s434695/en-ner-conll-2003
1
0
Fork 0
forked from kubapok/en-ner-conll-2003
Code Issues Pull Requests Projects Releases Wiki Activity

solution

Browse Source
This commit is contained in:
Tomasz Grzybowski 2021-06-09 03:01:30 +02:00
parent d6b3d1c0d1
commit c6aaaf6544
7 changed files with 1253 additions and 436009 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

436
.ipynb_checkpoints/Program-checkpoint.ipynb
View File

@ -1,6 +1,438 @@
{
"cells": [],
"metadata": {},
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"id": "e574fca4",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"C:\\Users\\grzyb\\anaconda3\\lib\\site-packages\\gensim\\similarities\\__init__.py:15: UserWarning: The gensim.similarities.levenshtein submodule is disabled, because the optional Levenshtein package <https://pypi.org/project/python-Levenshtein/> is unavailable. Install Levenhstein (e.g. `pip install python-Levenshtein`) to suppress this warning.\n",
" warnings.warn(msg)\n"
]
}
],
"source": [
"import pandas as pd\n",
"import numpy as np\n",
"import csv\n",
"import os.path\n",
"import shutil\n",
"import torch\n",
"from tqdm import tqdm\n",
"from itertools import islice\n",
"from sklearn.model_selection import train_test_split\n",
"from torchtext.vocab import Vocab\n",
"from collections import Counter\n",
"from nltk.tokenize import word_tokenize\n",
"import gensim.downloader as api\n",
"from gensim.models.word2vec import Word2Vec"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "b476f295",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Collecting gensim\n",
" Downloading gensim-4.0.1-cp38-cp38-win_amd64.whl (23.9 MB)\n",
"Requirement already satisfied: scipy>=0.18.1 in c:\\users\\grzyb\\anaconda3\\lib\\site-packages (from gensim) (1.6.2)\n",
"Collecting Cython==0.29.21\n",
" Downloading Cython-0.29.21-cp38-cp38-win_amd64.whl (1.7 MB)\n",
"Requirement already satisfied: numpy>=1.11.3 in c:\\users\\grzyb\\anaconda3\\lib\\site-packages (from gensim) (1.20.1)\n",
"Collecting smart-open>=1.8.1\n",
" Downloading smart_open-5.1.0-py3-none-any.whl (57 kB)\n",
"Installing collected packages: smart-open, Cython, gensim\n",
" Attempting uninstall: Cython\n",
" Found existing installation: Cython 0.29.23\n",
" Uninstalling Cython-0.29.23:\n",
" Successfully uninstalled Cython-0.29.23\n",
"Successfully installed Cython-0.29.21 gensim-4.0.1 smart-open-5.1.0\n"
]
}
],
"source": [
"!pip install gensim"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "fbe3a657",
"metadata": {},
"outputs": [],
"source": [
"class NERModel(torch.nn.Module):\n",
"\n",
" def __init__(self,):\n",
" super(NERModel, self).__init__()\n",
" self.emb = torch.nn.Embedding(23628,200)\n",
" self.fc1 = torch.nn.Linear(600,9)\n",
" \n",
"\n",
" def forward(self, x):\n",
" x = self.emb(x)\n",
" x = x.reshape(600) \n",
" x = self.fc1(x)\n",
" return x"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "3497a580",
"metadata": {},
"outputs": [],
"source": [
"def process_output(lines):\n",
" result = []\n",
" for line in lines:\n",
" last_label = None\n",
" new_line = []\n",
" for label in line:\n",
" if(label != \"O\" and label[0:2] == \"I-\"):\n",
" if last_label == None or last_label == \"O\":\n",
" label = label.replace('I-', 'B-')\n",
" else:\n",
" label = \"I-\" + last_label[2:]\n",
" last_label = label\n",
" new_line.append(label)\n",
" x = (\" \".join(new_line))\n",
" result.append(\" \".join(new_line))\n",
" return result"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "3e78d902",
