notebook: add example for the creation of Wikipedia filtered-out GermEval 2014 dataset
Browse files- CreateDataset.ipynb +530 -0
CreateDataset.ipynb
ADDED
@@ -0,0 +1,530 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"cell_type": "code",
|
5 |
+
"execution_count": 1,
|
6 |
+
"id": "1a6ca1fd-66c6-4da6-af7f-35879fb663ba",
|
7 |
+
"metadata": {},
|
8 |
+
"outputs": [
|
9 |
+
{
|
10 |
+
"name": "stdout",
|
11 |
+
"output_type": "stream",
|
12 |
+
"text": [
|
13 |
+
"Requirement already satisfied: tldextract in /home/stefan/.venvs/flair/lib/python3.12/site-packages (5.1.2)\n",
|
14 |
+
"Requirement already satisfied: idna in /home/stefan/.venvs/flair/lib/python3.12/site-packages (from tldextract) (3.7)\n",
|
15 |
+
"Requirement already satisfied: requests>=2.1.0 in /home/stefan/.venvs/flair/lib/python3.12/site-packages (from tldextract) (2.32.2)\n",
|
16 |
+
"Requirement already satisfied: requests-file>=1.4 in /home/stefan/.venvs/flair/lib/python3.12/site-packages (from tldextract) (2.1.0)\n",
|
17 |
+
"Requirement already satisfied: filelock>=3.0.8 in /home/stefan/.venvs/flair/lib/python3.12/site-packages (from tldextract) (3.13.1)\n",
|
18 |
+
"Requirement already satisfied: charset-normalizer<4,>=2 in /home/stefan/.venvs/flair/lib/python3.12/site-packages (from requests>=2.1.0->tldextract) (3.3.2)\n",
|
19 |
+
"Requirement already satisfied: urllib3<3,>=1.21.1 in /home/stefan/.venvs/flair/lib/python3.12/site-packages (from requests>=2.1.0->tldextract) (1.26.18)\n",
|
20 |
+
"Requirement already satisfied: certifi>=2017.4.17 in /home/stefan/.venvs/flair/lib/python3.12/site-packages (from requests>=2.1.0->tldextract) (2024.2.2)\n"
|
21 |
+
]
|
22 |
+
}
|
23 |
+
],
|
24 |
+
"source": [
|
25 |
+
"!pip3 install tldextract"
|
26 |
+
]
|
27 |
+
},
|
28 |
+
{
|
29 |
+
"cell_type": "code",
|
30 |
+
"execution_count": 2,
|
31 |
+
"id": "4906d36b-6a06-4987-b032-4de6a11bfc65",
|
32 |
+
"metadata": {},
|
33 |
+
"outputs": [],
|
34 |
+
"source": [
|
35 |
+
"import random\n",
|
36 |
+
"import tldextract\n",
|
37 |
+
"\n",
|
38 |
+
"from collections import Counter\n",
|
39 |
+
"from tabulate import tabulate"
|
40 |
+
]
|
41 |
+
},
|
42 |
+
{
|
43 |
+
"cell_type": "code",
|
44 |
+
"execution_count": 3,
|
45 |
+
"id": "0107c076-c40a-41ae-bf51-92ccddaf2e61",
|
46 |
+
"metadata": {},
|
47 |
+
"outputs": [],
|
48 |
+
"source": [
|
49 |
+
"dataset_splits = {\n",
|
50 |
+
" \"train\": \"./original/NER-de-train.tsv\",\n",
|
51 |
+
" \"dev\": \"./original/NER-de-dev.tsv\",\n",
|
52 |
+
" \"test\": \"./original/NER-de-test.tsv\",\n",
|
53 |
+
"}"
|
54 |
+
]
|
55 |
+
},
|
56 |
+
{
|
57 |
+
"cell_type": "markdown",
|
58 |
+
