haystack/tutorials/Tutorial1_Basic_QA_Pipeline.ipynb
2020-01-24 18:24:07 +01:00

294 lines
11 KiB
Plaintext

{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Task: Question Answering for Game of Thrones\n",
"\n",
"<img style=\"float: right;\" src=\"https://upload.wikimedia.org/wikipedia/en/d/d8/Game_of_Thrones_title_card.jpg\">\n",
"\n",
"Question Answering can be used in a variety of use cases. A very common one: Using it to navigate through complex knowledge bases or long documents (\"search setting\").\n",
"\n",
"A \"knowledge base\" could for example be your website, an internal wiki or a collection of financial reports. \n",
"In this tutorial we will work on a slightly different domain: \"Game of Thrones\". \n",
"\n",
"Let's see how we can use a bunch of wikipedia articles to answer a variety of questions about the \n",
"marvellous seven kingdoms... \n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# Let's start by adjust the working directory so that it is the root of the repository\n",
"# This should be run just once.\n",
"import os\n",
"os.chdir('../')\n",
"print(\"Current working directory is {}\".format(os.getcwd()))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"pycharm": {
"is_executing": false
}
},
"outputs": [],
"source": [
"from haystack import Finder\n",
"from haystack.database.sql import SQLDocumentStore\n",
"from haystack.indexing.cleaning import clean_wiki_text\n",
"from haystack.indexing.io import write_documents_to_db, fetch_archive_from_http\n",
"from haystack.reader.farm import FARMReader\n",
"from haystack.reader.transformers import TransformersReader\n",
"from haystack.retriever.tfidf import TfidfRetriever\n",
"from haystack.utils import print_answers"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Indexing & cleaning documents"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"pycharm": {
"is_executing": false
}
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"11/28/2019 12:02:51 - INFO - haystack.indexing.io - Fetching from https://s3.eu-central-1.amazonaws.com/deepset.ai-farm-qa/datasets/documents/wiki_gameofthrones_txt.zip to `data/article_txt_got`\n",
"100%|██████████| 1167348/1167348 [00:00<00:00, 8157729.91B/s]\n",
"11/28/2019 12:02:52 - INFO - haystack.indexing.io - Wrote 517 docs to DB\n"
]
}
],
"source": [
"# Let's first get some documents that we want to query\n",
"# Here: 517 Wikipedia articles for Game of Thrones\n",
"doc_dir = \"data/article_txt_got\"\n",
"s3_url = \"https://s3.eu-central-1.amazonaws.com/deepset.ai-farm-qa/datasets/documents/wiki_gameofthrones_txt.zip\"\n",
"fetch_archive_from_http(url=s3_url, output_dir=doc_dir)\n",
"\n",
"# The documents can be stored in different types of \"DocumentStores\".\n",
"# For dev we suggest a light-weight SQL DB\n",
"# For production we suggest elasticsearch\n",
"document_store = SQLDocumentStore(url=\"sqlite:///qa.db\")\n",
"\n",
"# Now, let's write the docs to our DB.\n",
"# You can optionally supply a cleaning function that is applied to each doc (e.g. to remove footers)\n",
"# It must take a str as input, and return a str.\n",
"write_documents_to_db(document_store=document_store, document_dir=doc_dir, clean_func=clean_wiki_text, only_empty_db=True)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Initalize Reader, Retriever & Finder"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"pycharm": {
"is_executing": false,
"name": "#%%\n"
}
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"11/28/2019 12:02:56 - INFO - haystack.retriever.tfidf - Found 2813 candidate paragraphs from 519 docs in DB\n"
]
}
],
"source": [
"# A retriever identifies the k most promising chunks of text that might contain the answer for our question\n",
"# Retrievers use some simple but fast algorithm, here: TF-IDF\n",
"retriever = TfidfRetriever(document_store=document_store)"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"pycharm": {
"is_executing": false
}
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"11/28/2019 12:02:57 - INFO - haystack.indexing.io - Fetching from https://s3.eu-central-1.amazonaws.com/deepset.ai-farm-models/0.3.0/bert-english-qa-large.tar.gz to `model`\n",
"100%|██████████| 1245547135/1245547135 [00:53<00:00, 23191957.12B/s]\n",
"11/28/2019 12:04:05 - WARNING - farm.data_handler.processor - Loading tokenizer from deprecated FARM config. If you used `custom_vocab` or `never_split_chars`, this won't work anymore.\n"