"metadata": {},
"outputs": [],
"source": [
"def build_vocab(dataset):\n",
" counter = Counter()\n",
" for document in dataset:\n",
" counter.update(document)\n",
" return Vocab(counter, specials=['<unk>', '<pad>', '<bos>', '<eos>'])"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "ec8537cf",
"metadata": {},
"outputs": [],
"source": [
"def data_process(dt):\n",
" return [ torch.tensor([vocab['<bos>']] +[vocab[token] for token in document ] + [vocab['<eos>']], dtype = torch.long) for document in dt]"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "847c958a",
"metadata": {},
"outputs": [],
"source": [
"def labels_process(dt):\n",
" return [ torch.tensor([0] + document + [0], dtype = torch.long) for document in dt]"
]
},
{
"cell_type": "code",
"execution_count": 24,
"id": "66bee163",
"metadata": {},
"outputs": [],
"source": [
"def predict(input_tokens, labels):\n",
"\n",
" results = []\n",
" \n",
" for i in range(len(input_tokens)):\n",
" line_results = []\n",
" for j in range(1, len(input_tokens[i]) - 1):\n",
" x = input_tokens[i][j-1: j+2].to(device_gpu)\n",
" predicted = ner_model(x.long())\n",
" result = torch.argmax(predicted)\n",
" label = labels[result]\n",
" line_results.append(label)\n",
" results.append(line_results)\n",
"\n",
" return results"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "39046f3f",
"metadata": {},
"outputs": [],
"source": [
"train = pd.read_csv('train/train.tsv.xz', sep='\\t', names=['a', 'b'])"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "9b40a8b6",
"metadata": {},
"outputs": [],
"source": [
"labels = ['O','B-LOC', 'I-LOC','B-MISC', 'I-MISC', 'B-ORG', 'I-ORG', 'B-PER', 'I-PER'] \n",
"train[\"a\"]=train[\"a\"].apply(lambda x: [labels.index(y) for y in x.split()])\n",
"train[\"b\"]=train[\"b\"].apply(lambda x: x.split())"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "02a12cbd",
"metadata": {},
"outputs": [],
"source": [
"vocab = build_vocab(train['b'])"
]
},
{
"cell_type": "code",
"execution_count": 10,
"id": "8cc6d19d",
"metadata": {},
"outputs": [],
"source": [
" tensors = []\n",
"\n",
" for sent in train[\"b\"]:\n",
" sent_tensor = torch.tensor(())\n",
" for word in sent:\n",
" temp = torch.tensor([word[0].isupper(), word[0].isdigit()])\n",
" sent_tensor = torch.cat((sent_tensor, temp))\n",
"\n",
" tensors.append(sent_tensor)"
]
},
{
"cell_type": "code",
"execution_count": 15,
"id": "690085f6",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"'NVIDIA GeForce RTX 2060'"
]
},
"execution_count": 15,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"torch.cuda.get_device_name(0)"
]
},
{
"cell_type": "code",
"execution_count": 16,
"id": "64b2d751",
"metadata": {},
"outputs": [],
"source": [
"device_gpu = torch.device(\"cuda:0\")\n",
"ner_model = NERModel().to(device_gpu)\n",
"criterion = torch.nn.CrossEntropyLoss()\n",
"optimizer = torch.optim.Adam(ner_model.parameters())"
]
},
{
"cell_type": "code",
"execution_count": 17,
"id": "094d7e69",
"metadata": {},
"outputs": [],
"source": [
"train_labels = labels_process(train['a'])\n",
"train_tokens_ids = data_process(train['b'])\n"
]
},
{
"cell_type": "code",
"execution_count": 18,
"id": "17291b41",
"metadata": {},
"outputs": [],
"source": [
"train_tensors = [torch.cat((token, tensors[i])) for i, token in enumerate(train_tokens_ids)]"
]
},
{
"cell_type": "code",
"execution_count": 19,
"id": "045b7186",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch: 0\n",
"f1: 0.6373470953763748\n",
