aitech-eks-pub/cw/11_NER_RNN.ipynb
2021-09-27 12:34:44 +02:00

849 lines
20 KiB
Plaintext

{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"![Logo 1](https://git.wmi.amu.edu.pl/AITech/Szablon/raw/branch/master/Logotyp_AITech1.jpg)\n",
"<div class=\"alert alert-block alert-info\">\n",
"<h1> Ekstrakcja informacji </h1>\n",
"<h2> 11. <i>NER RNN</i> [ćwiczenia]</h2> \n",
"<h3> Jakub Pokrywka (2021)</h3>\n",
"</div>\n",
"\n",
"![Logo 2](https://git.wmi.amu.edu.pl/AITech/Szablon/raw/branch/master/Logotyp_AITech2.jpg)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Podejście softmax z embeddingami na przykładzie NER"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"/media/kuba/ssdsam/anaconda3/lib/python3.8/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 numpy as np\n",
"import gensim\n",
"import torch\n",
"import pandas as pd\n",
"import seaborn as sns\n",
"from sklearn.model_selection import train_test_split\n",
"\n",
"from datasets import load_dataset\n",
"from torchtext.vocab import Vocab\n",
"from collections import Counter\n",
"\n",
"from sklearn.datasets import fetch_20newsgroups\n",
"# https://scikit-learn.org/0.19/datasets/twenty_newsgroups.html\n",
"\n",
"from sklearn.feature_extraction.text import TfidfVectorizer\n",
"from sklearn.metrics import accuracy_score\n",
"\n",
"from tqdm.notebook import tqdm\n",
"\n",
"import torch"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"scrolled": false
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Reusing dataset conll2003 (/home/kuba/.cache/huggingface/datasets/conll2003/conll2003/1.0.0/40e7cb6bcc374f7c349c83acd1e9352a4f09474eb691f64f364ee62eb65d0ca6)\n"
]
}
],
"source": [
"dataset = load_dataset(\"conll2003\")"
]
},
{
"cell_type": "code",
"execution_count": 3,
"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": 4,
"metadata": {},
"outputs": [],
"source": [
"vocab = build_vocab(dataset['train']['tokens'])"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"23627"
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"len(vocab.itos)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"15"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"vocab['on']"
]
},
{
"cell_type": "code",
"execution_count": 7,
"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": 8,
"metadata": {},
"outputs": [],
"source": [
"def labels_process(dt):\n",
" return [ torch.tensor([0] + document + [0], dtype = torch.long) for document in dt]\n"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [],
"source": [
"train_tokens_ids = data_process(dataset['train']['tokens'])"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [],
"source": [
"test_tokens_ids = data_process(dataset['test']['tokens'])"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [],
"source": [
"validation_tokens_ids = data_process(dataset['validation']['tokens'])"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {
"scrolled": true
},
"outputs": [],
"source": [
"train_labels = labels_process(dataset['train']['ner_tags'])"
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {},
"outputs": [],
"source": [
"validation_labels = labels_process(dataset['validation']['ner_tags'])"
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {},
"outputs": [],
"source": [
"test_labels = labels_process(dataset['test']['ner_tags'])"
]
},
{
"cell_type": "code",
"execution_count": 15,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"tensor([ 2, 966, 22409, 238, 773, 9, 4588, 212, 7686, 4,\n",
" 3])"
]
},
"execution_count": 15,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"train_tokens_ids[0]"
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"{'chunk_tags': [11, 21, 11, 12, 21, 22, 11, 12, 0],\n",
" 'id': '0',\n",
" 'ner_tags': [3, 0, 7, 0, 0, 0, 7, 0, 0],\n",
" 'pos_tags': [22, 42, 16, 21, 35, 37, 16, 21, 7],\n",
" 'tokens': ['EU',\n",
" 'rejects',\n",
" 'German',\n",
" 'call',\n",
" 'to',\n",
" 'boycott',\n",
" 'British',\n",
" 'lamb',\n",
" '.']}"
]
},
"execution_count": 16,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"dataset['train'][0]"
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {
"scrolled": true
},
"outputs": [
{
"data": {
"text/plain": [
"tensor([0, 3, 0, 7, 0, 0, 0, 7, 0, 0, 0])"
