Created
June 20, 2019 11:01
-
-
Save nikogamulin/173bc65dca79ff0a6cefe5e856c4e5d4 to your computer and use it in GitHub Desktop.
flair-test.ipynb
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| { | |
| "cells": [ | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:46:01.166626Z", | |
| "end_time": "2019-06-20T10:46:04.936750Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "from flair.data import Sentence\nfrom flair.models import SequenceTagger\nfrom numpy import dot\nfrom numpy.linalg import norm\n\n# make a sentence\nsentence = Sentence('I love Berlin .')\n\n# load the NER tagger\ntagger = SequenceTagger.load('ner')\n\n# run NER over sentence\ntagger.predict(sentence)", | |
| "execution_count": 30, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:46:01,169 loading file /Users/nikogamulin/.flair/models/en-ner-conll03-v0.4.pt\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "execute_result", | |
| "execution_count": 30, | |
| "data": { | |
| "text/plain": "[Sentence: \"I love Berlin .\" - 4 Tokens]" | |
| }, | |
| "metadata": {} | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "The Sentence now has entity annotations. Print the sentence to see what the tagger found." | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:51:46.524368Z", | |
| "end_time": "2019-06-20T09:51:46.529038Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "print(sentence)\nprint('The following NER tags are found:')\n\n# iterate over entities and print\nfor entity in sentence.get_spans('ner'):\n print(entity)", | |
| "execution_count": 4, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "Sentence: \"I love Berlin .\" - 4 Tokens\nThe following NER tags are found:\nLOC-span [3]: \"Berlin\"\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "# Tutorial 1: NLP Base Types" | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:52:46.936673Z", | |
| "end_time": "2019-06-20T09:52:46.941986Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "# The sentence objects holds a sentence that we may want to embed or tag\nfrom flair.data import Sentence\n\n# Make a sentence object by passing a whitespace tokenized string\nsentence = Sentence('The grass is green .')\n\n# Print the object to see what's in there\nprint(sentence)", | |
| "execution_count": 5, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "Sentence: \"The grass is green .\" - 5 Tokens\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:53:11.320247Z", | |
| "end_time": "2019-06-20T09:53:11.325115Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "# using the token id\nprint(sentence.get_token(4))\n# using the index itself\nprint(sentence[3])", | |
| "execution_count": 6, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "Token: 4 green\nToken: 4 green\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "## Tokenization" | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:54:10.798334Z", | |
| "end_time": "2019-06-20T09:54:10.803539Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "# Make a sentence object by passing an untokenized string and the 'use_tokenizer' flag\nsentence = Sentence('The grass is green.', use_tokenizer=True)\n\n# Print the object to see what's in there\nprint(sentence)", | |
| "execution_count": 7, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "Sentence: \"The grass is green .\" - 5 Tokens\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "## Adding Tags to Tokens" | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:54:56.470612Z", | |
| "end_time": "2019-06-20T09:54:56.474827Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "# add a tag to a word in the sentence\nsentence[3].add_tag('ner', 'color')\n\n# print the sentence with all tags of this type\nprint(sentence.to_tagged_string())", | |
| "execution_count": 8, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "The grass is green <color> .\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:55:33.444253Z", | |
| "end_time": "2019-06-20T09:55:33.448702Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "from flair.data import Label\n\ntag: Label = sentence[3].get_tag('ner')\n\nprint(f'\"{sentence[3]}\" is tagged as \"{tag.value}\" with confidence score \"{tag.score}\"')", | |
| "execution_count": 9, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "\"Token: 4 green\" is tagged as \"color\" with confidence score \"1.0\"\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "# Tutorial 2: Tagging your Text" | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:57:15.561146Z", | |
| "end_time": "2019-06-20T09:57:15.564011Z" | |
| } | |
| }, | |
| "cell_type": "markdown", | |
