [wordlike_tag.attrib["ana"].strip()]
53
 
54
- # If this happens, the word contains nested words indicating its normalized form
55
- else:
 
56
  words, lemmas, msds = [], [], []
57
  for _child in wordlike_tag:
58
- words.append(_child.attrib["norm"].strip())
59
- lemmas.append(_child.attrib["lemma"].strip())
60
- msds.append(_child.attrib["ana"].strip())
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
61
 
62
  return words, lemmas, msds, nes
63
 
 
46
 
47
  if wordlike_tag.tag in {f"{_namespace}w", f"{_namespace}pc"}:
48
  nes = None
 
 
 
 
49
 
50
+ children = list(iter(wordlike_tag))
51
+ if len(children) > 0:
52
+ # If this happens, the word contains nested words indicating its normalized form
53
  words, lemmas, msds = [], [], []
54
  for _child in wordlike_tag:
55
+ assert _child.tag in {f"{_namespace}w", f"{_namespace}pc"}, _child.tag
56
+
57
+ # Arbitrary words in the text have a normalized form that is formatted inconsistently and so it is
58
+ # unclear how to parse it correctly -> convention: always use information of the normalized words
59
+ if "norm" in _child.attrib:
60
+ words.append(_child.attrib["norm"].strip())
61
+ lemmas.append(_child.attrib["lemma"].strip())
62
+ msds.append(_child.attrib["ana"].strip())
63
+ else:
64
+ # These don't have linguistic annotations ¯\_(ツ)_/¯
65
+ words.append(_child.text.strip())
66
+ lemmas.append(_child.text.strip())
67
+ msds.append("UNK")
68
+
69
+ else:
70
+ words = [wordlike_tag.text.strip()]
71
+ lemmas = [wordlike_tag.attrib["lemma"].strip()]
72
+ msds = [wordlike_tag.attrib["ana"].strip()]
73
 
74
  return words, lemmas, msds, nes
75