]
},
"execution_count": 17,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"train_labels[0]"
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {},
"outputs": [],
"source": [
"def get_scores(y_true, y_pred):\n",
" acc_score = 0\n",
" tp = 0\n",
" fp = 0\n",
" selected_items = 0\n",
" relevant_items = 0 \n",
"\n",
" for p,t in zip(y_pred, y_true):\n",
" if p == t:\n",
" acc_score +=1\n",
"\n",
" if p > 0 and p == t:\n",
" tp +=1\n",
"\n",
" if p > 0:\n",
" selected_items += 1\n",
"\n",
" if t > 0 :\n",
" relevant_items +=1\n",
"\n",
" \n",
" \n",
" if selected_items == 0:\n",
" precision = 1.0\n",
" else:\n",
" precision = tp / selected_items\n",
" \n",
" \n",
" if relevant_items == 0:\n",
" recall = 1.0\n",
" else:\n",
" recall = tp / relevant_items\n",
" \n",
" \n",
" if precision + recall == 0.0 :\n",
" f1 = 0.0\n",
" else:\n",
" f1 = 2* precision * recall / (precision + recall)\n",
"\n",
" return precision, recall, f1"
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {},
"outputs": [],
"source": [
"num_tags = max([max(x) for x in dataset['train']['ner_tags'] ]) + 1 "
]
},
{
"cell_type": "code",
"execution_count": 20,
"metadata": {},
"outputs": [],
"source": [
"class LSTM(torch.nn.Module):\n",
"\n",
" def __init__(self):\n",
" super(LSTM, self).__init__()\n",
" self.emb = torch.nn.Embedding(len(vocab.itos),100)\n",
" self.rec = torch.nn.LSTM(100, 256, 1, batch_first = True)\n",
" self.fc1 = torch.nn.Linear( 256 , 9)\n",
"\n",
" def forward(self, x):\n",
" emb = torch.relu(self.emb(x))\n",
" \n",
" lstm_output, (h_n, c_n) = self.rec(emb)\n",
" \n",
" out_weights = self.fc1(lstm_output)\n",
"\n",
" return out_weights"
]
},
{
"cell_type": "code",
"execution_count": 21,
"metadata": {},
"outputs": [],
"source": [
"lstm = LSTM()"
]
},
{
"cell_type": "code",
"execution_count": 22,
"metadata": {},
"outputs": [],
"source": [
"criterion = torch.nn.CrossEntropyLoss()"
]
},
{
"cell_type": "code",
"execution_count": 23,
"metadata": {},
"outputs": [],
"source": [
"optimizer = torch.optim.Adam(lstm.parameters())"
]
},
{
"cell_type": "code",
"execution_count": 24,
"metadata": {},
"outputs": [],
"source": [
"def eval_model(dataset_tokens, dataset_labels, model):\n",
" Y_true = []\n",
" Y_pred = []\n",
" for i in tqdm(range(len(dataset_labels))):\n",
" batch_tokens = dataset_tokens[i].unsqueeze(0)\n",
" tags = list(dataset_labels[i].numpy())\n",
" Y_true += tags\n",
" \n",
" Y_batch_pred_weights = model(batch_tokens).squeeze(0)\n",
" Y_batch_pred = torch.argmax(Y_batch_pred_weights,1)\n",
" Y_pred += list(Y_batch_pred.numpy())\n",
" \n",
"\n",
" return get_scores(Y_true, Y_pred)\n",
" "
]
},
{
"cell_type": "code",
"execution_count": 25,
"metadata": {},
"outputs": [],
"source": [
"NUM_EPOCHS = 5"
]
},
{
"cell_type": "code",
"execution_count": 26,
"metadata": {
"scrolled": true
},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "75184b632ce54ae690b3444778f44651",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=14041.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "74a55a414fa948a3b251b89f780564d0",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=3250.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"(0.5068524970963996, 0.5072649075903755, 0.5070586184860281)\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "0c9c580076fb4ec48b7ea2f300878594",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=14041.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "0be8681c67f64aca95ce5d3c44f10538",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=3250.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"(0.653649243957614, 0.6381494827385795, 0.6458063757205035)\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "2dec403004bb4ae298bc73553ea3f4bc",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=14041.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "eebed0407ba343e29cf8c2d607f631dc",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=3250.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"(0.7140486069946651, 0.7001046146693014, 0.7070078647728607)\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "70792f22eea343c8916bcfcf9215c298",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=14041.