mirror of
https://github.com/docling-project/docling.git
synced 2025-06-27 05:20:05 +00:00
add node parser, JSONPath resolution to LI example, refactor
Signed-off-by: Panos Vagenas <35837085+vagenas@users.noreply.github.com>
This commit is contained in:
Panos Vagenas
parent
d8ddb559fa
commit
4090f9700b
@ -1,5 +1,12 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# Basic RAG pipeline with LangChain"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 1,
|
||||
@ -58,14 +65,7 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Setup"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Loader and splitter"
|
||||
"## Loader and splitter"
|
||||
]
|
||||
},
|
||||
{
|
||||
@ -146,7 +146,7 @@
|
||||
{
|
||||
"data": {
|
||||
"application/vnd.jupyter.widget-view+json": {
|
||||
"model_id": "1b38d07d5fed4618a44ecf261e1e5c44",
|
||||
"model_id": "02ec5bb258e94419b725a824b8af2acc",
|
||||
"version_major": 2,
|
||||
"version_minor": 0
|
||||
},
|
||||
@ -192,7 +192,7 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Embeddings"
|
||||
"## Embeddings"
|
||||
]
|
||||
},
|
||||
{
|
||||
@ -211,7 +211,7 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Vector store"
|
||||
"## Vector store"
|
||||
]
|
||||
},
|
||||
{
|
||||
@ -240,7 +240,7 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### LLM"
|
||||
"## LLM"
|
||||
]
|
||||
},
|
||||
{
|
||||
@ -318,7 +318,7 @@
|
||||
{
|
||||
"data": {
|
||||
"text/plain": [
|
||||
"'The human annotation of DocLayNet was performed on 80863 pages.\\n\\nExplanation:\\nThe information is found in the paragraph \"DocLayNet contains 80863 PDF pages\" in the context.'"
|
||||
"'\\n80K pages were human annotated for DocLayNet.'"
|
||||
]
|
||||
},
|
||||
"execution_count": 12,
|
||||
|
||||
@ -4,7 +4,7 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# Quackling — Basic Pipeline"
|
||||
"# Basic RAG pipeline with LlamaIndex"
|
||||
]
|
||||
},
|
||||
{
|
||||
@ -22,7 +22,7 @@
|
||||
],
|
||||
"source": [
|
||||
"# requirements for this example:\n",
|
||||
"%pip install -qq docling docling-core python-dotenv llama-index-embeddings-huggingface llama-index-llms-huggingface-api llama-index-vector-stores-milvus"
|
||||
"%pip install -qq docling docling-core python-dotenv llama-index-embeddings-huggingface llama-index-llms-huggingface-api llama-index-vector-stores-milvus jsonpath-ng"
|
||||
]
|
||||
},
|
||||
{
|
||||
@ -47,8 +47,12 @@
|
||||
"\n",
|
||||
"from dotenv import load_dotenv\n",
|
||||
"from pydantic import TypeAdapter\n",
|
||||
"from rich.console import Console\n",
|
||||
"from rich.pretty import pprint\n",
|
||||
"\n",
|
||||
"console = Console()\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"load_dotenv()"
|
||||
]
|
||||
},
|
||||
@ -68,23 +72,34 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Setup"
|
||||
"## LlamaIndex extensions"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Reader and node parser"
|
||||
"Below we define our framework extensions:\n",
|
||||
"- `DoclingPDFReader` which will be used to create LlamaIndex documents, and\n",
|
||||
"- `HierarchicalJSONNodeParser`, which will be used to create LlamaIndex nodes out of the documents"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"Below we set up:\n",
|
||||
"- a `Reader` which will be used to create LlamaIndex documents, and\n",
|
||||
"- a `NodeParser`, which will be used to create LlamaIndex nodes out of the documents"
|
||||
"```mermaid\n",
|
||||
"flowchart LR\n",
|
||||
" subgraph LI extensions\n",
|
||||
" direction LR\n",
|
||||
" Reader -->|LI Documents| Parser\n",
|
||||
" end\n",
|
||||
" Reader(DoclingPDFReader)\n",
|
||||
" Parser(HierarchicalJSONNodeParser)\n",
|
||||
" Encoder(EmbedModel)\n",
|
||||
" Store(VectorStore)\n",
|
||||
" Parser -->|LI Nodes| Encoder -->|LI Nodes| Store\n",
|
||||
"```"
|
||||
]
|
||||
},
|
||||
{
|
||||
@ -110,7 +125,7 @@
|
||||
"class DoclingPDFReader(BasePydanticReader):\n",
|
||||
" class ParseType(str, Enum):\n",
|
||||
" MARKDOWN = \"markdown\"\n",
|
||||
" # JSON = \"json\"\n",
|
||||
" JSON = \"json\"\n",
|
||||
"\n",
|
||||
" parse_type: ParseType = ParseType.MARKDOWN\n",
|
||||
"\n",
|
||||
@ -122,8 +137,8 @@
|
||||
" match self.parse_type:\n",
|
||||
" case self.ParseType.MARKDOWN:\n",
|
||||
" text = dl_doc.export_to_markdown()\n",
|
||||
" # case self.ParseType.JSON:\n",
|
||||
" # text = dl_doc.model_dump_json()\n",
|
||||
" case self.ParseType.JSON:\n",
|
||||
" text = dl_doc.model_dump_json()\n",
|
||||
" case _:\n",
|
||||
" raise RuntimeError(\n",
|
||||
" f\"Unexpected parse type encountered: {self.parse_type}\"\n",
|
||||
@ -147,10 +162,140 @@
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from llama_index.core.node_parser import MarkdownNodeParser\n",
|
||||
"from typing import Any, Iterable, Sequence\n",
|
||||
"\n",
|
||||
"reader = DoclingPDFReader(parse_type=DoclingPDFReader.ParseType.MARKDOWN)\n",
|
||||
"node_parser = MarkdownNodeParser()\n",
|
||||
"from docling_core.transforms.chunker import HierarchicalChunker\n",
|
||||
"from docling_core.types import Document as DLDocument\n",
|
||||
"from llama_index.core import Document as LIDocument\n",
|
||||
"from llama_index.core.node_parser.interface import NodeParser\n",
|
||||
"from llama_index.core.schema import (\n",
|
||||
" BaseNode,\n",
|
||||
" NodeRelationship,\n",
|
||||
" RelatedNodeType,\n",
|
||||
" TextNode,\n",
|
||||
")\n",
|
||||
"from llama_index.core.utils import get_tqdm_iterable\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"class NodeMetadata(BaseModel):\n",
|
||||
" path: str\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"class HierarchicalJSONNodeParser(NodeParser):\n",
|
||||
"\n",
|
||||
" include_metadata: bool = False\n",
|
||||
"\n",
|
||||
" def _parse_nodes(\n",
|
||||
" self,\n",
|
||||
" nodes: Sequence[BaseNode],\n",
|
||||
" show_progress: bool = False,\n",
|
||||
" **kwargs: Any,\n",
|
||||
" ) -> list[BaseNode]:\n",
|
||||
" nodes_with_progress: Iterable[BaseNode] = get_tqdm_iterable(\n",
|
||||
" items=nodes, show_progress=show_progress, desc=\"Parsing nodes\"\n",
|
||||
" )\n",
|
||||
" all_nodes: list[BaseNode] = []\n",
|
||||
" chunker = HierarchicalChunker()\n",
|
||||
" for input_node in nodes_with_progress:\n",
|
||||
" li_doc = LIDocument.model_validate(input_node)\n",
|
||||