]
}
],
"source": [
"# A reader scans the text chunks in detail and extracts the k best answers\n",
"# Reader use more powerful but slower deep learning models\n",
"# You can select a local model or any of the QA models published on huggingface's model hub (https://huggingface.co/models)\n",
"# here: a medium sized BERT QA model trained via FARM on Squad 2.0\n",
"reader = FARMReader(model_name_or_path=\"deepset/bert-base-cased-squad2\", use_gpu=False)\n",
"\n",
"# OR: use alternatively a reader from huggingface's transformers package (https://github.com/huggingface/transformers)\n",
"# reader = TransformersReader(model=\"distilbert-base-uncased-distilled-squad\", tokenizer=\"distilbert-base-uncased\", use_gpu=-1)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"pycharm": {
"is_executing": false
}
},
"outputs": [],
"source": [
"# The Finder sticks together retriever and retriever in a pipeline to answer our actual questions\n",
"finder = Finder(reader, retriever)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Voilá! Ask a question!"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"pycharm": {
"is_executing": false
}
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"11/28/2019 12:04:26 - INFO - haystack - Identified 10 candidates via retriever:\n",
" paragraph_id document_id text\n",
" 2723 506 \\n===Arya Stark===\\n'''Arya Stark''' portrayed by Maisie Williams. Arya Star...\n",
" 2212 407 \\n====Season 8====\\nArya reunites with Jon, Gendry, and the Hound, who have ...\n",
" 2205 407 \\n====Season 1====\\nArya accompanies her father Ned and her sister Sansa to ...\n",
" 548 105 \\n===''A Game of Thrones''===\\nSansa Stark begins the novel by being betroth...\n",
" 1437 258 \\n===In Braavos===\\nLady Crane returns to her chambers to find a wounded Ary...\n",
" 462 92 \\n== Characters ==\\nThe tale is told through the eyes of 9 recurring POV cha...\n",
" 2198 407 \\n==== ''A Game of Thrones'' ====\\nArya adopts a direwolf cub, which she nam...\n",
" 2211 407 \\n====Season 7====\\nTaking the face of Walder Frey, Arya gathers the men of ...\n",
" 570 106 \\n=== Arya Stark ===\\nArya Stark is the third child and younger daughter of ...\n",
" 313 65 \\n===On the Kingsroad===\\nCity Watchmen search the caravan for Gendry but ar...\n",
"11/28/2019 12:04:27 - INFO - haystack - Applying the reader now to look for the answer in detail ...\n",
"Inferencing: 100%|██████████| 1/1 [00:21<00:00, 21.82s/it]\n"
]
}
],
"source": [
"# You can configure how many candidates the reader and retriever shall return\n",
"# The higher top_k_retriever, the better (but also the slower) your answers. \n",
"prediction = finder.get_answers(question=\"Who is the father of Arya Stark?\", top_k_retriever=10, top_k_reader=5)"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
"#prediction = finder.get_answers(question=\"Who created the Dothraki vocabulary?\", top_k_reader=5)\n",
"#prediction = finder.get_answers(question=\"Who is the sister of Sansa?\", top_k_reader=5)"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {
"pycharm": {
"is_executing": false,
"name": "#%%\n"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[ { 'answer': 'Eddard',\n",
" 'context': 'ry warrior queen. She travels with her father, Eddard, to '\n",
" \"King's Landing when he is made Hand of the\"},\n",
" { 'answer': 'Ned',\n",
" 'context': '\\n'\n",
" '====Season 1====\\n'\n",
" 'Arya accompanies her father Ned and her sister Sansa to '\n",
" \"King's Landing. Before the\"},\n",
" { 'answer': 'Lord Eddard',\n",
" 'context': ' is the younger daughter and third child of Lord Eddard '\n",
" 'and Catelyn Stark of Winterfell. Ever the to'},\n",
" { 'answer': 'Lord Eddard Stark',\n",
" 'context': ' Tourney of the Hand to honour her father Lord Eddard '\n",
" 'Stark, Sansa Stark is enchanted by the knights'},\n",
" { 'answer': 'Eddard and Catelyn Stark',\n",
" 'context': 'e third child and younger daughter of Eddard and Catelyn '\n",
" 'Stark. She serves as a POV character for 33'}]\n"
]
}
],
"source": [
"print_answers(prediction, details=\"minimal\")"
]
}
],
"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": "3.7.4"
},
"pycharm": {
"stem_cell": {
"cell_type": "raw",
"source": [],
"metadata": {
"collapsed": false
}
}
}
},
"nbformat": 4,
"nbformat_minor": 2
}