| "source": "## Tagging with Pre-Trained Sequence Tagging Models" | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:57:29.257212Z", | |
| "end_time": "2019-06-20T09:57:31.188681Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "from flair.models import SequenceTagger\n\ntagger = SequenceTagger.load('ner')", | |
| "execution_count": 11, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 11:57:29,259 loading file /Users/nikogamulin/.flair/models/en-ner-conll03-v0.4.pt\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:57:49.695126Z", | |
| "end_time": "2019-06-20T09:57:49.919593Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "sentence = Sentence('George Washington went to Washington .')\n\n# predict NER tags\ntagger.predict(sentence)\n\n# print sentence with predicted tags\nprint(sentence.to_tagged_string())", | |
| "execution_count": 12, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "George <B-PER> Washington <E-PER> went to Washington <S-LOC> .\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:58:19.095615Z", | |
| "end_time": "2019-06-20T09:58:19.099759Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "for entity in sentence.get_spans('ner'):\n print(entity)", | |
| "execution_count": 13, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "PER-span [1,2]: \"George Washington\"\nLOC-span [5]: \"Washington\"\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:59:24.533186Z", | |
| "end_time": "2019-06-20T09:59:24.537410Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "sentence = Sentence(\"Burger King is not a best choice for a healthy food.\", use_tokenizer=True)", | |
| "execution_count": 14, | |
| "outputs": [] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T09:59:48.188423Z", | |
| "end_time": "2019-06-20T09:59:48.527928Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "# predict NER tags\ntagger.predict(sentence)\nfor entity in sentence.get_spans('ner'):\n print(entity)", | |
| "execution_count": 15, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "ORG-span [1,2]: \"Burger King\"\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:00:29.771210Z", | |
| "end_time": "2019-06-20T10:00:29.775402Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "print(sentence.to_dict(tag_type='ner'))", | |
| "execution_count": 16, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "{'text': 'Burger King is not a best choice for a healthy food.', 'labels': [], 'entities': [{'text': 'Burger King', 'start_pos': 0, 'end_pos': 11, 'type': 'ORG', 'confidence': 0.6567533910274506}]}\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "## Embedding" | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "### Word Embedding" | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:04:30.966438Z", | |
| "end_time": "2019-06-20T10:06:44.197381Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "from flair.embeddings import WordEmbeddings\n\n# init embedding\nglove_embedding = WordEmbeddings('glove')", | |
| "execution_count": 17, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:04:31,118 https://s3.eu-central-1.amazonaws.com/alan-nlp/resources/embeddings/glove.gensim.vectors.npy not found in cache, downloading to /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmpxi_2v3d1\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "100%|██████████| 160000128/160000128 [01:54<00:00, 1402777.43B/s]", | |
| "name": "stderr" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:06:25,499 copying /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmpxi_2v3d1 to cache at /Users/nikogamulin/.flair/embeddings/glove.gensim.vectors.npy\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "\n", | |
| "name": "stderr" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:06:25,970 removing temp file /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmpxi_2v3d1\n2019-06-20 12:06:26,517 https://s3.eu-central-1.amazonaws.com/alan-nlp/resources/embeddings/glove.gensim not found in cache, downloading to /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmplxnhk39b\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "100%|██████████| 21494764/21494764 [00:16<00:00, 1318189.15B/s]", | |
| "name": "stderr" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:06:43,180 copying /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmplxnhk39b to cache at /Users/nikogamulin/.flair/embeddings/glove.gensim\n2019-06-20 12:06:43,231 removing temp file /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmplxnhk39b\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "\n/Users/nikogamulin/workspace/pytorch-lab/venv/lib/python3.7/site-packages/smart_open/smart_open_lib.py:398: UserWarning: This function is deprecated, use smart_open.open instead. See the migration notes for details: https://github.com/RaRe-Technologies/smart_open/blob/master/README.rst#migrating-to-the-new-open-function\n 'See the migration notes for details: %s' % _MIGRATION_NOTES_URL\n", | |