"id": "b11a81f2-0e71-4bbc-9538-b1f3295c4b38",
|
59 |
+
"metadata": {},
|
60 |
+
"source": [
|
61 |
+
"# Dataset Stats"
|
62 |
+
]
|
63 |
+
},
|
64 |
+
{
|
65 |
+
"cell_type": "code",
|
66 |
+
"execution_count": 4,
|
67 |
+
"id": "530c2634-19ee-4814-842e-b834024fa8c9",
|
68 |
+
"metadata": {},
|
69 |
+
"outputs": [
|
70 |
+
{
|
71 |
+
"name": "stdout",
|
72 |
+
"output_type": "stream",
|
73 |
+
"text": [
|
74 |
+
"GermEval 2014 Dataset Stats for train split:\n",
|
75 |
+
"| TLD | Number of examples (Percentage) |\n",
|
76 |
+
"|----------------------|-----------------------------------|\n",
|
77 |
+
"| wikipedia.org | 12007 (50.03%) |\n",
|
78 |
+
"| welt.de | 662 (2.76%) |\n",
|
79 |
+
"| spiegel.de | 512 (2.13%) |\n",
|
80 |
+
"| tagesspiegel.de | 424 (1.77%) |\n",
|
81 |
+
"| handelsblatt.com | 369 (1.54%) |\n",
|
82 |
+
"| fr-aktuell.de | 344 (1.43%) |\n",
|
83 |
+
"| sueddeutsche.de | 308 (1.28%) |\n",
|
84 |
+
"| abendblatt.de | 283 (1.18%) |\n",
|
85 |
+
"| berlinonline.de | 255 (1.06%) |\n",
|
86 |
+
"| szon.de | 249 (1.04%) |\n",
|
87 |
+
"| n-tv.de | 195 (0.81%) |\n",
|
88 |
+
"| yahoo.com | 192 (0.8%) |\n",
|
89 |
+
"| feedsportal.com | 173 (0.72%) |\n",
|
90 |
+
"| ngz-online.de | 173 (0.72%) |\n",
|
91 |
+
"| faz.net | 156 (0.65%) |\n",
|
92 |
+
"| nzz.ch | 146 (0.61%) |\n",
|
93 |
+
"| morgenweb.de | 134 (0.56%) |\n",
|
94 |
+
"| rp-online.de | 132 (0.55%) |\n",
|
95 |
+
"| gea.de | 131 (0.55%) |\n",
|
96 |
+
"| sat1.de | 126 (0.53%) |\n",
|
97 |
+
"| tagesschau.de | 124 (0.52%) |\n",
|
98 |
+
"| pnp.de | 101 (0.42%) |\n",
|
99 |
+
"| orf.at | 98 (0.41%) |\n",
|
100 |
+
"| n24.de | 98 (0.41%) |\n",
|
101 |
+
"| finanznachrichten.de | 91 (0.38%) |\n",
|
102 |
+
"\n",
|
103 |
+
"GermEval 2014 Dataset Stats for dev split:\n",
|
104 |
+
"| TLD | Number of examples (Percentage) |\n",
|
105 |
+
"|----------------------|-----------------------------------|\n",
|
106 |
+
"| wikipedia.org | 1119 (50.86%) |\n",
|
107 |
+
"| welt.de | 46 (2.09%) |\n",
|
108 |
+
"| spiegel.de | 43 (1.95%) |\n",
|
109 |
+
"| fr-aktuell.de | 38 (1.73%) |\n",
|
110 |
+
"| tagesspiegel.de | 37 (1.68%) |\n",
|
111 |
+
"| handelsblatt.com | 35 (1.59%) |\n",
|
112 |
+
"| sueddeutsche.de | 28 (1.27%) |\n",
|
113 |
+
"| szon.de | 25 (1.14%) |\n",
|
114 |
+
"| feedsportal.com | 24 (1.09%) |\n",
|
115 |
+
"| berlinonline.de | 22 (1.0%) |\n",
|
116 |
+
"| rp-online.de | 21 (0.95%) |\n",
|
117 |
+
"| abendblatt.de | 20 (0.91%) |\n",
|
118 |
+
"| ngz-online.de | 19 (0.86%) |\n",
|
119 |
+
"| n-tv.de | 18 (0.82%) |\n",
|
120 |
+
"| yahoo.com | 15 (0.68%) |\n",
|
121 |
+
"| sat1.de | 15 (0.68%) |\n",
|
122 |
+
"| orf.at | 13 (0.59%) |\n",
|
123 |
+