"acc: 0.9116419913061858\n",
"epoch: 1\n",
"f1: 0.7973076923076923\n",
"acc: 0.9540771782783307\n",
"epoch: 2\n",
"f1: 0.8640167364016735\n",
"acc: 0.9702287410511612\n",
"epoch: 3\n",
"f1: 0.9038441719055962\n",
"acc: 0.9793820591289644\n",
"epoch: 4\n",
"f1: 0.928903400400047\n",
"acc: 0.9850890978100043\n"
]
}
],
"source": [
"for epoch in range(5):\n",
" acc_score = 0\n",
" prec_score = 0\n",
" selected_items = 0\n",
" recall_score = 0\n",
" relevant_items = 0\n",
" items_total = 0\n",
" ner_model.train()\n",
" for i in range(len(train_labels)):\n",
" for j in range(1, len(train_labels[i]) - 1):\n",
" X = train_tensors[i][j - 1: j + 2].to(device_gpu)\n",
"\n",
" Y = train_labels[i][j: j + 1].to(device_gpu)\n",
"\n",
" Y_predictions = ner_model(X.long())\n",
"\n",
" acc_score += int(torch.argmax(Y_predictions) == Y)\n",
" if torch.argmax(Y_predictions) != 0:\n",
" selected_items += 1\n",
" if torch.argmax(Y_predictions) != 0 and torch.argmax(Y_predictions) == Y.item():\n",
" prec_score += 1\n",
" if Y.item() != 0:\n",
" relevant_items += 1\n",
" if Y.item() != 0 and torch.argmax(Y_predictions) == Y.item():\n",
" recall_score += 1\n",
"\n",
" items_total += 1\n",
" optimizer.zero_grad()\n",
" loss = criterion(Y_predictions.unsqueeze(0), Y)\n",
" loss.backward()\n",
" optimizer.step()\n",
"\n",
" precision = prec_score / selected_items\n",
" recall = recall_score / relevant_items\n",
" f1_score = (2 * precision * recall) / (precision + recall)\n",
" print(f'epoch: {epoch}')\n",
" print(f'f1: {f1_score}')\n",
" print(f'acc: {acc_score / items_total}')"
]
},
{
"cell_type": "code",
"execution_count": 28,
"id": "f75aa5e2",
"metadata": {},
"outputs": [],
"source": [
"def create_tensors_list(data):\n",
" tensors = []\n",
"\n",
" for sent in data[\"a\"]:\n",
" sent_tensor = torch.tensor(())\n",
" for word in sent:\n",
" temp = torch.tensor([word[0].isupper(), word[0].isdigit()])\n",
" sent_tensor = torch.cat((sent_tensor, temp))\n",
"\n",
" tensors.append(sent_tensor)\n",
"\n",
" return tensors"
]
},
{
"cell_type": "code",
"execution_count": 29,
"id": "49215802",
"metadata": {},
"outputs": [],
"source": [
"dev = pd.read_csv('dev-0/in.tsv', sep='\\t', names=['a'])\n",
"dev[\"a\"] = dev[\"a\"].apply(lambda x: x.split())\n",
"\n",
"dev_tokens_ids = data_process(dev[\"a\"])\n",
"\n",
"dev_extra_tensors = create_tensors_list(dev)\n",
"\n",
"dev_tensors = [torch.cat((token, dev_extra_tensors[i])) for i, token in enumerate(dev_tokens_ids)]\n",
"\n",
"results = predict(dev_tensors, labels)\n",
"results_processed = process_output(results)\n",
"\n",
"with open(\"dev-0/out.tsv\", \"w\") as f:\n",
" for line in results_processed:\n",
" f.write(line + \"\\n\")"
]
},
{
"cell_type": "code",
"execution_count": 30,
"id": "8c5b007e",
"metadata": {},
"outputs": [],
"source": [
"test = pd.read_csv('test-A/in.tsv', sep='\\t', names=['a'])\n",
"test[\"a\"] = test[\"a\"].apply(lambda x: x.split())\n",
"\n",
"test_tokens_ids = data_process(test[\"a\"])\n",
"\n",
"test_extra_tensors = create_tensors_list(test)\n",
"\n",
"test_tensors = [torch.cat((token, test_extra_tensors[i])) for i, token in enumerate(test_tokens_ids)]\n",
"\n",
"results = predict(test_tensors, labels)\n",
"results_processed = process_output(results)\n",
"\n",
"with open(\"test-A/out.tsv\", \"w\") as f:\n",
" for line in results_processed:\n",
" f.write(line + \"\\n\")"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.8"
}
},
"nbformat": 4,
"nbformat_minor": 5
}