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "5d400bf1b656433ba2091cf750ec2d78",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=3250.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"(0.756327964151629, 0.725909566430315, 0.7408066429418744)\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "604c4fa13c03435d81bf68be37977d74",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=14041.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "2f78871f366f4fd1b7de6c4be5303906",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=3250.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"(0.7963248522230789, 0.7203301174009067, 0.7564235581324383)\n"
]
}
],
"source": [
"for i in range(NUM_EPOCHS):\n",
" lstm.train()\n",
" #for i in tqdm(range(500)):\n",
" for i in tqdm(range(len(train_labels))):\n",
" batch_tokens = train_tokens_ids[i].unsqueeze(0)\n",
" tags = train_labels[i].unsqueeze(1)\n",
" \n",
" \n",
" predicted_tags = lstm(batch_tokens)\n",
"\n",
" \n",
" optimizer.zero_grad()\n",
" loss = criterion(predicted_tags.squeeze(0),tags.squeeze(1))\n",
" \n",
" loss.backward()\n",
" optimizer.step()\n",
" \n",
" lstm.eval()\n",
" print(eval_model(validation_tokens_ids, validation_labels, lstm))"
]
},
{
"cell_type": "code",
"execution_count": 27,
"metadata": {
"scrolled": true
},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "5159f7a61c3a439bab45573f15ea55b2",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=3250.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n"
]
},
{
"data": {
"text/plain": [
"(0.7963248522230789, 0.7203301174009067, 0.7564235581324383)"
]
},
"execution_count": 27,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"eval_model(validation_tokens_ids, validation_labels, lstm)"
]
},
{
"cell_type": "code",
"execution_count": 28,
"metadata": {},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "4b604bbb796f4d4cb99528fad98cfdff",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"HBox(children=(FloatProgress(value=0.0, max=3453.0), HTML(value='')))"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n"
]
},
{
"data": {
"text/plain": [
"(0.7450810185185185, 0.6348619329388561, 0.685569755058573)"
]
},
"execution_count": 28,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"eval_model(test_tokens_ids, test_labels, lstm)"
]
},
{
"cell_type": "code",
"execution_count": 29,
"metadata": {
"scrolled": true
},
"outputs": [
{
"data": {
"text/plain": [
"14041"
]
},
"execution_count": 29,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"len(train_tokens_ids)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## pytania\n",
"\n",
"- co zrobić z trenowaniem na batchach > 1 ?\n",
"- co zrobić, żeby sieć uwzględniała następne tokeny, a nie tylko poprzednie?\n",
"- w jaki sposób wykorzystać taką sieć do zadania zwykłej klasyfikacji?"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Zadanie na zajęcia ( 20 minut)\n",
"\n",
"zmodyfikować sieć tak, żeby była używała dwuwarstwowej, dwukierunkowej warstwy GRU oraz dropoutu. Dropout ma nałożony na embeddingi.\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Zadanie domowe\n",
"\n",
"\n",
"- sklonować repozytorium https://git.wmi.amu.edu.pl/kubapok/en-ner-conll-2003\n",
"- stworzyć model seq labelling bazujący na sieci neuronowej opisanej w punkcie niżej (można bazować na tym jupyterze lub nie).\n",
"- model sieci to GRU (o dowolnych parametrach) + CRF w pytorchu korzystając z modułu CRF z poprzednich zajęć- - stworzyć predykcje w plikach dev-0/out.tsv oraz test-A/out.tsv\n",
"- wynik fscore sprawdzony za pomocą narzędzia geval (patrz poprzednie zadanie) powinien wynosić conajmniej 0.65\n",
"- proszę umieścić predykcję oraz skrypty generujące (w postaci tekstowej a nie jupyter) w repo, a w MS TEAMS umieścić link do swojego repo\n",
"termin 22.06, 60 punktów, za najlepszy wynik- 100 punktów\n",
" "
]
}
],
"metadata": {
"author": "Jakub Pokrywka",
"email": "kubapok@wmi.amu.edu.pl",
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"lang": "pl",
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.3"
},
"subtitle": "11.NER RNN[ćwiczenia]",
"title": "Ekstrakcja informacji",
"year": "2021"
},
"nbformat": 4,
"nbformat_minor": 4
}