" dl_doc: DLDocument = DLDocument.model_validate_json(li_doc.get_content())\n",
|
||||
" chunk_iter = chunker.chunk(dl_doc=dl_doc)\n",
|
||||
" for chunk in chunk_iter:\n",
|
||||
" rels: dict[NodeRelationship, RelatedNodeType] = {\n",
|
||||
" NodeRelationship.SOURCE: li_doc.as_related_node_info(),\n",
|
||||
" }\n",
|
||||
" excl_metadata_keys = [\"path\"]\n",
|
||||
" node = TextNode(\n",
|
||||
" text=chunk.text,\n",
|
||||
" excluded_embed_metadata_keys=excl_metadata_keys,\n",
|
||||
" excluded_llm_metadata_keys=excl_metadata_keys,\n",
|
||||
" relationships=rels,\n",
|
||||
" )\n",
|
||||
" node.metadata = NodeMetadata(\n",
|
||||
" path=chunk.path,\n",
|
||||
" ).model_dump()\n",
|
||||
" all_nodes.append(node)\n",
|
||||
" return all_nodes"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Reader and node parser"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Using JSON"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"To leverage Docling's rich document structure format, we namely set the parse type to JSON and use a HierarchicalJSONNodeParser accordingly:"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 6,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"reader = DoclingPDFReader(parse_type=DoclingPDFReader.ParseType.JSON)\n",
|
||||
"node_parser = HierarchicalJSONNodeParser()"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Using Markdown"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"Alternatively, to just use the flat Markdown export instead of the native document format, one can uncomment and use the following:"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 7,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# from llama_index.core.node_parser import MarkdownNodeParser\n",
|
||||
"\n",
|
||||
"# reader = DoclingPDFReader(parse_type=DoclingPDFReader.ParseType.MARKDOWN)\n",
|
||||
"# node_parser = MarkdownNodeParser()\n",
|
||||
"# transformations = [node_parser]"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Transformations"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"Our transformations currently include the `node_parser`:"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 8,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"transformations = [node_parser]"
|
||||
]
|
||||
},
|
||||
@ -163,7 +308,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 6,
|
||||
"execution_count": 9,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
@ -180,12 +325,12 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Embed model"
|
||||
"## Embed model"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 7,
|
||||
"execution_count": 10,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
@ -198,12 +343,12 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Vector store"
|
||||
"## Vector store"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 8,
|
||||
"execution_count": 11,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
@ -212,7 +357,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 9,
|
||||
"execution_count": 12,
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
@ -245,13 +390,13 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 10,
|
||||
"execution_count": 13,
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"application/vnd.jupyter.widget-view+json": {
|
||||
"model_id": "536daee038de4d52a793445c6d853c72",
|
||||
"model_id": "b79e2579c1a544159099e372010f2692",
|
||||
"version_major": 2,
|
||||
"version_minor": 0
|
||||
},
|
||||
@ -262,6 +407,19 @@
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #00ff00; text-decoration-color: #00ff00\">─────────────────────────────────────────────── </span>example `Document`:<span style=\"color: #00ff00; text-decoration-color: #00ff00\"> ───────────────────────────────────────────────</span>\n",
|
||||
"</pre>\n"
|
||||
],
|
||||
"text/plain": [
|
||||
"\u001b[92m─────────────────────────────────────────────── \u001b[0mexample `Document`:\u001b[92m ───────────────────────────────────────────────\u001b[0m\n"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
@ -273,7 +431,7 @@
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">excluded_embed_metadata_keys</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008000; text-decoration-color: #008000\">'dl_doc_hash'</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">excluded_llm_metadata_keys</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008000; text-decoration-color: #008000\">'dl_doc_hash'</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">relationships</span>=<span style=\"font-weight: bold\">{}</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">text</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'## DocLayNet: A Large Human-Annotated Dataset for '</span>+<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">50593</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">text</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'{\"_name\":\"\",\"type\":\"pdf-document\",\"description\":{\"'</span>+<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">171405</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">mimetype</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'text/plain'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">start_char_idx</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">end_char_idx</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
@ -293,7 +451,7 @@
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33mexcluded_embed_metadata_keys\u001b[0m=\u001b[1m[\u001b[0m\u001b[32m'dl_doc_hash'\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33mexcluded_llm_metadata_keys\u001b[0m=\u001b[1m[\u001b[0m\u001b[32m'dl_doc_hash'\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33mrelationships\u001b[0m=\u001b[1m{\u001b[0m\u001b[1m}\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33mtext\u001b[0m=\u001b[32m'## DocLayNet: A Large Human-Annotated Dataset for '\u001b[0m+\u001b[1;36m50593\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33mtext\u001b[0m=\u001b[32m'\u001b[0m\u001b[32m{\u001b[0m\u001b[32m\"_name\":\"\",\"type\":\"pdf-document\",\"description\":\u001b[0m\u001b[32m{\u001b[0m\u001b[32m\"'\u001b[0m+\u001b[1;36m171405\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33mmimetype\u001b[0m=\u001b[32m'text/plain'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33mstart_char_idx\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33mend_char_idx\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