"| finanznachrichten.de | 13 (0.59%) |\n",
|
124 |
+
"| tagesschau.de | 13 (0.59%) |\n",
|
125 |
+
"| nzz.ch | 12 (0.55%) |\n",
|
126 |
+
"| faz.net | 12 (0.55%) |\n",
|
127 |
+
"| morgenweb.de | 12 (0.55%) |\n",
|
128 |
+
"| 20min.ch | 11 (0.5%) |\n",
|
129 |
+
"| pnp.de | 11 (0.5%) |\n",
|
130 |
+
"| focus.de | 10 (0.45%) |\n",
|
131 |
+
"\n",
|
132 |
+
"GermEval 2014 Dataset Stats for test split:\n",
|
133 |
+
"| TLD | Number of examples (Percentage) |\n",
|
134 |
+
"|----------------------|-----------------------------------|\n",
|
135 |
+
"| wikipedia.org | 2547 (49.94%) |\n",
|
136 |
+
"| welt.de | 139 (2.73%) |\n",
|
137 |
+
"| spiegel.de | 88 (1.73%) |\n",
|
138 |
+
"| tagesspiegel.de | 86 (1.69%) |\n",
|
139 |
+
"| handelsblatt.com | 84 (1.65%) |\n",
|
140 |
+
"| sueddeutsche.de | 78 (1.53%) |\n",
|
141 |
+
"| abendblatt.de | 72 (1.41%) |\n",
|
142 |
+
"| fr-aktuell.de | 62 (1.22%) |\n",
|
143 |
+
"| berlinonline.de | 59 (1.16%) |\n",
|
144 |
+
"| szon.de | 57 (1.12%) |\n",
|
145 |
+
"| feedsportal.com | 52 (1.02%) |\n",
|
146 |
+
"| n-tv.de | 47 (0.92%) |\n",
|
147 |
+
"| sat1.de | 42 (0.82%) |\n",
|
148 |
+
"| nzz.ch | 39 (0.76%) |\n",
|
149 |
+
"| yahoo.com | 38 (0.75%) |\n",
|
150 |
+
"| ngz-online.de | 37 (0.73%) |\n",
|
151 |
+
"| faz.net | 37 (0.73%) |\n",
|
152 |
+
"| morgenweb.de | 36 (0.71%) |\n",
|
153 |
+
"| taz.de | 28 (0.55%) |\n",
|
154 |
+
"| finanznachrichten.de | 25 (0.49%) |\n",
|
155 |
+
"| tagesschau.de | 24 (0.47%) |\n",
|
156 |
+
"| gea.de | 24 (0.47%) |\n",
|
157 |
+
"| bernerzeitung.ch | 23 (0.45%) |\n",
|
158 |
+
"| ftd.de | 22 (0.43%) |\n",
|
159 |
+
"| orf.at | 21 (0.41%) |\n",
|
160 |
+
"\n"
|
161 |
+
]
|
162 |
+
}
|
163 |
+
],
|
164 |
+
"source": [
|
165 |
+
"def print_stats(dataset_split, dataset_path, limit=25):\n",
|
166 |
+
" hostname_counter = Counter()\n",
|
167 |
+
"\n",
|
168 |
+
" with open(dataset_path, \"rt\") as f_p:\n",
|
169 |
+
" for line in f_p:\n",
|
170 |
+
" if not line.startswith(\"#\"):\n",
|
171 |
+
" continue\n",
|
172 |
+
" \n",
|
173 |
+
" current_url = line.split(\"\\t\")[1].split(\" \")[0]\n",
|
174 |
+
" \n",
|
175 |
+
" ext = tldextract.extract(current_url)\n",
|
176 |
+
" \n",
|
177 |
+
" hostname = ext.registered_domain\n",
|
178 |
+
" \n",
|
179 |
+
" hostname_counter[hostname] += 1\n",
|
180 |
+
"\n",
|
181 |
+
" # Print nice table\n",
|
182 |
+
" headers = [\"TLD\", \"Number of examples (Percentage)\"]\n",
|
183 |
+
"\n",
|
184 |
+
" table = []\n",
|
185 |
+
"\n",
|
186 |
+
" total_examples = sum(hostname_counter.values())\n",
|
187 |
+
" \n",
|
188 |
+
" for tld_name, examples in hostname_counter.most_common(limit):\n",
|
189 |
+