| "name": "stderr" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "### BERT, ELMo, and Flair Embeddings" | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:08:04.262075Z", | |
| "end_time": "2019-06-20T10:08:57.968889Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "from flair.embeddings import FlairEmbeddings\n\n# init embedding\nflair_embedding_forward = FlairEmbeddings('news-forward')\n\n# create a sentence\nsentence = Sentence('The grass is green .')\n\n# embed words in sentence\nflair_embedding_forward.embed(sentence)", | |
| "execution_count": 18, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:08:04,476 https://s3.eu-central-1.amazonaws.com/alan-nlp/resources/embeddings-v0.4.1/big-news-forward--h2048-l1-d0.05-lr30-0.25-20/news-forward-0.4.1.pt not found in cache, downloading to /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmpj1qzymrj\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "100%|██████████| 73034624/73034624 [00:52<00:00, 1387700.78B/s]", | |
| "name": "stderr" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:08:57,501 copying /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmpj1qzymrj to cache at /Users/nikogamulin/.flair/embeddings/news-forward-0.4.1.pt\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "\n", | |
| "name": "stderr" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:08:57,674 removing temp file /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmpj1qzymrj\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "execute_result", | |
| "execution_count": 18, | |
| "data": { | |
| "text/plain": "[Sentence: \"The grass is green .\" - 5 Tokens]" | |
| }, | |
| "metadata": {} | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "#### Recommended Flair Usage" | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:09:55.593948Z", | |
| "end_time": "2019-06-20T10:10:50.242773Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "from flair.embeddings import WordEmbeddings, FlairEmbeddings, StackedEmbeddings\n\n# create a StackedEmbedding object that combines glove and forward/backward flair embeddings\nstacked_embeddings = StackedEmbeddings([\n WordEmbeddings('glove'), \n FlairEmbeddings('news-forward'), \n FlairEmbeddings('news-backward'),\n ])", | |
| "execution_count": 19, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "/Users/nikogamulin/workspace/pytorch-lab/venv/lib/python3.7/site-packages/smart_open/smart_open_lib.py:398: UserWarning: This function is deprecated, use smart_open.open instead. See the migration notes for details: https://github.com/RaRe-Technologies/smart_open/blob/master/README.rst#migrating-to-the-new-open-function\n 'See the migration notes for details: %s' % _MIGRATION_NOTES_URL\n", | |
| "name": "stderr" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:09:57,028 https://s3.eu-central-1.amazonaws.com/alan-nlp/resources/embeddings-v0.4.1/big-news-backward--h2048-l1-d0.05-lr30-0.25-20/news-backward-0.4.1.pt not found in cache, downloading to /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmp5ao07d0m\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "100%|██████████| 73034575/73034575 [00:52<00:00, 1392467.21B/s]", | |
| "name": "stderr" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:10:49,739 copying /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmp5ao07d0m to cache at /Users/nikogamulin/.flair/embeddings/news-backward-0.4.1.pt\n", | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "\n", | |
| "name": "stderr" | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": "2019-06-20 12:10:49,996 removing temp file /var/folders/nn/wsc7gmvs5cz7pf2mdj7d2pwh0000gn/T/tmp5ao07d0m\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:11:02.200773Z", | |
| "end_time": "2019-06-20T10:11:02.324187Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "sentence = Sentence('The grass is green .')\n\n# just embed a sentence using the StackedEmbedding as you would with any single embedding.\nstacked_embeddings.embed(sentence)\n\n# now check out the embedded tokens.\nfor token in sentence:\n print(token)\n print(token.embedding)", | |
| "execution_count": 20, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "Token: 1 The\ntensor([-3.8194e-02, -2.4487e-01, 7.2812e-01, ..., -4.4014e-04,\n -3.9301e-02, 1.0601e-02])\nToken: 2 grass\ntensor([-8.1353e-01, 9.4042e-01, -2.4048e-01, ..., -3.7749e-04,\n -2.3563e-02, 1.1700e-02])\nToken: 3 is\ntensor([-0.5426, 0.4148, 1.0322, ..., -0.0061, 0.0112, 0.0100])\nToken: 4 green\ntensor([-0.6791, 0.3491, -0.2398, ..., -0.0026, -0.0118, 0.0455])\nToken: 5 .\ntensor([-3.3979e-01, 2.0941e-01, 4.6348e-01, ..., -2.3405e-04,\n 3.8688e-03, 5.7725e-03])\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": {}, | |