202
Program.ipynb
View File

@ -5,7 +5,16 @@
"execution_count": 1,
"id": "e574fca4",
"metadata": {},
"outputs": [],
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"C:\\Users\\grzyb\\anaconda3\\lib\\site-packages\\gensim\\similarities\\__init__.py:15: UserWarning: The gensim.similarities.levenshtein submodule is disabled, because the optional Levenshtein package <https://pypi.org/project/python-Levenshtein/> is unavailable. Install Levenhstein (e.g. `pip install python-Levenshtein`) to suppress this warning.\n",
" warnings.warn(msg)\n"
]
}
],
"source": [
"import pandas as pd\n",
"import numpy as np\n",
@ -23,6 +32,37 @@
"from gensim.models.word2vec import Word2Vec"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "b476f295",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Collecting gensim\n",
" Downloading gensim-4.0.1-cp38-cp38-win_amd64.whl (23.9 MB)\n",
"Requirement already satisfied: scipy>=0.18.1 in c:\\users\\grzyb\\anaconda3\\lib\\site-packages (from gensim) (1.6.2)\n",
"Collecting Cython==0.29.21\n",
" Downloading Cython-0.29.21-cp38-cp38-win_amd64.whl (1.7 MB)\n",
"Requirement already satisfied: numpy>=1.11.3 in c:\\users\\grzyb\\anaconda3\\lib\\site-packages (from gensim) (1.20.1)\n",
"Collecting smart-open>=1.8.1\n",
" Downloading smart_open-5.1.0-py3-none-any.whl (57 kB)\n",
"Installing collected packages: smart-open, Cython, gensim\n",
" Attempting uninstall: Cython\n",
" Found existing installation: Cython 0.29.23\n",
" Uninstalling Cython-0.29.23:\n",
" Successfully uninstalled Cython-0.29.23\n",
"Successfully installed Cython-0.29.21 gensim-4.0.1 smart-open-5.1.0\n"
]
}
],
"source": [
"!pip install gensim"
]
},
{
"cell_type": "code",
"execution_count": 2,
@ -106,6 +146,30 @@
" return [ torch.tensor([0] + document + [0], dtype = torch.long) for document in dt]"
]
},
{
"cell_type": "code",
"execution_count": 24,
"id": "66bee163",
"metadata": {},
"outputs": [],
"source": [
"def predict(input_tokens, labels):\n",
"\n",
" results = []\n",
" \n",
" for i in range(len(input_tokens)):\n",
" line_results = []\n",
" for j in range(1, len(input_tokens[i]) - 1):\n",
" x = input_tokens[i][j-1: j+2].to(device_gpu)\n",
" predicted = ner_model(x.long())\n",
" result = torch.argmax(predicted)\n",
" label = labels[result]\n",
" line_results.append(label)\n",
" results.append(line_results)\n",
"\n",
" return results"
]
},
{
"cell_type": "code",
"execution_count": 7,
@ -113,9 +177,7 @@
"metadata": {},
"outputs": [],
"source": [
"train = pd.read_csv('train/train.tsv.xz', sep='\\t', names=['a', 'b'])\n",
"dev = pd.read_csv('dev-0/in.tsv', sep='\\t', names=['a'])\n",
"test = pd.read_csv('test-A/in.tsv', sep='\\t', names=['a'])"
"train = pd.read_csv('train/train.tsv.xz', sep='\\t', names=['a', 'b'])"
]
},
{
@ -142,7 +204,7 @@
},
{
"cell_type": "code",
"execution_count": 18,
"execution_count": 10,
"id": "8cc6d19d",
"metadata": {},
"outputs": [],
@ -160,20 +222,41 @@
},
{
"cell_type": "code",
"execution_count": 10,
"execution_count": 15,
"id": "690085f6",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"'NVIDIA GeForce RTX 2060'"
]
},
"execution_count": 15,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"torch.cuda.get_device_name(0)"
]
},
{
"cell_type": "code",
"execution_count": 16,
"id": "64b2d751",
"metadata": {},
"outputs": [],
"source": [
"device_cpu = torch.device(\"cpu\")\n",
"ner_model = NERModel().to(device_cpu)\n",