" current_percentage = round(examples / total_examples * 100, 2)\n",
|
190 |
+
" table.append([tld_name, f\"{examples} ({current_percentage}%)\"])\n",
|
191 |
+
"\n",
|
192 |
+
" print(tabulate(table, headers=headers, tablefmt=\"github\"))\n",
|
193 |
+
"\n",
|
194 |
+
"for dataset_split in dataset_splits.keys():\n",
|
195 |
+
" print(f\"GermEval 2014 Dataset Stats for {dataset_split} split:\")\n",
|
196 |
+
" print_stats(dataset_split, dataset_splits[dataset_split])\n",
|
197 |
+
" print(\"\")"
|
198 |
+
]
|
199 |
+
},
|
200 |
+
{
|
201 |
+
"cell_type": "markdown",
|
202 |
+
"id": "d97e3e86-ae0e-4a54-af15-de195608a071",
|
203 |
+
"metadata": {},
|
204 |
+
"source": [
|
205 |
+
"# Generate New Dataset Split without Wikipedia"
|
206 |
+
]
|
207 |
+
},
|
208 |
+
{
|
209 |
+
"cell_type": "code",
|
210 |
+
"execution_count": 5,
|
211 |
+
"id": "4a081af0-f2fe-4e1a-ae56-da10b2bd7be1",
|
212 |
+
"metadata": {},
|
213 |
+
"outputs": [
|
214 |
+
{
|
215 |
+
"name": "stdout",
|
216 |
+
"output_type": "stream",
|
217 |
+
"text": [
|
218 |
+
"Writing out train split...\n",
|
219 |
+
"Writing out dev split...\n",
|
220 |
+
"Writing out test split...\n"
|
221 |
+
]
|
222 |
+
}
|
223 |
+
],
|
224 |
+
"source": [
|
225 |
+
"def filter_out_wikipedia(dataset_split):\n",
|
226 |
+
" with open(dataset_splits[dataset_split], \"rt\") as f_p:\n",
|
227 |
+
" all_sentences = []\n",
|
228 |
+
" current_sentence = []\n",
|
229 |
+
" for line in f_p:\n",
|
230 |
+
" line = line.strip()\n",
|
231 |
+
" if not line:\n",
|
232 |
+
" # We found new sentence, yeah!\n",
|
233 |
+
"\n",
|
234 |
+
" if len(current_sentence) == 0:\n",
|
235 |
+
" continue\n",
|
236 |
+
" \n",
|
237 |
+
" all_sentences.append(current_sentence)\n",
|
238 |
+
" current_sentence = []\n",
|
239 |
+
" continue\n",
|
240 |
+
"\n",
|
241 |
+
" current_sentence.append(line)\n",
|
242 |
+
"\n",
|
243 |
+
" if len(current_sentence) > 0:\n",
|
244 |
+
" all_sentences.append(current_sentence)\n",
|
245 |
+
" \n",
|
246 |
+
" with open(f\"NER-de-without-wikipedia-{dataset_split}.tsv\", \"wt\") as f_out:\n",
|
247 |
+
" for sentence in all_sentences:\n",
|
248 |
+
"\n",
|
249 |
+
" header = sentence[0]\n",
|
250 |
+
" assert header.startswith(\"#\")\n",
|
251 |
+
"\n",
|
252 |
+
" current_url = header.split(\"\\t\")[1].split(\" \")[0]\n",
|
253 |
+
" ext = tldextract.extract(current_url)\n",
|
254 |
+
" hostname = ext.registered_domain\n",
|
255 |
+
"\n",
|
256 |
+
" if hostname == \"wikipedia.org\":\n",
|
257 |
+
" continue\n",
|
258 |
+
"\n",
|
259 |
+
" f_out.write(\"\\n\".join(sentence) + \"\\n\\n\")\n",
|
260 |
+
"\n",
|
261 |
+
"for dataset_split in dataset_splits.keys():\n",
|
262 |
+
" print(f\"Writing out {dataset_split} split...\")\n",
|
263 |
+
" filter_out_wikipedia(dataset_split)"
|
264 |
+
]
|
265 |
+
},
|
266 |
+
{
|
267 |
+
"cell_type": "markdown",
|
268 |
+
"id": "84321c99-7085-467d-8d9c-7966bddd21de",
|
269 |
+
"metadata": {},
|
270 |
+
"source": [
|
271 |
+
"# Load with Flair"
|
272 |
+
]
|
273 |
+
},
|
274 |
+
{
|
275 |
+
"cell_type": "code",
|
276 |
+
"execution_count": 6,
|
277 |
+
"id": "be671ee3-e562-43a5-8734-144b3a129993",
|
278 |
+
"metadata": {},
|
279 |
+
"outputs": [
|
280 |
+
{
|
281 |
+
"name": "stderr",
|
282 |
+
"output_type": "stream",
|
283 |
+
"text": [
|
284 |
+
"2024-05-29 17:08:26.593572: I external/local_tsl/tsl/cuda/cudart_stub.cc:32] Could not find cuda drivers on your machine, GPU will not be used.\n",
|
285 |
+
"2024-05-29 17:08:26.597226: I external/local_tsl/tsl/cuda/cudart_stub.cc:32] Could not find cuda drivers on your machine, GPU will not be used.\n",
|
286 |
+
"2024-05-29 17:08:26.644331: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
|
287 |
+
"To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
|
288 |
+
"2024-05-29 17:08:27.578466: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n"
|
289 |
+
]
|
290 |
+
}
|
291 |
+
],
|
292 |
+
"source": [
|
293 |
+
"import flair\n",
|
294 |
+
"\n",
|
295 |
+
"from flair.datasets import NER_GERMAN_GERMEVAL\n",
|
296 |
+
"from flair.datasets.sequence_labeling import ColumnCorpus\n",
|
297 |
+
"from flair.file_utils import cached_path\n",
|
298 |
+
"\n",
|
299 |
+
"from pathlib import Path\n",
|
300 |
+
"from typing import Optional, Union\n",
|
301 |
+
"\n",
|
302 |
+
"\n",
|
303 |
+
"class NER_GERMEVAL_2014_NO_WIKIPEDIA(ColumnCorpus):\n",
|
304 |
+
" def __init__(\n",
|
305 |
+
" self,\n",
|
306 |
+
" base_path: Optional[Union[str, Path]] = None,\n",
|
307 |
+
" in_memory: bool = True,\n",
|
308 |
+
" **corpusargs,\n",
|
309 |
+
" ) -> None:\n",
|
310 |
+
" base_path = flair.cache_root / \"datasets\" if not base_path else Path(base_path)\n",
|
311 |
+
" dataset_name = self.__class__.__name__.lower()\n",
|
312 |
+
" data_folder = base_path / dataset_name\n",
|
313 |
+
" data_path = flair.cache_root / \"datasets\" / dataset_name\n",
|
314 |
+
"\n",
|
315 |
+
" column_format = {1: \"text\", 2: \"ner\"}\n",
|
316 |
+
"\n",
|
317 |
+
" #hf_download_path = \"https://huggingface.co/datasets/stefan-it/germeval14_no_wikipedia/resolve/main\"\n",
|
318 |
+
"\n",
|
319 |
+
" #for split in [\"train\", \"dev\", \"test\"]:\n",
|
320 |
+
" # cached_path(f\"{hf_download_path}/NER-de-without-wikipedia-{split}.tsv\", data_path)\n",
|
321 |
+
" \n",
|
322 |
+
" super().__init__(\n",
|
323 |
+
" \"./\", #data_folder,\n",
|
324 |
+
