Spaces:
Sleeping
Sleeping
File size: 7,842 Bytes
579ab0b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 |
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Notebook for processing the text data (chunking, cleaning, embeddings)"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"import json\n",
"from typing import Iterable\n",
"from langchain.docstore.document import Document\n",
"from typing import List\n",
"\n",
"# Helper methods for storing and loading already generated documents\n",
"def store_documents(documents, file_path: str) -> None:\n",
" with open(file_path, \"w\") as jsonl_file:\n",
" for doc in documents:\n",
" jsonl_file.write(doc.json() + \"\\n\")\n",
"\n",
"\n",
"def load_documents(file_path: str) -> List[Document]:\n",
" documents = []\n",
" with open(file_path, \"r\") as jsonl_file:\n",
" for line in jsonl_file:\n",
" data = json.loads(line)\n",
" obj = Document(**data)\n",
" documents.append(obj)\n",
" return documents"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"def get_pdf_documents(all_docs: bool):\n",
" \"\"\"\n",
" Method for returning the documents of the PDFs. Processing and updating takes place in update_pdf_documents.\n",
" all_docs parameter defines whether to load all documents or only new ones. Only new ones can be used if the index is already build and new documents should be added.\n",
" \"\"\"\n",
" pdf_documents = []\n",
" if all_docs:\n",
" pdf_documents = load_documents(\"./../input_data/PDF/documents/all_documents\")\n",
" else:\n",
" pdf_documents = load_documents(\"./../input_data/PDF/documents/new_documents\")\n",
"\n",
" return pdf_documents\n",
"\n",
"def get_web_documents(all_docs: bool) -> List[Document]:\n",
" \"\"\"\n",
" Method for returning the already processed documents. FIRST need to call get_web_docs_for_cleaning and clean manually. As it is a manual cleaning process, the methods are need to be called asynchronously.\n",
" \"\"\"\n",
" web_documents = []\n",
" if all_docs:\n",
" web_documents = load_documents(\"./../input_data/Web/documents/all_documents\")\n",
" else:\n",
" web_documents = load_documents(\"./../input_data/Web/documents/new_documents\")\n",
"\n",
" return web_documents\n",
"\n",
"def get_template_documents(all_docs: bool) -> List[Document]:\n",
" \"\"\"\n",
" Method for returning the documents of the templates.\n",
" \"\"\"\n",
" template_documents = []\n",
" if all_docs:\n",
" template_documents = load_documents(\"./../input_data/Templates/documents/all_documents\")\n",
" else:\n",
" template_documents = load_documents(\"./../input_data/Templates/documents/new_documents\")\n",
"\n",
" return template_documents\n",
"\n",
"def get_dataset_documents() -> List[Document]:\n",
" \"\"\"\n",
" Method for returning the documents of the templates.\n",
" \"\"\"\n",
" template_documents = []\n",
" template_documents = load_documents(\"./../input_data/QA_dataset/all_documents\")\n",
"\n",
" return template_documents"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"def get_documents_from_files(all_docs: bool):\n",
" \"\"\"\n",
" Gets documents from all document types.\n",
" \"\"\"\n",
" documents_all = []\n",
" documents_PDF = get_pdf_documents(all_docs)\n",
" document_web = get_web_documents(all_docs)\n",
" document_template = get_template_documents(all_docs)\n",
" document_dataset = get_dataset_documents()\n",
" \n",
" documents_all.extend(documents_PDF)\n",
" documents_all.extend(document_web)\n",
" documents_all.extend(document_template)\n",
" documents_all.extend(document_dataset)\n",
" \n",
" print(\"Number of documents: \" + str(len(documents_all)) + \"\\n\")\n",
" return documents_all"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"from langchain.text_splitter import RecursiveCharacterTextSplitter\n",
"\n",
"def split_docs(documents: List[Document], chunk_size: int, chunk_overlap: int):\n",
"\n",
" text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap, separators=[\" \"])\n",
" chunkedDocuments = text_splitter.split_documents(documents)\n",
" return chunkedDocuments"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"import re\n",
"\n",
"def clean_text(text: str) -> str:\n",
" # Replace multiple whitespaces (except newlines) with a single space\n",
" text = re.sub(r\"(?!\\n)\\s+\", \" \", text)\n",
" # Replace multiple newlines with a single newline\n",
" text = re.sub(r\"\\n+\", \"\\n\", text)\n",
" # Remove leading and trailing whitespace\n",
" text = text.strip()\n",
" return text\n",
"\n",
"def clean_and_process_chunked_documents(chunkedDocuments: List[Document]) -> List[Document]:\n",
" counter = 1\n",
" for i in chunkedDocuments:\n",
" i.page_content = clean_text(i.page_content)\n",
" i.metadata[\"original_text\"] = i.page_content\n",
" i.metadata[\"doc_ID\"] = counter\n",
" counter += 1\n",
"\n",
" i.page_content = i.page_content.lower() \n",
"\n",
" return chunkedDocuments"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"from langchain.embeddings import HuggingFaceEmbeddings\n",
"\n",
"def get_embedding_model():\n",
" path = \"Basti8499/bge-large-en-v1.5-ISO-27001\"\n",
" model = HuggingFaceEmbeddings(model_name=path)\n",
" return model"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
"def create_embedding_vectors(embedding_model, documents: List[Document]):\n",
" texts = []\n",
" for document in documents:\n",
" texts.append(document.page_content)\n",
"\n",
" embeddings = embedding_model.embed_documents(texts)\n",
"\n",
" return embeddings"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"def preprocess_data(chunk_size: int, chunk_overlap: int, all_docs: bool):\n",
" documents = get_documents_from_files(all_docs)\n",
" chunked_documents = split_docs(documents, chunk_size=chunk_size, chunk_overlap=chunk_overlap)\n",
" chunked_cleaned_documents = clean_and_process_chunked_documents(chunked_documents)\n",
" embedding_model = get_embedding_model()\n",
" embeddings = create_embedding_vectors(embedding_model, chunked_cleaned_documents)\n",
"\n",
" return chunked_cleaned_documents, embedding_model, embeddings"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "venv",
"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": "3.11.5"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
|