Datasets:
Tasks:
Token Classification
Modalities:
Text
Formats:
json
Sub-tasks:
named-entity-recognition
Languages:
Greek
Size:
10K - 100K
Tags:
legal
License:
joelniklaus
commited on
Commit
•
715e3e7
1
Parent(s):
c22cb0e
changed notation scheme to IOB
Browse files- README.md +3 -1
- convert_to_hf_dataset.py +101 -0
- test.jsonl +2 -2
- train.jsonl +2 -2
- validation.jsonl +2 -2
README.md
CHANGED
@@ -3,7 +3,7 @@ annotations_creators:
|
|
3 |
- other
|
4 |
language_creators:
|
5 |
- found
|
6 |
-
|
7 |
- el
|
8 |
license:
|
9 |
- cc-by-nc-sa-4.0
|
@@ -93,6 +93,8 @@ The files contain the following data fields
|
|
93 |
- `PUBLIC-DOCS`: Public Document Reference; any reference to documents or decisions that have been published by a public institution (organization) that are not considered a primary source of legislation (e.g., local decisions, announcements, memorandums, directives).
|
94 |
- `O`: No entity annotation present
|
95 |
|
|
|
|
|
96 |
### Data Splits
|
97 |
|
98 |
The dataset has three splits: *train*, *validation* and *test*.
|
|
|
3 |
- other
|
4 |
language_creators:
|
5 |
- found
|
6 |
+
language:
|
7 |
- el
|
8 |
license:
|
9 |
- cc-by-nc-sa-4.0
|
|
|
93 |
- `PUBLIC-DOCS`: Public Document Reference; any reference to documents or decisions that have been published by a public institution (organization) that are not considered a primary source of legislation (e.g., local decisions, announcements, memorandums, directives).
|
94 |
- `O`: No entity annotation present
|
95 |
|
96 |
+
The final tagset (in IOB notation) is the following: `['O', 'B-ORG', 'I-ORG', 'B-GPE', 'I-GPE', 'B-LEG-REFS', 'I-LEG-REFS', 'B-PUBLIC-DOCS', 'I-PUBLIC-DOCS', 'B-PERSON', 'I-PERSON', 'B-FACILITY', 'I-FACILITY', 'B-LOCATION-UNK', 'I-LOCATION-UNK', 'B-LOCATION-NAT', 'I-LOCATION-NAT']`
|
97 |
+
|
98 |
### Data Splits
|
99 |
|
100 |
The dataset has three splits: *train*, *validation* and *test*.
|
convert_to_hf_dataset.py
ADDED
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
from glob import glob
|
3 |
+
from pathlib import Path
|
4 |
+
|
5 |
+
from typing import List
|
6 |
+
|
7 |
+
import pandas as pd
|
8 |
+
|
9 |
+
from spacy.lang.el import Greek
|
10 |
+
|
11 |
+
pd.set_option('display.max_colwidth', None)
|
12 |
+
pd.set_option('display.max_columns', None)
|
13 |
+
|
14 |
+
base_path = Path("DATASETS/ENTITY RECOGNITION")
|
15 |
+
tokenizer = Greek().tokenizer
|
16 |
+
|
17 |
+
|
18 |
+
# A and D are different government gazettes
|
19 |
+
# A is the general one, publishing standard legislation, and D is meant for legislation on urban planning and such things
|
20 |
+
|
21 |
+
def process_document(ann_file: str, text_file: Path, metadata: dict, tokenizer) -> List[dict]:
|
22 |
+
"""Processes one document (.ann file and .txt file) and returns a list of annotated sentences"""
|
23 |
+
# read the ann file into a df
|
24 |
+
ann_df = pd.read_csv(ann_file, sep="\t", header=None, names=["id", "entity_with_span", "entity_text"])
|
25 |
+
sentences = [sent for sent in text_file.read_text().split("\n") if sent] # remove empty sentences
|
26 |
+
|
27 |
+
# split into individual columns
|
28 |
+
ann_df[["entity", "start", "end"]] = ann_df["entity_with_span"].str.split(" ", expand=True)
|
29 |
+
ann_df.start = ann_df.start.astype(int)
|
30 |
+
ann_df.end = ann_df.end.astype(int)
|
31 |
+
|
32 |
+
not_found_entities = 0
|
33 |
+
annotated_sentences = []
|
34 |
+
current_start_index = 0
|
35 |
+
for sentence in sentences:
|
36 |
+
ann_sent = {**metadata}
|
37 |
+
|
38 |
+
doc = tokenizer(sentence)
|
39 |
+
doc_start_index = current_start_index
|
40 |
+
doc_end_index = current_start_index + len(sentence)
|
41 |
+
current_start_index = doc_end_index + 1
|
42 |
+
|
43 |
+
relevant_annotations = ann_df[(ann_df.start >= doc_start_index) & (ann_df.end <= doc_end_index)]
|
44 |
+
for _, row in relevant_annotations.iterrows():
|
45 |
+
sent_start_index = row["start"] - doc_start_index
|
46 |
+
sent_end_index = row["end"] - doc_start_index
|
47 |
+
char_span = doc.char_span(sent_start_index, sent_end_index, label=row["entity"], alignment_mode="expand")
|
48 |
+
# ent_span = Span(doc, char_span.start, char_span.end, row["entity"])
|
49 |
+
if char_span:
|
50 |
+
doc.set_ents([char_span])
|
51 |
+
else:
|
52 |
+
not_found_entities += 1
|
53 |
+
print(f"Could not find entity `{row['entity_text']}` in sentence `{sentence}`")
|
54 |
+
|
55 |
+
ann_sent["words"] = [str(tok) for tok in doc]
|
56 |
+
ann_sent["ner"] = [tok.ent_iob_ + "-" + tok.ent_type_ if tok.ent_type_ else "O" for tok in doc]
|
57 |
+
|
58 |
+
annotated_sentences.append(ann_sent)
|
59 |
+
|
60 |
+
print(f"Did not find entities in {not_found_entities} cases")
|
61 |
+
return annotated_sentences
|
62 |
+
|
63 |
+
|
64 |
+
def read_to_df(split):
|
65 |
+
"""Reads the different documents and saves metadata"""
|
66 |
+
ann_files = glob(str(base_path / split / "ANN" / "*/*/*.ann"))
|
67 |
+
sentences = []
|
68 |
+
for ann_file in ann_files:
|
69 |
+
path = Path(ann_file)
|
70 |
+
year = path.parent.stem
|
71 |
+
file_name = path.stem
|
72 |
+
_, gazette, gazette_number, _, date = tuple(file_name.split(' '))
|
73 |
+
text_file = base_path / split / "TXT" / f"{gazette}/{year}/{file_name}.txt"
|
74 |
+
metadata = {
|
75 |
+
"date": date,
|
76 |
+
"gazette": gazette,
|
77 |
+
# "gazette_number": gazette_number,
|
78 |
+
}
|
79 |
+
sentences.extend(process_document(ann_file, text_file, metadata, tokenizer))
|
80 |
+
return pd.DataFrame(sentences)
|
81 |
+
|
82 |
+
|
83 |
+
splits = ["TRAIN", "VALIDATION", "TEST"]
|
84 |
+
train = read_to_df("TRAIN")
|
85 |
+
validation = read_to_df("VALIDATION")
|
86 |
+
test = read_to_df("TEST")
|
87 |
+
|
88 |
+
df = pd.concat([train, validation, test])
|
89 |
+
print(f"The final tagset (in IOB notation) is the following: `{list(df.ner.explode().unique())}`")
|
90 |
+
|
91 |
+
|
92 |
+
# save splits
|
93 |
+
def save_splits_to_jsonl(config_name):
|
94 |
+
# save to jsonl files for huggingface
|
95 |
+
if config_name: os.makedirs(config_name, exist_ok=True)
|
96 |
+
train.to_json(os.path.join(config_name, "train.jsonl"), lines=True, orient="records", force_ascii=False)
|
97 |
+
validation.to_json(os.path.join(config_name, "validation.jsonl"), lines=True, orient="records", force_ascii=False)
|
98 |
+
test.to_json(os.path.join(config_name, "test.jsonl"), lines=True, orient="records", force_ascii=False)
|
99 |
+
|
100 |
+
|
101 |
+
save_splits_to_jsonl("")
|
test.jsonl
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5ad0654caac860a9d6d3d7d7834f913afaafc6cc5506c379c50c3d86ef12d559
|
3 |
+
size 2278020
|
train.jsonl
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1e0a17fd23cf770e960e14d4972c3245bc9a8f65cf616a8db4a7daaf4fd0e519
|
3 |
+
size 9226475
|
validation.jsonl
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d91aab2c40560e498d8948fd4133eda83f519ebb2575d59901b343b8f91dda15
|
3 |
+
size 2445108
|