"device_gpu = torch.device(\"cuda:0\")\n",
"ner_model = NERModel().to(device_gpu)\n",
"criterion = torch.nn.CrossEntropyLoss()\n",
"optimizer = torch.optim.Adam(ner_model.parameters())"
]
},
{
"cell_type": "code",
"execution_count": 19,
"execution_count": 17,
"id": "094d7e69",
"metadata": {},
"outputs": [],
@ -184,7 +267,7 @@
},
{
"cell_type": "code",
"execution_count": 22,
"execution_count": 18,
"id": "17291b41",
"metadata": {},
"outputs": [],
@ -194,10 +277,32 @@
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 19,
"id": "045b7186",
"metadata": {},
"outputs": [],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch: 0\n",
"f1: 0.6373470953763748\n",
"acc: 0.9116419913061858\n",
"epoch: 1\n",
"f1: 0.7973076923076923\n",
"acc: 0.9540771782783307\n",
"epoch: 2\n",
"f1: 0.8640167364016735\n",
"acc: 0.9702287410511612\n",
"epoch: 3\n",
"f1: 0.9038441719055962\n",
"acc: 0.9793820591289644\n",
"epoch: 4\n",
"f1: 0.928903400400047\n",
"acc: 0.9850890978100043\n"
]
}
],
"source": [
"for epoch in range(5):\n",
" acc_score = 0\n",
@ -209,9 +314,9 @@
" ner_model.train()\n",
" for i in range(len(train_labels)):\n",
" for j in range(1, len(train_labels[i]) - 1):\n",
" X = train_tensors[i][j - 1: j + 2].to(device_cpu)\n",
" X = train_tensors[i][j - 1: j + 2].to(device_gpu)\n",
"\n",
" Y = train_labels[i][j: j + 1].to(device_cpu)\n",
" Y = train_labels[i][j: j + 1].to(device_gpu)\n",
"\n",
" Y_predictions = ner_model(X.long())\n",
"\n",
@ -238,6 +343,75 @@
" print(f'f1: {f1_score}')\n",
" print(f'acc: {acc_score / items_total}')"
]
},
{
"cell_type": "code",
"execution_count": 28,
"id": "f75aa5e2",
"metadata": {},
"outputs": [],
"source": [
"def create_tensors_list(data):\n",
" tensors = []\n",
"\n",
" for sent in data[\"a\"]:\n",
" sent_tensor = torch.tensor(())\n",
" for word in sent:\n",
" temp = torch.tensor([word[0].isupper(), word[0].isdigit()])\n",
" sent_tensor = torch.cat((sent_tensor, temp))\n",
"\n",
" tensors.append(sent_tensor)\n",
"\n",
" return tensors"
]
},
{
"cell_type": "code",
"execution_count": 29,
"id": "49215802",
"metadata": {},
"outputs": [],
"source": [
"dev = pd.read_csv('dev-0/in.tsv', sep='\\t', names=['a'])\n",
"dev[\"a\"] = dev[\"a\"].apply(lambda x: x.split())\n",
"\n",
"dev_tokens_ids = data_process(dev[\"a\"])\n",
"\n",
"dev_extra_tensors = create_tensors_list(dev)\n",
"\n",
"dev_tensors = [torch.cat((token, dev_extra_tensors[i])) for i, token in enumerate(dev_tokens_ids)]\n",
"\n",
"results = predict(dev_tensors, labels)\n",
"results_processed = process_output(results)\n",
"\n",
"with open(\"dev-0/out.tsv\", \"w\") as f:\n",
" for line in results_processed:\n",
" f.write(line + \"\\n\")"
]
},
{
"cell_type": "code",
"execution_count": 30,
"id": "8c5b007e",
"metadata": {},
"outputs": [],
"source": [
"test = pd.read_csv('test-A/in.tsv', sep='\\t', names=['a'])\n",
"test[\"a\"] = test[\"a\"].apply(lambda x: x.split())\n",
"\n",
"test_tokens_ids = data_process(test[\"a\"])\n",
"\n",
"test_extra_tensors = create_tensors_list(test)\n",
"\n",
"test_tensors = [torch.cat((token, test_extra_tensors[i])) for i, token in enumerate(test_tokens_ids)]\n",
"\n",
"results = predict(test_tensors, labels)\n",
"results_processed = process_output(results)\n",
"\n",
"with open(\"test-A/out.tsv\", \"w\") as f:\n",
" for line in results_processed:\n",
" f.write(line + \"\\n\")"
]
}
],
"metadata": {