| "cell_type": "markdown", | |
| "source": "## Document Embeddings" | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:13:01.042761Z", | |
| "end_time": "2019-06-20T10:13:01.045592Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "from flair.embeddings import WordEmbeddings, FlairEmbeddings, DocumentPoolEmbeddings, Sentence", | |
| "execution_count": 21, | |
| "outputs": [] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:13:12.709443Z", | |
| "end_time": "2019-06-20T10:13:14.846507Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "# initialize the word embeddings\nglove_embedding = WordEmbeddings('glove')\nflair_embedding_forward = FlairEmbeddings('news-forward')\nflair_embedding_backward = FlairEmbeddings('news-backward')\n\n# initialize the document embeddings, mode = mean\ndocument_embeddings = DocumentPoolEmbeddings([glove_embedding,\n flair_embedding_backward,\n flair_embedding_forward])", | |
| "execution_count": 22, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "/Users/nikogamulin/workspace/pytorch-lab/venv/lib/python3.7/site-packages/smart_open/smart_open_lib.py:398: UserWarning: This function is deprecated, use smart_open.open instead. See the migration notes for details: https://github.com/RaRe-Technologies/smart_open/blob/master/README.rst#migrating-to-the-new-open-function\n 'See the migration notes for details: %s' % _MIGRATION_NOTES_URL\n", | |
| "name": "stderr" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:13:42.452459Z", | |
| "end_time": "2019-06-20T10:13:42.673425Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "# create an example sentence\nsentence = Sentence('The grass is green . And the sky is blue .')\n\n# embed the sentence with our document embedding\ndocument_embeddings.embed(sentence)\n\n# now check out the embedded sentence.\nprint(sentence.get_embedding())", | |
| "execution_count": 23, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "tensor([-0.3197, 0.2621, 0.4037, ..., -0.0013, -0.0026, 0.0170],\n grad_fn=<CatBackward>)\n", | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:23:08.051327Z", | |
| "end_time": "2019-06-20T10:23:08.054910Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "def cosine_similarity(a, b):\n cos_sim = dot(a, b)/(norm(a)*norm(b))\n return cos_sim", | |
| "execution_count": 24, | |
| "outputs": [] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:25:00.584929Z", | |
| "end_time": "2019-06-20T10:25:00.593291Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "sentence.get_embedding().data.numpy()", | |
| "execution_count": 26, | |
| "outputs": [ | |
| { | |
| "output_type": "execute_result", | |
| "execution_count": 26, | |
| "data": { | |
| "text/plain": "array([-0.31969544, 0.26205996, 0.4037069 , ..., -0.00134025,\n -0.00258876, 0.01702889], dtype=float32)" | |
| }, | |
| "metadata": {} | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:25:28.303332Z", | |
| "end_time": "2019-06-20T10:25:28.306833Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "coupons = ['Buy one pizza, get other as a gift. Limited time! Available on all crusts',\n 'Get a free pair of socks for every friend you refer',\n '$1 mozarella sticks with purcase of an extra value meal',\n 'Swiss Watch Expo: $100 off + free shipping if you get this item now. This timepieve is yours for the next 24 hours',\n 'Dell: save 20% on your next purchase',\n 'Unlock Special Bonus for Premium Dog Food',\n 'shhh... here\\'s $8 off this purchase only',\n 'Sign up and receive 10% off your first undies purchase',\n 'Would you like 10% off your first purchase in any Wallmart store?',\n 'Check out now & receive 10% off your first food order',\n 'Get your device repaired today and save 15% at Radio Shack',\n 'Spend $150 and get free US standard delivery',\n 'Belk mobile extra $10, $20 or $30 off | ends Saturday']", | |
| "execution_count": 27, | |
| "outputs": [] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:49:35.761554Z", | |
| "end_time": "2019-06-20T10:49:39.271230Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "sentences = [Sentence(coupon, use_tokenizer=True) for coupon in coupons]\nfor sentence in sentences:\n document_embeddings.embed(sentence)", | |
| "execution_count": 35, | |
| "outputs": [] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:49:51.513311Z", | |