" column_format = {0: \"text\", 1: \"ner\"},\n",
|
325 |
+
" column_delimiter=\"\\t\",\n",
|
326 |
+
" document_separator_token=\"-DOCSTART-\",\n",
|
327 |
+
" in_memory=in_memory,\n",
|
328 |
+
" comment_symbol=\"# \",\n",
|
329 |
+
" **corpusargs,\n",
|
330 |
+
" )"
|
331 |
+
]
|
332 |
+
},
|
333 |
+
{
|
334 |
+
"cell_type": "markdown",
|
335 |
+
"id": "7c9d0275-dc4e-4ca0-9a49-b582a42d6bca",
|
336 |
+
"metadata": {},
|
337 |
+
"source": [
|
338 |
+
"# New Corpus Stats"
|
339 |
+
]
|
340 |
+
},
|
341 |
+
{
|
342 |
+
"cell_type": "code",
|
343 |
+
"execution_count": 7,
|
344 |
+
"id": "8b1a34ab-42c5-4e72-a63e-823f7f2b3cbe",
|
345 |
+
"metadata": {},
|
346 |
+
"outputs": [
|
347 |
+
{
|
348 |
+
"name": "stdout",
|
349 |
+
"output_type": "stream",
|
350 |
+
"text": [
|
351 |
+
"2024-05-29 17:08:30,120 Reading data from .\n",
|
352 |
+
"2024-05-29 17:08:30,121 Train: NER-de-without-wikipedia-train.tsv\n",
|
353 |
+
"2024-05-29 17:08:30,123 Dev: NER-de-without-wikipedia-dev.tsv\n",
|
354 |
+
"2024-05-29 17:08:30,124 Test: NER-de-without-wikipedia-test.tsv\n",
|
355 |
+
"2024-05-29 17:08:34,319 Reading data from /home/stefan/.flair/datasets/ner_german_germeval\n",
|
356 |
+
"2024-05-29 17:08:34,320 Train: /home/stefan/.flair/datasets/ner_german_germeval/train.tsv\n",
|
357 |
+
"2024-05-29 17:08:34,321 Dev: /home/stefan/.flair/datasets/ner_german_germeval/dev.tsv\n",
|
358 |
+
"2024-05-29 17:08:34,321 Test: /home/stefan/.flair/datasets/ner_german_germeval/test.tsv\n"
|
359 |
+
]
|
360 |
+
}
|
361 |
+
],
|
362 |
+
"source": [
|
363 |
+
"corpus = NER_GERMEVAL_2014_NO_WIKIPEDIA()\n",
|
364 |
+
"original_corpus = NER_GERMAN_GERMEVAL()"
|
365 |
+
]
|
366 |
+
},
|
367 |
+
{
|
368 |
+
"cell_type": "code",
|
369 |
+
"execution_count": 8,
|
370 |
+
"id": "86047123-c882-433b-8d19-92626e878afd",
|
371 |
+
"metadata": {},
|
372 |
+
"outputs": [
|
373 |
+
{
|
374 |
+
"name": "stdout",
|
375 |
+
"output_type": "stream",
|
376 |
+
"text": [
|
377 |
+
"Original GermEval 2014 stats: Corpus: 24000 train + 2200 dev + 5100 test sentences\n",
|
378 |
+
"Filtered-out GermEval 2014 stats: Corpus: 11993 train + 1081 dev + 2553 test sentences\n"
|
379 |
+
]
|
380 |
+
}
|
381 |
+
],
|
382 |
+
"source": [
|
383 |
+
"print(\"Original GermEval 2014 stats:\", str(original_corpus))\n",
|
384 |
+
"print(\"Filtered-out GermEval 2014 stats:\", str(corpus))"
|
385 |
+
]
|
386 |
+
},
|
387 |
+
{
|
388 |
+
"cell_type": "code",
|
389 |
+
"execution_count": 9,
|
390 |
+
"id": "9e6ffb09-b1c0-4559-b954-df8373492876",
|
391 |
+
"metadata": {},
|
392 |
+
"outputs": [],
|
393 |
+
"source": [
|
394 |
+
"new_dataset_splits = {\n",
|
395 |
+
" \"train\": \"./NER-de-without-wikipedia-train.tsv\",\n",
|
396 |
+
" \"dev\": \"./NER-de-without-wikipedia-dev.tsv\",\n",
|
397 |
+
" \"test\": \"./NER-de-without-wikipedia-test.tsv\",\n",
|
398 |
+
"}"
|
399 |
+
]
|
400 |
+
},
|
401 |
+
{
|
402 |
+
"cell_type": "code",
|
403 |
+
"execution_count": 11,
|
404 |
+
"id": "9ea60c74-70e2-473c-96c7-b6cdfc601297",
|
405 |
+
"metadata": {},
|
406 |
+
"outputs": [
|
407 |
+
{
|
408 |
+
"name": "stdout",
|
409 |
+
"output_type": "stream",
|
410 |
+
"text": [
|
411 |
+
"New GermEval 2014 Dataset Stats for train split:\n",
|
412 |
+
"| TLD | Number of examples (Percentage) |\n",
|
413 |
+
"|----------------------|-----------------------------------|\n",
|
414 |
+
"| welt.de | 662 (5.52%) |\n",
|
415 |
+
"| spiegel.de | 512 (4.27%) |\n",
|
416 |
+
"| tagesspiegel.de | 424 (3.54%) |\n",
|
417 |
+
"| handelsblatt.com | 369 (3.08%) |\n",
|
418 |
+
"| fr-aktuell.de | 344 (2.87%) |\n",
|
419 |
+
"| sueddeutsche.de | 308 (2.57%) |\n",
|
420 |
+
"| abendblatt.de | 283 (2.36%) |\n",
|
421 |
+
"| berlinonline.de | 255 (2.13%) |\n",
|
422 |
+
"| szon.de | 249 (2.08%) |\n",
|
423 |
+
"| n-tv.de | 195 (1.63%) |\n",
|
424 |
+
"| yahoo.com | 192 (1.6%) |\n",
|
425 |
+
"| feedsportal.com | 173 (1.44%) |\n",
|
426 |
+
"| ngz-online.de | 173 (1.44%) |\n",
|
427 |
+
"| faz.net | 156 (1.3%) |\n",
|
428 |
+
"| nzz.ch | 146 (1.22%) |\n",
|
429 |
+
"| morgenweb.de | 134 (1.12%) |\n",
|
430 |
+
"| rp-online.de | 132 (1.1%) |\n",
|
431 |
+
"| gea.de | 131 (1.09%) |\n",
|
432 |
+
"| sat1.de | 126 (1.05%) |\n",
|
433 |
+
"| tagesschau.de | 124 (1.03%) |\n",
|
434 |
+
"| pnp.de | 101 (0.84%) |\n",
|
435 |
+
"| orf.at | 98 (0.82%) |\n",
|
436 |
+
"| n24.de | 98 (0.82%) |\n",
|
437 |
+
"| finanznachrichten.de | 91 (0.76%) |\n",
|
438 |
+
"| taz.de | 91 (0.76%) |\n",
|
439 |
+
"\n",
|
440 |
+
"New GermEval 2014 Dataset Stats for dev split:\n",
|
441 |
+
"| TLD | Number of examples (Percentage) |\n",
|
442 |
+
"|----------------------|-----------------------------------|\n",
|
443 |
+
"| welt.de | 46 (4.26%) |\n",
|
444 |
+
"| spiegel.de | 43 (3.98%) |\n",
|
445 |
+
"| fr-aktuell.de | 38 (3.52%) |\n",
|
446 |
+
"| tagesspiegel.de | 37 (3.42%) |\n",
|
447 |
+
"| handelsblatt.com | 35 (3.24%) |\n",
|
448 |
+
"| sueddeutsche.de | 28 (2.59%) |\n",
|
449 |
+
"| szon.de | 25 (2.31%) |\n",
|
450 |
+
"| feedsportal.com | 24 (2.22%) |\n",
|
451 |
+
"| berlinonline.de | 22 (2.04%) |\n",
|
452 |
+
"| rp-online.de | 21 (1.94%) |\n",
|
453 |
+
"| abendblatt.de | 20 (1.85%) |\n",
|
454 |
+
"| ngz-online.de | 19 (1.76%) |\n",
|
455 |
+
"| n-tv.de | 18 (1.67%) |\n",
|
456 |
+
"| yahoo.com | 15 (1.39%) |\n",
|
457 |
+
"| sat1.de | 15 (1.39%) |\n",
|
458 |
+
"| orf.at | 13 (1.2%) |\n",
|
459 |
+
"| finanznachrichten.de | 13 (1.2%) |\n",
|
460 |
+
"| tagesschau.de | 13 (1.2%) |\n",
|
461 |
+
"| nzz.ch | 12 (1.11%) |\n",
|
462 |
+
"| faz.net | 12 (1.11%) |\n",
|
463 |
+
"| morgenweb.de | 12 (1.11%) |\n",
|
464 |
+
"| 20min.ch | 11 (1.02%) |\n",
|
465 |
+
"| pnp.de | 11 (1.02%) |\n",
|
466 |
+
"| focus.de | 10 (0.93%) |\n",
|
467 |
+
"| ftd.de | 9 (0.83%) |\n",
|
468 |
+
"\n",
|
469 |
+
"New GermEval 2014 Dataset Stats for test split:\n",
|
470 |
+
"| TLD | Number of examples (Percentage) |\n",
|
471 |
+
"|----------------------|-----------------------------------|\n",
|
472 |
+
"| welt.de | 139 (5.44%) |\n",
|
473 |
+
"| spiegel.de | 88 (3.45%) |\n",
|
474 |
+
"| tagesspiegel.de | 86 (3.37%) |\n",
|
475 |
+
"| handelsblatt.com | 84 (3.29%) |\n",
|
476 |
+
"| sueddeutsche.de | 78 (3.06%) |\n",
|
477 |
+
"| abendblatt.de | 72 (2.82%) |\n",
|
478 |
+
"| fr-aktuell.de | 62 (2.43%) |\n",
|
479 |
+
"| berlinonline.de | 59 (2.31%) |\n",
|
480 |
+
"| szon.de | 57 (2.23%) |\n",
|
481 |
+
"| feedsportal.com | 52 (2.04%) |\n",
|
482 |
+
"| n-tv.de | 47 (1.84%) |\n",
|
483 |
+
"| sat1.de | 42 (1.65%) |\n",
|
484 |
+
"| nzz.ch | 39 (1.53%) |\n",
|
485 |
+
"| yahoo.com | 38 (1.49%) |\n",
|
486 |
+
"| ngz-online.de | 37 (1.45%) |\n",
|
487 |
+
"| faz.net | 37 (1.45%) |\n",
|
488 |
+
"| morgenweb.de | 36 (1.41%) |\n",
|
489 |
+
"| taz.de | 28 (1.1%) |\n",
|
490 |
+
"| finanznachrichten.de | 25 (0.98%) |\n",
|
491 |
+
"| tagesschau.de | 24 (0.94%) |\n",
|
492 |
+
"| gea.de | 24 (0.94%) |\n",
|
493 |
+
"| bernerzeitung.ch | 23 (0.9%) |\n",
|
494 |
+
"| ftd.de | 22 (0.86%) |\n",
|
495 |
+
"| orf.at | 21 (0.82%) |\n",
|
496 |
+
"| rp-online.de | 21 (0.82%) |\n",
|
497 |
+
"\n"
|
498 |
+
]
|
499 |
+
}
|
500 |
+
],
|
501 |
+
"source": [
|
502 |
+
"for dataset_split in new_dataset_splits.keys():\n",
|
503 |
+
" print(f\"New GermEval 2014 Dataset Stats for {dataset_split} split:\")\n",
|
504 |
+
" print_stats(dataset_split, new_dataset_splits[dataset_split])\n",
|
505 |
+
" print(\"\")"
|
506 |
+
]
|
507 |
+
}
|
508 |
+
],
|
509 |
+
"metadata": {
|
510 |
+
"kernelspec": {
|
511 |
+
"display_name": "Python 3 (ipykernel)",
|
512 |
+
"language": "python",
|
513 |
+
"name": "python3"
|
514 |
+
},
|
515 |
+
"language_info": {
|
516 |
+
"codemirror_mode": {
|
517 |
+
"name": "ipython",
|
518 |
+
"version": 3
|
519 |
+
},
|
520 |
+
"file_extension": ".py",
|
521 |
+
"mimetype": "text/x-python",
|
522 |
+
"name": "python",
|
523 |
+
"nbconvert_exporter": "python",
|
524 |
+
"pygments_lexer": "ipython3",
|
525 |
+
"version": "3.12.3"
|
526 |
+
}
|
527 |
+
},
|
528 |
+
"nbformat": 4,
|
529 |
+
"nbformat_minor": 5
|
530 |
+
}
|