331
Untitled.ipynb
View File

@ -1,331 +0,0 @@
{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"id": "0895b7c8",
"metadata": {},
"outputs": [],
"source": [
"# Informacje na temat zakomentowanego kodu oraz wyników znajdują się w README.md \n",
"\n",
"import pandas as pd\n",
"import os.path\n",
"import shutil\n",
"import torch\n",
"import pandas as pd\n",
"from torchtext.vocab import Vocab\n",
"from collections import Counter\n",
"\n",
"# class NERModelWithAlpha(torch.nn.Module):\n",
"# def __init__(self,):\n",
"# super(NERModel, self).__init__()\n",
"# self.emb = torch.nn.Embedding(23629,200)\n",
"# self.fc1 = torch.nn.Linear(1200,9) \n",
"\n",
"# def forward(self, x):\n",
"# x = self.emb(x)\n",
"# x = x.reshape(1200) \n",
"# x = self.fc1(x)\n",
"# return x\n",
"\n",
"class NERModel(torch.nn.Module):\n",
" def __init__(self,):\n",
" super(NERModel, self).__init__()\n",
" self.emb = torch.nn.Embedding(23628,200)\n",
" self.fc1 = torch.nn.Linear(600,9) \n",
"\n",
" def forward(self, x):\n",
" x = self.emb(x)\n",
" x = x.reshape(600) \n",
" x = self.fc1(x)\n",
" return x\n",
"\n",
"def data_process(dt):\n",
" return [ torch.tensor([vocab['<bos>']] +[vocab[token] for token in document ] + [vocab['<eos>']], dtype = torch.long) for document in dt]\n",
" \n",
"# def data_process(dt):\n",
"# result = []\n",
"# for document in dt:\n",
"# sentence = [vocab['<bos>'],vocab['<alpha>']]\n",
"# for token in document:\n",
"# sentence += [vocab[token]]\n",
"# sentence += [vocab['<alpha>'] if token.isalpha() else vocab['<notalpha>']]\n",
"# sentence += [vocab['<eos>'],vocab['<alpha>']]\n",
"# result.append(torch.tensor(sentence, dtype = torch.long))\n",
"# return result\n",
"\n",
"def build_vocab(dataset):\n",
" counter = Counter()\n",
" for document in dataset:\n",
" counter.update(document)\n",
" return Vocab(counter, specials=['<unk>', '<pad>', '<bos>', '<eos>']) #, '<alpha>', '<notalpha>'])\n",
"\n",
"def labels_process(dt):\n",
" return [ torch.tensor([0] + document + [0], dtype = torch.long) for document in dt]\n",
"\n",
"def process(model, x):\n",
" predicted = model(x)\n",
" result = torch.argmax(predicted)\n",
" return labels[result]\n",
"\n",
"def process_dataset(model, path):\n",
" with open(path, 'r') as f:\n",
" lines = f.readlines()\n",
" X = [x.split() for x in lines]\n",
" data_tokens_ids = data_process(X)\n",
" results = []\n",
" for i in range(len(data_tokens_ids)):\n",
" line_results = []\n",
" for j in range(1, len(data_tokens_ids[i]) - 1):\n",
"# for j in range(2, len(data_tokens_ids[i]) - 3, 2):\n",
" #x = data_tokens_ids[i][j-2: j+4].to(device_gpu)\n",
" x = data_tokens_ids[i][j-1: j+2].to(device_cpu)\n",
" label = process(model, x)\n",
" line_results.append(label)\n",
" results.append(line_results)\n",
" return results\n",
"\n",
"# Przetwarzanie danych z wyjścia modelu (gdy B- i I- nie dotyczą tej samej etykiety)\n",
"def process_output(lines):\n",
" result = []\n",
" for line in lines:\n",
" last_label = None\n",
" new_line = []\n",
" for label in line:\n",
" if(label != \"O\" and label[0:2] == \"I-\"):\n",
" if last_label == None or last_label == \"O\":\n",
" label = label.replace('I-', 'B-')\n",
" else:\n",
" label = \"I-\" + last_label[2:]\n",
" last_label = label\n",
" new_line.append(label)\n",
" result.append(\" \".join(new_line))\n",
" return result\n"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "b2f73f9e",
"metadata": {},
"outputs": [],
"source": [
"labels = ['O','B-LOC', 'I-LOC','B-MISC', 'I-MISC', 'B-ORG', 'I-ORG', 'B-PER', 'I-PER'] "
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "2a94110d",
"metadata": {},
"outputs": [],
"source": [
"if not os.path.isfile('train/train.tsv'):\n",
" import lzma\n",
" with lzma.open('train/train.tsv.xz', 'rb') as f_in:\n",
" with open('train/train.tsv', 'wb') as f_out:\n",
" shutil.copyfileobj(f_in, f_out)"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "02b81af3",
"metadata": {},
"outputs": [],
"source": [
"data = pd.read_csv('train/train.tsv', sep='\\t', names=['iob', 'tokens'])\n",
"data[\"iob\"]=data[\"iob\"].apply(lambda x: [labels.index(y) for y in x.split()])\n",