| "end_time": "2019-06-20T10:49:51.526272Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "max_similarities = {i: None for i in range(len(sentences))}\nfor i, sentence_i in enumerate(sentences):\n vec_i = sentence_i.get_embedding().data.numpy()\n for j, sentence_j in enumerate(sentences):\n vec_j = sentence_j.get_embedding().data.numpy()\n if i != j:\n similarity = cosine_similarity(vec_i, vec_j)\n if max_similarities[i] is None:\n max_similarities[i] = (j, similarity)\n if max_similarities[i][1] < similarity:\n max_similarities[i] = (j, similarity)", | |
| "execution_count": 37, | |
| "outputs": [] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:49:55.049609Z", | |
| "end_time": "2019-06-20T10:49:55.054361Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "max_similarities", | |
| "execution_count": 38, | |
| "outputs": [ | |
| { | |
| "output_type": "execute_result", | |
| "execution_count": 38, | |
| "data": { | |
| "text/plain": "{0: (3, 0.9133937),\n 1: (0, 0.8866179),\n 2: (0, 0.82934916),\n 3: (0, 0.9133937),\n 4: (7, 0.8908964),\n 5: (0, 0.7471674),\n 6: (3, 0.89610726),\n 7: (9, 0.9255427),\n 8: (3, 0.89906967),\n 9: (7, 0.9255427),\n 10: (3, 0.8647915),\n 11: (3, 0.8663579),\n 12: (11, 0.84938776)}" | |
| }, | |
| "metadata": {} | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:56:14.493637Z", | |
| "end_time": "2019-06-20T10:56:15.769765Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "from flair.embeddings import WordEmbeddings, DocumentRNNEmbeddings\n\nglove_embedding = WordEmbeddings('glove')\n\ndocument_embeddings = DocumentRNNEmbeddings([glove_embedding])", | |
| "execution_count": 39, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": "/Users/nikogamulin/workspace/pytorch-lab/venv/lib/python3.7/site-packages/smart_open/smart_open_lib.py:398: UserWarning: This function is deprecated, use smart_open.open instead. See the migration notes for details: https://github.com/RaRe-Technologies/smart_open/blob/master/README.rst#migrating-to-the-new-open-function\n 'See the migration notes for details: %s' % _MIGRATION_NOTES_URL\n", | |
| "name": "stderr" | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:57:09.290210Z", | |
| "end_time": "2019-06-20T10:57:09.341190Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "sentences = [Sentence(coupon, use_tokenizer=True) for coupon in coupons]\nfor sentence in sentences:\n document_embeddings.embed(sentence)", | |
| "execution_count": 41, | |
| "outputs": [] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:57:24.553601Z", | |
| "end_time": "2019-06-20T10:57:24.565270Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "max_similarities = {i: None for i in range(len(sentences))}\nfor i, sentence_i in enumerate(sentences):\n vec_i = sentence_i.get_embedding().data.numpy()\n for j, sentence_j in enumerate(sentences):\n vec_j = sentence_j.get_embedding().data.numpy()\n if i != j:\n similarity = cosine_similarity(vec_i, vec_j)\n if max_similarities[i] is None:\n max_similarities[i] = (j, similarity)\n if max_similarities[i][1] < similarity:\n max_similarities[i] = (j, similarity)", | |
| "execution_count": 42, | |
| "outputs": [] | |
| }, | |
| { | |
| "metadata": { | |
| "ExecuteTime": { | |
| "start_time": "2019-06-20T10:57:33.242734Z", | |
| "end_time": "2019-06-20T10:57:33.247518Z" | |
| }, | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "max_similarities", | |
| "execution_count": 43, | |
| "outputs": [ | |
| { | |
| "output_type": "execute_result", | |
| "execution_count": 43, | |
| "data": { | |
| "text/plain": "{0: (1, 0.28898498),\n 1: (9, 0.3556913),\n 2: (8, 0.39637336),\n 3: (6, 0.40637004),\n 4: (9, 0.5623293),\n 5: (9, 0.39802852),\n 6: (3, 0.40637004),\n 7: (4, 0.37906048),\n 8: (2, 0.39637336),\n 9: (4, 0.5623293),\n 10: (9, 0.349115),\n 11: (9, 0.54480976),\n 12: (3, 0.3571025)}" | |
| }, | |
| "metadata": {} | |
| } | |
| ] | |
| }, | |
| { | |
| "metadata": { | |
| "trusted": true | |
| }, | |
| "cell_type": "code", | |
| "source": "", | |
| "execution_count": null, | |
| "outputs": [] | |
| } | |
| ], | |
| "metadata": { | |
| "kernelspec": { | |
| "name": "python3", | |
| "display_name": "Python 3", | |
| "language": "python" | |
| }, | |
| "language_info": { | |
| "name": "python", | |
| "version": "3.7.3", | |
| "mimetype": "text/x-python", | |
| "codemirror_mode": { | |
| "name": "ipython", | |
| "version": 3 | |
| }, | |
| "pygments_lexer": "ipython3", | |
| "nbconvert_exporter": "python", | |
| "file_extension": ".py" | |
| }, | |
| "gist": { | |
| "id": "", | |
| "data": { | |
| "description": "flair-test.ipynb", | |
| "public": true | |
| } | |
| } | |
| }, | |
| "nbformat": 4, | |
| "nbformat_minor": 2 | |
| } |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment