Skip to content

Commit

Permalink
added final prototype
Browse files Browse the repository at this point in the history
  • Loading branch information
@jeanroether
jeanroether committed Jun 24, 2024
1 parent 99f2293 commit 60a3e2d
Showing 1 changed file with 277 additions and 0 deletions.
277 changes: 277 additions & 0 deletions prototypes/Final-Prototyp.ipynb
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,277 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Import All Libaries"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"PyTerrier 0.10.0 has loaded Terrier 5.8 (built by craigm on 2023-11-01 18:05) and terrier-helper 0.0.8\n",
"\n",
"No etc/terrier.properties, using terrier.default.properties for bootstrap configuration.\n"
]
}
],
"source": [
"# Imports\n",
"from tira.third_party_integrations import ensure_pyterrier_is_loaded, persist_and_normalize_run\n",
"from tira.rest_api_client import Client\n",
"ensure_pyterrier_is_loaded()\n",
"import pandas as pd\n",
"import pyterrier as pt\n",
"from tqdm import tqdm\n",
"from jnius import autoclass\n",
"import gzip\n",
"import json\n",
"import re\n",
"\n",
"# Create a REST client to the TIRA platform for retrieving the pre-indexed data.\n",
"tira = Client()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Load the Dataset and the Index\n"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"# The dataset: the union of the IR Anthology and the ACL Anthology\n",
"# This line creates an IRDSDataset object and registers it under the name provided as an argument.\n",
"dataset = 'antique-test-20230107-training'\n",
"pt_dataset = pt.get_dataset(f'irds:ir-benchmarks/{dataset}')\n",
"bm25 = tira.pt.from_submission('ir-benchmarks/tira-ir-starter/BM25 Re-Rank (tira-ir-starter-pyterrier)', dataset)\n",
"\n",
"# A (pre-built) PyTerrier index loaded from TIRA\n",
"index = tira.pt.index('ir-lab-sose-2024/tira-ir-starter/Index (tira-ir-starter-pyterrier)', pt_dataset)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Stopwords"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"ir-benchmarks/antique-test-20230107-training documents: 0%| | 1912/403666 [00:02<03:27, 1939.85it/s]"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"13:47:56.872 [ForkJoinPool-1-worker-3] WARN org.terrier.structures.indexing.Indexer - Adding an empty document to the index (2824443_2) - further warnings are suppressed\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"ir-benchmarks/antique-test-20230107-training documents: 100%|██████████| 403666/403666 [00:44<00:00, 9033.35it/s] \n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"13:48:43.988 [ForkJoinPool-1-worker-3] WARN org.terrier.structures.indexing.Indexer - Indexed 1570 empty documents\n"
]
}
],
"source": [
"def create_index(documents, stopwords):\n",
" indexer = pt.IterDictIndexer(\"/tmp/index\", overwrite=True, meta={'docno': 100, 'text': 20480}, stopwords=stopwords)\n",
" index_ref = indexer.index(documents)\n",
" return pt.IndexFactory.of(index_ref)\n",
"\n",
"chatGPTStopwords =[\n",
" 'i', 'me', 'my', 'myself', 'we', 'our', 'ours', 'ourselves', 'you', 'your', 'yours', \n",
" 'yourself', 'yourselves', 'he', 'him', 'his', 'himself', 'she', 'her', 'hers', 'herself', \n",
" 'it', 'its', 'itself', 'they', 'them', 'their', 'theirs', 'themselves', 'what', 'which', \n",
" 'who', 'whom', 'this', 'that', 'these', 'those', 'am', 'is', 'are', 'was', 'were', 'be', \n",
" 'been', 'being', 'have', 'has', 'had', 'having', 'do', 'does', 'did', 'doing', 'a', 'an', \n",
" 'the', 'and', 'but', 'if', 'or', 'because', 'as', 'until', 'while', 'of', 'at', 'by', \n",
" 'for', 'with', 'about', 'against', 'between', 'into', 'through', 'during', 'before', \n",
" 'after', 'above', 'below', 'to', 'from', 'up', 'down', 'in', 'out', 'on', 'off', 'over', \n",
" 'under', 'again', 'further', 'then', 'once', 'here', 'there', 'when', 'where', 'why', \n",
" 'how', 'all', 'any', 'both', 'each', 'few', 'more', 'most', 'other', 'some', 'such', \n",
" 'no', 'nor', 'not', 'only', 'own', 'same', 'so', 'than', 'too', 'very', 's', 't', 'can', \n",
" 'will', 'just', 'don', 'should', 'now'\n",
"]\n",
"\n",
"index = create_index(pt_dataset.get_corpus_iter(), chatGPTStopwords)\n",
"\n",
"bm25_chatGPTStopwords = pt.BatchRetrieve(index, wmodel=\"BM25\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Query Expansion with Large Language Models"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"llama_sq_zs = tira.pt.transform_queries('ir-benchmarks/tu-dresden-03/qe-llama-sq-zs', dataset, prefix='llm_expansion_')\n"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"tokeniser = pt.autoclass(\"org.terrier.indexing.tokenisation.Tokeniser\").getTokeniser()\n",
"\n",
"def pt_tokenize(text):\n",
" return ' '.join(tokeniser.getTokens(text))\n",
"\n",
"def expand_query(topic):\n",
" ret = ' '.join([topic['query'], topic['query'], topic['query'], topic['query'], topic['query'], topic['llm_expansion_query']])\n",
"\n",
" # apply the tokenization\n",
" return pt_tokenize(ret)\n",
"\n",
"# we wrap this into an pyterrier transformer\n",
"# Documentation: https://pyterrier.readthedocs.io/en/latest/apply.html\n",
"pt_expand_query = pt.apply.query(expand_query)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"pipeline_llama_sq_zs = (llama_sq_zs >> pt_expand_query) >> bm25_chatGPTStopwords"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Evaluation"
]
},
{
"cell_type": "code",
"execution_count": 7,
"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>name</th>\n",
" <th>recall_1000</th>\n",
" <th>ndcg_cut_5</th>\n",
" <th>ndcg_cut.10</th>\n",
" <th>recip_rank</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>BM25_chatgptstopwords+Llama-SQ-ZS \\t</td>\n",
" <td>0.808404</td>\n",
" <td>0.566703</td>\n",
" <td>0.53322</td>\n",
" <td>0.928343</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" name recall_1000 ndcg_cut_5 ndcg_cut.10 \\\n",
"0 BM25_chatgptstopwords+Llama-SQ-ZS \\t 0.808404 0.566703 0.53322 \n",
"\n",
" recip_rank \n",
"0 0.928343 "
]
},
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"pt.Experiment(\n",
" retr_systems=[pipeline_llama_sq_zs],\n",
" topics=pt_dataset.get_topics('text'),\n",
" qrels=pt_dataset.get_qrels(),\n",
" names=['BM25_chatgptstopwords+Llama-SQ-ZS'],\n",
" eval_metrics=['recall_1000', 'ndcg_cut_5', 'ndcg_cut.10', 'recip_rank']\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": "3.10.12"
}
},
"nbformat": 4,
"nbformat_minor": 2
}

0 comments on commit 60a3e2d

Please sign in to comment.