"data[\"tokens\"]=data[\"tokens\"].apply(lambda x: x.split())"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "f005db98",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>iob</th>\n",
" <th>tokens</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>[5, 0, 3, 0, 0, 0, 3, 0, 0, 0, 7, 8, 0, 1, 0, ...</td>\n",
" <td>[EU, rejects, German, call, to, boycott, Briti...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>[0, 7, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, ...</td>\n",
" <td>[Rare, Hendrix, song, draft, sells, for, almos...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>[1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, ...</td>\n",
" <td>[China, says, Taiwan, spoils, atmosphere, for,...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>[1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, ...</td>\n",
" <td>[China, says, time, right, for, Taiwan, talks,...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>[3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, ...</td>\n",
" <td>[German, July, car, registrations, up, 14.2, p...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>...</th>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>940</th>\n",
" <td>[0, 0, 7, 0, 0, 0, 0, 0, 0, 0, 7, 8, 0, 1, 0, ...</td>\n",
" <td>[CYCLING, -, BALLANGER, KEEPS, SPRINT, TITLE, ...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>941</th>\n",
" <td>[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, ...</td>\n",
" <td>[CYCLING, -, WORLD, TRACK, CHAMPIONSHIP, RESUL...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>942</th>\n",
" <td>[0, 0, 3, 0, 7, 0, 5, 0, 0, 1, 0, 1, 0, 0, 3, ...</td>\n",
" <td>[SOCCER, -, FRENCH, DEFENDER, KOMBOUARE, JOINS...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>943</th>\n",
" <td>[0, 0, 1, 2, 3, 4, 0, 0, 0, 0, 1, 0, 1, 0, 0, ...</td>\n",
" <td>[MOTORCYCLING, -, SAN, MARINO, GRAND, PRIX, PR...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>944</th>\n",
" <td>[0, 0, 3, 4, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, ...</td>\n",
" <td>[GOLF, -, BRITISH, MASTERS, THIRD, ROUND, SCOR...</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"<p>945 rows × 2 columns</p>\n",
"</div>"
],
"text/plain": [
" iob \\\n",
"0 [5, 0, 3, 0, 0, 0, 3, 0, 0, 0, 7, 8, 0, 1, 0, ... \n",
"1 [0, 7, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, ... \n",
"2 [1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, ... \n",
"3 [1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, ... \n",
"4 [3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, ... \n",
".. ... \n",
"940 [0, 0, 7, 0, 0, 0, 0, 0, 0, 0, 7, 8, 0, 1, 0, ... \n",
"941 [0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, ... \n",
"942 [0, 0, 3, 0, 7, 0, 5, 0, 0, 1, 0, 1, 0, 0, 3, ... \n",
"943 [0, 0, 1, 2, 3, 4, 0, 0, 0, 0, 1, 0, 1, 0, 0, ... \n",
"944 [0, 0, 3, 4, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, ... \n",
"\n",
" tokens \n",
"0 [EU, rejects, German, call, to, boycott, Briti... \n",
"1 [Rare, Hendrix, song, draft, sells, for, almos... \n",
"2 [China, says, Taiwan, spoils, atmosphere, for,... \n",
"3 [China, says, time, right, for, Taiwan, talks,... \n",
"4 [German, July, car, registrations, up, 14.2, p... \n",
".. ... \n",
"940 [CYCLING, -, BALLANGER, KEEPS, SPRINT, TITLE, ... \n",
"941 [CYCLING, -, WORLD, TRACK, CHAMPIONSHIP, RESUL... \n",
"942 [SOCCER, -, FRENCH, DEFENDER, KOMBOUARE, JOINS... \n",
"943 [MOTORCYCLING, -, SAN, MARINO, GRAND, PRIX, PR... \n",
"944 [GOLF, -, BRITISH, MASTERS, THIRD, ROUND, SCOR... \n",
"\n",
"[945 rows x 2 columns]"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"data"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "4a114973",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<torchtext.vocab.Vocab at 0x7ff2dd0edac0>"
]
},
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"vocab = build_vocab(data['tokens'])\n",
"vocab"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "c666872d",
"metadata": {},
"outputs": [],
"source": [
"device_cpu = torch.device(\"cpu\")\n",
"ner_model = NERModel().to(device_cpu)\n",
"criterion = torch.nn.CrossEntropyLoss()\n",
"optimizer = torch.optim.Adam(ner_model.parameters())"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.8"
}
},
"nbformat": 4,
"nbformat_minor": 5
}

435217
Untitled1.ipynb
View File

File diff suppressed because it is too large Load Diff

430
dev-0/out.tsv
View File

File diff suppressed because one or more lines are too long

186
solution.py Normal file
View File

@ -0,0 +1,186 @@
import pandas as pd
import numpy as np
import csv
import os.path
import shutil
import torch
from tqdm import tqdm
from itertools import islice
from sklearn.model_selection import train_test_split
from torchtext.vocab import Vocab
from collections import Counter
from nltk.tokenize import word_tokenize
import gensim.downloader as api
from gensim.models.word2vec import Word2Vec
class NERModel(torch.nn.Module):
def __init__(self,):
super(NERModel, self).__init__()
self.emb = torch.nn.Embedding(23628,200)
self.fc1 = torch.nn.Linear(600,9)
def forward(self, x):
x = self.emb(x)
x = x.reshape(600)
x = self.fc1(x)
return x
def process_output(lines):
result = []
for line in lines:
last_label = None
new_line = []
for label in line:
if(label != "O" and label[0:2] == "I-"):
if last_label == None or last_label == "O":
label = label.replace('I-', 'B-')
else:
label = "I-" + last_label[2:]
last_label = label
new_line.append(label)
x = (" ".join(new_line))
result.append(" ".join(new_line))
return result
def build_vocab(dataset):
counter = Counter()
for document in dataset:
counter.update(document)
return Vocab(counter, specials=['<unk>', '<pad>', '<bos>', '<eos>'])
def data_process(dt):
return [ torch.tensor([vocab['<bos>']] +[vocab[token] for token in document ] + [vocab['<eos>']], dtype = torch.long) for document in dt]
def labels_process(dt):
return [ torch.tensor([0] + document + [0], dtype = torch.long) for document in dt]
def predict(input_tokens, labels):
results = []
for i in range(len(input_tokens)):
line_results = []
for j in range(1, len(input_tokens[i]) - 1):
x = input_tokens[i][j-1: j+2].to(device_gpu)
predicted = ner_model(x.long())
result = torch.argmax(predicted)
label = labels[result]
line_results.append(label)
results.append(line_results)
return results
train = pd.read_csv('train/train.tsv.xz', sep='\t', names=['a', 'b'])
labels = ['O','B-LOC', 'I-LOC','B-MISC', 'I-MISC', 'B-ORG', 'I-ORG', 'B-PER', 'I-PER']
train["a"]=train["a"].apply(lambda x: [labels.index(y) for y in x.split()])
train["b"]=train["b"].apply(lambda x: x.split())
vocab = build_vocab(train['b'])
tensors = []
for sent in train["b"]:
sent_tensor = torch.tensor(())
for word in sent:
temp = torch.tensor([word[0].isupper(), word[0].isdigit()])
sent_tensor = torch.cat((sent_tensor, temp))
tensors.append(sent_tensor)
device_gpu = torch.device("cuda:0")
ner_model = NERModel().to(device_gpu)
criterion = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(ner_model.parameters())
train_labels = labels_process(train['a'])
train_tokens_ids = data_process(train['b'])
train_tensors = [torch.cat((token, tensors[i])) for i, token in enumerate(train_tokens_ids)]
for epoch in range(5):
acc_score = 0
prec_score = 0
selected_items = 0
recall_score = 0
relevant_items = 0
items_total = 0
ner_model.train()
for i in range(len(train_labels)):
for j in range(1, len(train_labels[i]) - 1):
X = train_tensors[i][j - 1: j + 2].to(device_gpu)
Y = train_labels[i][j: j + 1].to(device_gpu)
Y_predictions = ner_model(X.long())
acc_score += int(torch.argmax(Y_predictions) == Y)
if torch.argmax(Y_predictions) != 0:
selected_items += 1
if torch.argmax(Y_predictions) != 0 and torch.argmax(Y_predictions) == Y.item():
prec_score += 1
if Y.item() != 0:
relevant_items += 1
if Y.item() != 0 and torch.argmax(Y_predictions) == Y.item():
recall_score += 1
items_total += 1
optimizer.zero_grad()
loss = criterion(Y_predictions.unsqueeze(0), Y)
loss.backward()
optimizer.step()
precision = prec_score / selected_items
recall = recall_score / relevant_items
f1_score = (2 * precision * recall) / (precision + recall)
print(f'epoch: {epoch}')
print(f'f1: {f1_score}')
print(f'acc: {acc_score / items_total}')
def create_tensors_list(data):
tensors = []
for sent in data["a"]:
sent_tensor = torch.tensor(())
for word in sent:
temp = torch.tensor([word[0].isupper(), word[0].isdigit()])
sent_tensor = torch.cat((sent_tensor, temp))
tensors.append(sent_tensor)
return tensors
dev = pd.read_csv('dev-0/in.tsv', sep='\t', names=['a'])
dev["a"] = dev["a"].apply(lambda x: x.split())
dev_tokens_ids = data_process(dev["a"])
dev_extra_tensors = create_tensors_list(dev)
dev_tensors = [torch.cat((token, dev_extra_tensors[i])) for i, token in enumerate(dev_tokens_ids)]
results = predict(dev_tensors, labels)
results_processed = process_output(results)
with open("dev-0/out.tsv", "w") as f:
for line in results_processed:
f.write(line + "\n")
test = pd.read_csv('test-A/in.tsv', sep='\t', names=['a'])
test["a"] = test["a"].apply(lambda x: x.split())
test_tokens_ids = data_process(test["a"])
test_extra_tensors = create_tensors_list(test)
test_tensors = [torch.cat((token, test_extra_tensors[i])) for i, token in enumerate(test_tokens_ids)]
results = predict(test_tensors, labels)
results_processed = process_output(results)
with open("test-A/out.tsv", "w") as f:
for line in results_processed:
f.write(line + "\n")

460
test-A/out.tsv
View File

File diff suppressed because one or more lines are too long