@ -306,24 +464,123 @@
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
},
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"num of Nodes (chunks): 76\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #00ff00; text-decoration-color: #00ff00\">────────────────────────────────────────────── </span>example Node <span style=\"font-weight: bold\">(</span>chunk<span style=\"font-weight: bold\">)</span>:<span style=\"color: #00ff00; text-decoration-color: #00ff00\"> ──────────────────────────────────────────────</span>\n",
|
||||
"</pre>\n"
|
||||
],
|
||||
"text/plain": [
|
||||
"\u001b[92m────────────────────────────────────────────── \u001b[0mexample Node \u001b[1m(\u001b[0mchunk\u001b[1m)\u001b[0m:\u001b[92m ──────────────────────────────────────────────\u001b[0m\n"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">TextNode</span><span style=\"font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">id_</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'715fc3ff-9d97-45fd-91bf-f57e3ea3ac94'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">embedding</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata</span>=<span style=\"font-weight: bold\">{</span><span style=\"color: #008000; text-decoration-color: #008000\">'path'</span>: <span style=\"color: #008000; text-decoration-color: #008000\">'$.main-text[7]'</span><span style=\"font-weight: bold\">}</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">excluded_embed_metadata_keys</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008000; text-decoration-color: #008000\">'path'</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">excluded_llm_metadata_keys</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008000; text-decoration-color: #008000\">'path'</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">relationships</span>=<span style=\"font-weight: bold\">{</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"font-weight: bold\"><</span><span style=\"color: #ff00ff; text-decoration-color: #ff00ff; font-weight: bold\">NodeRelationship.SOURCE:</span><span style=\"color: #000000; text-decoration-color: #000000\"> </span><span style=\"color: #008000; text-decoration-color: #008000\">'1'</span><span style=\"color: #000000; text-decoration-color: #000000\">>: </span><span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">RelatedNodeInfo</span><span style=\"color: #000000; text-decoration-color: #000000; font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">node_id</span><span style=\"color: #000000; text-decoration-color: #000000\">=</span><span style=\"color: #008000; text-decoration-color: #008000\">'5dfbd8c115a15fd3396b68409124cfee29fc8efac7b5c846634ff924e635e0dc'</span><span style=\"color: #000000; text-decoration-color: #000000\">,</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">node_type</span><span style=\"color: #000000; text-decoration-color: #000000\">=<ObjectType.DOCUMENT: </span><span style=\"color: #008000; text-decoration-color: #008000\">'4'</span><span style=\"color: #000000; text-decoration-color: #000000\">>,</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata</span><span style=\"color: #000000; text-decoration-color: #000000\">=</span><span style=\"color: #000000; text-decoration-color: #000000; font-weight: bold\">{</span><span style=\"color: #808000; text-decoration-color: #808000\">...</span><span style=\"color: #000000; text-decoration-color: #000000; font-weight: bold\">}</span><span style=\"color: #000000; text-decoration-color: #000000\">,</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">hash</span><span style=\"color: #000000; text-decoration-color: #000000\">=</span><span style=\"color: #008000; text-decoration-color: #008000\">'70041f7130ef2e27cc1e6156ffa78d3a1f3e0c3cb7763cc314cda6a47a029350'</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #000000; text-decoration-color: #000000; font-weight: bold\">)</span><span style=\"color: #000000; text-decoration-color: #000000\">,</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #000000; text-decoration-color: #000000\"><NodeRelationship.PREVIOUS: </span><span style=\"color: #008000; text-decoration-color: #008000\">'2'</span><span style=\"color: #000000; text-decoration-color: #000000\">>: </span><span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">RelatedNodeInfo</span><span style=\"color: #000000; text-decoration-color: #000000; font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">node_id</span><span style=\"color: #000000; text-decoration-color: #000000\">=</span><span style=\"color: #008000; text-decoration-color: #008000\">'06c2ded5-e46a-4278-92ea-eed9bc78b9f6'</span><span style=\"color: #000000; text-decoration-color: #000000\">,</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">node_type</span><span style=\"color: #000000; text-decoration-color: #000000\">=<ObjectType.TEXT: </span><span style=\"color: #008000; text-decoration-color: #008000\">'1'</span><span style=\"font-weight: bold\">></span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata</span>=<span style=\"font-weight: bold\">{</span><span style=\"color: #808000; text-decoration-color: #808000\">...</span><span style=\"font-weight: bold\">}</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">hash</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'c51973e676e01da1fc3504107653ce74d198dbafa50eedd32cf5d5e3c7a82e0b'</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"font-weight: bold\">)</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">...</span> +<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">1</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"font-weight: bold\">}</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">text</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'ABSTRACT\\nAccurate document layout analysis is a key requirement for highquality PDF document conversion. '</span>+<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">1499</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">mimetype</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'text/plain'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">start_char_idx</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">end_char_idx</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">text_template</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'{metadata_str}\\n\\n{content}'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata_template</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'{key}: {value}'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata_seperator</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'\\n'</span>\n",
|
||||
"<span style=\"font-weight: bold\">)</span>\n",
|
||||
"</pre>\n"
|
||||
],
|
||||
"text/plain": [
|
||||
"\u001b[1;35mTextNode\u001b[0m\u001b[1m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mid_\u001b[0m=\u001b[32m'715fc3ff-9d97-45fd-91bf-f57e3ea3ac94'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33membedding\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mmetadata\u001b[0m=\u001b[1m{\u001b[0m\u001b[32m'path'\u001b[0m: \u001b[32m'$.main-text\u001b[0m\u001b[32m[\u001b[0m\u001b[32m7\u001b[0m\u001b[32m]\u001b[0m\u001b[32m'\u001b[0m\u001b[1m}\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mexcluded_embed_metadata_keys\u001b[0m=\u001b[1m[\u001b[0m\u001b[32m'path'\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mexcluded_llm_metadata_keys\u001b[0m=\u001b[1m[\u001b[0m\u001b[32m'path'\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mrelationships\u001b[0m=\u001b[1m{\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1m<\u001b[0m\u001b[1;95mNodeRelationship.SOURCE:\u001b[0m\u001b[39m \u001b[0m\u001b[32m'1'\u001b[0m\u001b[39m>: \u001b[0m\u001b[1;35mRelatedNodeInfo\u001b[0m\u001b[1;39m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mnode_id\u001b[0m\u001b[39m=\u001b[0m\u001b[32m'5dfbd8c115a15fd3396b68409124cfee29fc8efac7b5c846634ff924e635e0dc'\u001b[0m\u001b[39m,\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mnode_type\u001b[0m\u001b[39m=<ObjectType.DOCUMENT: \u001b[0m\u001b[32m'4'\u001b[0m\u001b[39m>,\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mmetadata\u001b[0m\u001b[39m=\u001b[0m\u001b[1;39m{\u001b[0m\u001b[33m...\u001b[0m\u001b[1;39m}\u001b[0m\u001b[39m,\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mhash\u001b[0m\u001b[39m=\u001b[0m\u001b[32m'70041f7130ef2e27cc1e6156ffa78d3a1f3e0c3cb7763cc314cda6a47a029350'\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1;39m)\u001b[0m\u001b[39m,\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[39m<NodeRelationship.PREVIOUS: \u001b[0m\u001b[32m'2'\u001b[0m\u001b[39m>: \u001b[0m\u001b[1;35mRelatedNodeInfo\u001b[0m\u001b[1;39m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mnode_id\u001b[0m\u001b[39m=\u001b[0m\u001b[32m'06c2ded5-e46a-4278-92ea-eed9bc78b9f6'\u001b[0m\u001b[39m,\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mnode_type\u001b[0m\u001b[39m=<ObjectType.TEXT: \u001b[0m\u001b[32m'1'\u001b[0m\u001b[1m>\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mmetadata\u001b[0m=\u001b[1m{\u001b[0m\u001b[33m...\u001b[0m\u001b[1m}\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mhash\u001b[0m=\u001b[32m'c51973e676e01da1fc3504107653ce74d198dbafa50eedd32cf5d5e3c7a82e0b'\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1m)\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33m...\u001b[0m +\u001b[1;36m1\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[1m}\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mtext\u001b[0m=\u001b[32m'ABSTRACT\\nAccurate document layout analysis is a key requirement for highquality PDF document conversion. '\u001b[0m+\u001b[1;36m1499\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mmimetype\u001b[0m=\u001b[32m'text/plain'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mstart_char_idx\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mend_char_idx\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mtext_template\u001b[0m=\u001b[32m'\u001b[0m\u001b[32m{\u001b[0m\u001b[32mmetadata_str\u001b[0m\u001b[32m}\u001b[0m\u001b[32m\\n\\n\u001b[0m\u001b[32m{\u001b[0m\u001b[32mcontent\u001b[0m\u001b[32m}\u001b[0m\u001b[32m'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mmetadata_template\u001b[0m=\u001b[32m'\u001b[0m\u001b[32m{\u001b[0m\u001b[32mkey\u001b[0m\u001b[32m}\u001b[0m\u001b[32m: \u001b[0m\u001b[32m{\u001b[0m\u001b[32mvalue\u001b[0m\u001b[32m}\u001b[0m\u001b[32m'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mmetadata_seperator\u001b[0m=\u001b[32m'\\n'\u001b[0m\n",
|
||||
"\u001b[1m)\u001b[0m\n"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"from llama_index.core import StorageContext, VectorStoreIndex\n",
|
||||
"from llama_index.core.ingestion import IngestionPipeline\n",
|
||||
"\n",
|
||||
"if INGEST:\n",
|
||||
" # in this case we ingest the data into the vector store\n",
|
||||
" docs = reader.load_data(\n",
|
||||
" file_path=\"https://arxiv.org/pdf/2206.01062\", # DocLayNet paper\n",
|
||||
" )\n",
|
||||
" console.rule(f\"example `Document`:\")\n",
|
||||
" pprint(docs, max_length=1, max_string=50, max_depth=4)\n",
|
||||
" storage_context = StorageContext.from_defaults(vector_store=vector_store)\n",
|
||||
" index = VectorStoreIndex.from_documents(\n",
|
||||
" documents=docs,\n",
|
||||
" embed_model=embed_model,\n",
|
||||
" storage_context=storage_context,\n",
|
||||
" pipeline = IngestionPipeline(\n",
|
||||
" transformations=transformations,\n",
|
||||
" vector_store=vector_store,\n",
|
||||
" )\n",
|
||||
" nodes = pipeline.run(documents=docs)\n",
|
||||
" print(f\"num of Nodes (chunks): {len(nodes)}\")\n",
|
||||
" console.rule(f\"example Node (chunk):\")\n",
|
||||
" pprint(\n",
|
||||
" nodes[6],\n",
|
||||
" max_length=2,\n",
|
||||
" max_string=105,\n",
|
||||
" max_depth=3,\n",
|
||||
" )\n",
|
||||
" index = VectorStoreIndex(nodes=nodes, embed_model=embed_model)\n",
|
||||
"\n",
|
||||
"else:\n",
|
||||
" # in this case we just load the vector store index\n",
|
||||
" index = VectorStoreIndex.from_vector_store(\n",
|
||||
@ -336,12 +593,12 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### LLM"
|
||||
"## LLM"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 11,
|
||||
"execution_count": 14,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
@ -364,76 +621,64 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 12,
|
||||
"execution_count": 15,
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">Response</span><span style=\"font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">response</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'80863 pages were human annotated.'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">response</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'The 80K pages were human annotated.'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">source_nodes</span>=<span style=\"font-weight: bold\">[</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">NodeWithScore</span><span style=\"font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">node</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">TextNode</span><span style=\"font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">id_</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'8874a117-d181-4f4f-a30b-0b5604370d77'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">id_</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'431c2d9a-fc55-40ad-a635-d12aeb6b1cec'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">embedding</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata</span>=<span style=\"font-weight: bold\">{</span><span style=\"color: #808000; text-decoration-color: #808000\">...</span><span style=\"font-weight: bold\">}</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">excluded_embed_metadata_keys</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #808000; text-decoration-color: #808000\">...</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">excluded_llm_metadata_keys</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #808000; text-decoration-color: #808000\">...</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">relationships</span>=<span style=\"font-weight: bold\">{</span><span style=\"color: #808000; text-decoration-color: #808000\">...</span><span style=\"font-weight: bold\">}</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">text</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'3 THE DOCLAYNET DATASET\\n\\nDocLayNet contains 80863 PDF pages. Among these, 7059 carry two instances of human annotations, and 1591 carry three. This amounts to 91104 total annotation instances. The annotations provide layout information in the shape o'</span>+<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">5775</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata</span>=<span style=\"font-weight: bold\">{</span><span style=\"color: #008000; text-decoration-color: #008000\">'path'</span>: <span style=\"color: #008000; text-decoration-color: #008000\">'$.main-text[74]'</span><span style=\"font-weight: bold\">}</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">excluded_embed_metadata_keys</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008000; text-decoration-color: #008000\">'path'</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">excluded_llm_metadata_keys</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008000; text-decoration-color: #008000\">'path'</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">relationships</span>=<span style=\"font-weight: bold\">{<</span><span style=\"color: #ff00ff; text-decoration-color: #ff00ff; font-weight: bold\">NodeRelationship.SOURCE:</span><span style=\"color: #000000; text-decoration-color: #000000\"> </span><span style=\"color: #008000; text-decoration-color: #008000\">'1'</span><span style=\"font-weight: bold\">></span>: <span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">RelatedNodeInfo</span><span style=\"font-weight: bold\">(</span><span style=\"color: #808000; text-decoration-color: #808000\">...</span><span style=\"font-weight: bold\">)</span>, <span style=\"color: #808000; text-decoration-color: #808000\">...</span> +<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">2</span><span style=\"font-weight: bold\">}</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">text</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'4 ANNOTATION CAMPAIGN\\nThe complete annotation guideline is over 100 pages long and a detailed description is obviously out of scope for this paper. Nevertheless, it will be made publicly available alongside with DocLayNet for future reference.'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">mimetype</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'text/plain'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">start_char_idx</span>=<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">9089</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">end_char_idx</span>=<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">15114</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">start_char_idx</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">end_char_idx</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">text_template</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'{metadata_str}\\n\\n{content}'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata_template</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'{key}: {value}'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata_seperator</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'\\n'</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"font-weight: bold\">)</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">score</span>=<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">0.7367570400238037</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">score</span>=<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">0.7986579795165596</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"font-weight: bold\">)</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">...</span> +<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">1</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata</span>=<span style=\"font-weight: bold\">{</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #008000; text-decoration-color: #008000\">'8874a117-d181-4f4f-a30b-0b5604370d77'</span>: <span style=\"font-weight: bold\">{</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #008000; text-decoration-color: #008000\">'dl_doc_hash'</span>: <span style=\"color: #008000; text-decoration-color: #008000\">'5dfbd8c115a15fd3396b68409124cfee29fc8efac7b5c846634ff924e635e0dc'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">...</span> +<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">1</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"font-weight: bold\">}</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">...</span> +<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">1</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"font-weight: bold\">}</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">metadata</span>=<span style=\"font-weight: bold\">{</span><span style=\"color: #008000; text-decoration-color: #008000\">'431c2d9a-fc55-40ad-a635-d12aeb6b1cec'</span>: <span style=\"font-weight: bold\">{</span><span style=\"color: #008000; text-decoration-color: #008000\">'path'</span>: <span style=\"color: #008000; text-decoration-color: #008000\">'$.main-text[74]'</span><span style=\"font-weight: bold\">}</span>, <span style=\"color: #808000; text-decoration-color: #808000\">...</span> +<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">1</span><span style=\"font-weight: bold\">}</span>\n",
|
||||
"<span style=\"font-weight: bold\">)</span>\n",
|
||||
"</pre>\n"
|
||||
],
|
||||
"text/plain": [
|
||||
"\u001b[1;35mResponse\u001b[0m\u001b[1m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mresponse\u001b[0m=\u001b[32m'80863 pages were human annotated.'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mresponse\u001b[0m=\u001b[32m'The 80K pages were human annotated.'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33msource_nodes\u001b[0m=\u001b[1m[\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1;35mNodeWithScore\u001b[0m\u001b[1m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mnode\u001b[0m=\u001b[1;35mTextNode\u001b[0m\u001b[1m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mid_\u001b[0m=\u001b[32m'8874a117-d181-4f4f-a30b-0b5604370d77'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mid_\u001b[0m=\u001b[32m'431c2d9a-fc55-40ad-a635-d12aeb6b1cec'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33membedding\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mmetadata\u001b[0m=\u001b[1m{\u001b[0m\u001b[33m...\u001b[0m\u001b[1m}\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mexcluded_embed_metadata_keys\u001b[0m=\u001b[1m[\u001b[0m\u001b[33m...\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mexcluded_llm_metadata_keys\u001b[0m=\u001b[1m[\u001b[0m\u001b[33m...\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mrelationships\u001b[0m=\u001b[1m{\u001b[0m\u001b[33m...\u001b[0m\u001b[1m}\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mtext\u001b[0m=\u001b[32m'3 THE DOCLAYNET DATASET\\n\\nDocLayNet contains 80863 PDF pages. Among these, 7059 carry two instances of human annotations, and 1591 carry three. This amounts to 91104 total annotation instances. The annotations provide layout information in the shape o'\u001b[0m+\u001b[1;36m5775\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mmetadata\u001b[0m=\u001b[1m{\u001b[0m\u001b[32m'path'\u001b[0m: \u001b[32m'$.main-text\u001b[0m\u001b[32m[\u001b[0m\u001b[32m74\u001b[0m\u001b[32m]\u001b[0m\u001b[32m'\u001b[0m\u001b[1m}\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mexcluded_embed_metadata_keys\u001b[0m=\u001b[1m[\u001b[0m\u001b[32m'path'\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mexcluded_llm_metadata_keys\u001b[0m=\u001b[1m[\u001b[0m\u001b[32m'path'\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mrelationships\u001b[0m=\u001b[1m{\u001b[0m\u001b[1m<\u001b[0m\u001b[1;95mNodeRelationship.SOURCE:\u001b[0m\u001b[39m \u001b[0m\u001b[32m'1'\u001b[0m\u001b[1m>\u001b[0m: \u001b[1;35mRelatedNodeInfo\u001b[0m\u001b[1m(\u001b[0m\u001b[33m...\u001b[0m\u001b[1m)\u001b[0m, \u001b[33m...\u001b[0m +\u001b[1;36m2\u001b[0m\u001b[1m}\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mtext\u001b[0m=\u001b[32m'4 ANNOTATION CAMPAIGN\\nThe complete annotation guideline is over 100 pages long and a detailed description is obviously out of scope for this paper. Nevertheless, it will be made publicly available alongside with DocLayNet for future reference.'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mmimetype\u001b[0m=\u001b[32m'text/plain'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mstart_char_idx\u001b[0m=\u001b[1;36m9089\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mend_char_idx\u001b[0m=\u001b[1;36m15114\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mstart_char_idx\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mend_char_idx\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mtext_template\u001b[0m=\u001b[32m'\u001b[0m\u001b[32m{\u001b[0m\u001b[32mmetadata_str\u001b[0m\u001b[32m}\u001b[0m\u001b[32m\\n\\n\u001b[0m\u001b[32m{\u001b[0m\u001b[32mcontent\u001b[0m\u001b[32m}\u001b[0m\u001b[32m'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mmetadata_template\u001b[0m=\u001b[32m'\u001b[0m\u001b[32m{\u001b[0m\u001b[32mkey\u001b[0m\u001b[32m}\u001b[0m\u001b[32m: \u001b[0m\u001b[32m{\u001b[0m\u001b[32mvalue\u001b[0m\u001b[32m}\u001b[0m\u001b[32m'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ │ \u001b[0m\u001b[33mmetadata_seperator\u001b[0m=\u001b[32m'\\n'\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[1m)\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mscore\u001b[0m=\u001b[1;36m0\u001b[0m\u001b[1;36m.7367570400238037\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mscore\u001b[0m=\u001b[1;36m0\u001b[0m\u001b[1;36m.7986579795165596\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1m)\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33m...\u001b[0m +\u001b[1;36m1\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mmetadata\u001b[0m=\u001b[1m{\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[32m'8874a117-d181-4f4f-a30b-0b5604370d77'\u001b[0m: \u001b[1m{\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[32m'dl_doc_hash'\u001b[0m: \u001b[32m'5dfbd8c115a15fd3396b68409124cfee29fc8efac7b5c846634ff924e635e0dc'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33m...\u001b[0m +\u001b[1;36m1\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1m}\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[33m...\u001b[0m +\u001b[1;36m1\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[1m}\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mmetadata\u001b[0m=\u001b[1m{\u001b[0m\u001b[32m'431c2d9a-fc55-40ad-a635-d12aeb6b1cec'\u001b[0m: \u001b[1m{\u001b[0m\u001b[32m'path'\u001b[0m: \u001b[32m'$.main-text\u001b[0m\u001b[32m[\u001b[0m\u001b[32m74\u001b[0m\u001b[32m]\u001b[0m\u001b[32m'\u001b[0m\u001b[1m}\u001b[0m, \u001b[33m...\u001b[0m +\u001b[1;36m1\u001b[0m\u001b[1m}\u001b[0m\n",
|
||||
"\u001b[1m)\u001b[0m\n"
|
||||
]
|
||||
},
|
||||
@ -442,9 +687,173 @@
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"query_engine = index.as_query_engine(llm=llm)\n",
|
||||
"from llama_index.core import PromptTemplate\n",
|
||||
"\n",
|
||||
"TEXT_QA_TEMPLATE_STR = \"Context information is below.\\n---------------------\\n{context_str}\\n---------------------\\nGiven the context information and not prior knowledge, answer the query.\\nQuery: {query_str}\\nAnswer:\\n\"\n",
|
||||
"\n",
|
||||
"query_engine = index.as_query_engine(\n",
|
||||
" llm=llm,\n",
|
||||
" text_qa_template=PromptTemplate(TEXT_QA_TEMPLATE_STR),\n",
|
||||
")\n",
|
||||
"query_res = query_engine.query(\"How many pages were human annotated?\")\n",
|
||||
"pprint(query_res, max_length=1, max_string=250, max_depth=4)"
|
||||
"pprint(query_res, max_length=1, max_string=250, max_depth=5)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Fetching Docling native layout info\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 16,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import jsonpath_ng\n",
|
||||
"from docling_core.types import BaseText, Document, Ref\n",
|
||||
"\n",
|
||||
"NativeNode = BaseText | Ref\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"def get_native_node(dl_doc: Document, path: str) -> NativeNode:\n",
|
||||
" jsonpath_expr = jsonpath_ng.parse(path)\n",
|
||||
" jsonpath_res = [match.value for match in jsonpath_expr.find(dl_doc.model_dump())]\n",
|
||||
" if (num_res := len(jsonpath_res)) == 0:\n",
|
||||
" raise RuntimeError(f\"No results found for {path}\")\n",
|
||||
" elif num_res > 1:\n",
|
||||
" # currently only single result supported\n",
|
||||
" raise RuntimeError(f\"Multiple results found for {path}\")\n",
|
||||
" jres = jsonpath_res[0]\n",
|
||||
" return TypeAdapter(NativeNode).validate_python(jres)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 17,
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #00ff00; text-decoration-color: #00ff00\">────────────────────────────────── </span><span style=\"color: #808000; text-decoration-color: #808000\">dl_doc_hash</span>=<span style=\"color: #800080; text-decoration-color: #800080\">5dfbd8c</span><span style=\"color: #808000; text-decoration-color: #808000\">...</span> <span style=\"color: #808000; text-decoration-color: #808000\">path</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'$.main-text[74]'</span><span style=\"color: #00ff00; text-decoration-color: #00ff00\"> ──────────────────────────────────</span>\n",
|
||||
"</pre>\n"
|
||||
],
|
||||
"text/plain": [
|
||||
"\u001b[92m────────────────────────────────── \u001b[0m\u001b[33mdl_doc_hash\u001b[0m=\u001b[35m5dfbd8c\u001b[0m\u001b[33m...\u001b[0m \u001b[33mpath\u001b[0m=\u001b[32m'$.main-text\u001b[0m\u001b[32m[\u001b[0m\u001b[32m74\u001b[0m\u001b[32m]\u001b[0m\u001b[32m'\u001b[0m\u001b[92m ──────────────────────────────────\u001b[0m\n"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">BaseText</span><span style=\"font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">text</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'The complete annotation guideline is over 100 pages long and a detailed description is obviously out of scope for this paper. Nevertheless, it will be made publicly available alongside with DocLayNet for future reference.'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">obj_type</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'paragraph'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">name</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'Text'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">font</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">prov</span>=<span style=\"font-weight: bold\">[</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">Prov</span><span style=\"font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">bbox</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">52.994422912597656</span>, <span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">217.798828125</span>, <span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">295.5625305175781</span>, <span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">259.6097106933594</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">page</span>=<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">5</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">span</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">0</span>, <span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">221</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">ref_s3_data</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"font-weight: bold\">)</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"font-weight: bold\">]</span>\n",
|
||||
"<span style=\"font-weight: bold\">)</span>\n",
|
||||
"</pre>\n"
|
||||
],
|
||||
"text/plain": [
|
||||
"\u001b[1;35mBaseText\u001b[0m\u001b[1m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mtext\u001b[0m=\u001b[32m'The complete annotation guideline is over 100 pages long and a detailed description is obviously out of scope for this paper. Nevertheless, it will be made publicly available alongside with DocLayNet for future reference.'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mobj_type\u001b[0m=\u001b[32m'paragraph'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mname\u001b[0m=\u001b[32m'Text'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mfont\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mprov\u001b[0m=\u001b[1m[\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1;35mProv\u001b[0m\u001b[1m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mbbox\u001b[0m=\u001b[1m[\u001b[0m\u001b[1;36m52.994422912597656\u001b[0m, \u001b[1;36m217.798828125\u001b[0m, \u001b[1;36m295.5625305175781\u001b[0m, \u001b[1;36m259.6097106933594\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mpage\u001b[0m=\u001b[1;36m5\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mspan\u001b[0m=\u001b[1m[\u001b[0m\u001b[1;36m0\u001b[0m, \u001b[1;36m221\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mref_s3_data\u001b[0m=\u001b[3;35mNone\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1m)\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[1m]\u001b[0m\n",
|
||||
"\u001b[1m)\u001b[0m\n"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #00ff00; text-decoration-color: #00ff00\">────────────────────────────────── </span><span style=\"color: #808000; text-decoration-color: #808000\">dl_doc_hash</span>=<span style=\"color: #800080; text-decoration-color: #800080\">5dfbd8c</span><span style=\"color: #808000; text-decoration-color: #808000\">...</span> <span style=\"color: #808000; text-decoration-color: #808000\">path</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'$.main-text[79]'</span><span style=\"color: #00ff00; text-decoration-color: #00ff00\"> ──────────────────────────────────</span>\n",
|
||||
"</pre>\n"
|
||||
],
|
||||
"text/plain": [
|
||||
"\u001b[92m────────────────────────────────── \u001b[0m\u001b[33mdl_doc_hash\u001b[0m=\u001b[35m5dfbd8c\u001b[0m\u001b[33m...\u001b[0m \u001b[33mpath\u001b[0m=\u001b[32m'$.main-text\u001b[0m\u001b[32m[\u001b[0m\u001b[32m79\u001b[0m\u001b[32m]\u001b[0m\u001b[32m'\u001b[0m\u001b[92m ──────────────────────────────────\u001b[0m\n"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"text/html": [
|
||||
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">BaseText</span><span style=\"font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">text</span>=<span style=\"color: #008000; text-decoration-color: #008000\">\"Phase 4: Production annotation. The previously selected 80K pages were annotated with the defined 11 class labels by 32 annotators. This production phase took around three months to complete. All annotations were created online through CCS, which visualises the programmatic PDF text-cells as an overlay on the page. The page annotation are obtained by drawing rectangular bounding-boxes, as shown in Figure 3. With regard to the annotation practices, we implemented a few constraints and capabilities on the tooling level. First, we only allow non-overlapping, vertically oriented, rectangular boxes. For the large majority of documents, this constraint was sufficient and it speeds up the annotation considerably in comparison with arbitrary segmentation shapes. Second, annotator staff were not able to see each other's annotations. This was enforced by design to avoid any bias in the annotation, which could skew the numbers of the inter-annotator agreement (see Table 1). We wanted\"</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">obj_type</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'paragraph'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">name</span>=<span style=\"color: #008000; text-decoration-color: #008000\">'Text'</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">font</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"color: #808000; text-decoration-color: #808000\">prov</span>=<span style=\"font-weight: bold\">[</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\">Prov</span><span style=\"font-weight: bold\">(</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">bbox</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">317.00592041015625</span>, <span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">82.7375717163086</span>, <span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">559.7149047851562</span>, <span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">245.28392028808594</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">page</span>=<span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">5</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">span</span>=<span style=\"font-weight: bold\">[</span><span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">0</span>, <span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">987</span><span style=\"font-weight: bold\">]</span>,\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">ref_s3_data</span>=<span style=\"color: #800080; text-decoration-color: #800080; font-style: italic\">None</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ │ </span><span style=\"font-weight: bold\">)</span>\n",
|
||||
"<span style=\"color: #7fbf7f; text-decoration-color: #7fbf7f\">│ </span><span style=\"font-weight: bold\">]</span>\n",
|
||||
"<span style=\"font-weight: bold\">)</span>\n",
|
||||
"</pre>\n"
|
||||
],
|
||||
"text/plain": [
|
||||
"\u001b[1;35mBaseText\u001b[0m\u001b[1m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mtext\u001b[0m=\u001b[32m\"Phase\u001b[0m\u001b[32m 4: Production annotation. The previously selected 80K pages were annotated with the defined 11 class labels by 32 annotators. This production phase took around three months to complete. All annotations were created online through CCS, which visualises the programmatic PDF text-cells as an overlay on the page. The page annotation are obtained by drawing rectangular bounding-boxes, as shown in Figure 3. With regard to the annotation practices, we implemented a few constraints and capabilities on the tooling level. First, we only allow non-overlapping, vertically oriented, rectangular boxes. For the large majority of documents, this constraint was sufficient and it speeds up the annotation considerably in comparison with arbitrary segmentation shapes. Second, annotator staff were not able to see each other's annotations. This was enforced by design to avoid any bias in the annotation, which could skew the numbers of the inter-annotator agreement \u001b[0m\u001b[32m(\u001b[0m\u001b[32msee Table 1\u001b[0m\u001b[32m)\u001b[0m\u001b[32m. We wanted\"\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mobj_type\u001b[0m=\u001b[32m'paragraph'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mname\u001b[0m=\u001b[32m'Text'\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mfont\u001b[0m=\u001b[3;35mNone\u001b[0m,\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[33mprov\u001b[0m=\u001b[1m[\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1;35mProv\u001b[0m\u001b[1m(\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mbbox\u001b[0m=\u001b[1m[\u001b[0m\u001b[1;36m317.00592041015625\u001b[0m, \u001b[1;36m82.7375717163086\u001b[0m, \u001b[1;36m559.7149047851562\u001b[0m, \u001b[1;36m245.28392028808594\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mpage\u001b[0m=\u001b[1;36m5\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mspan\u001b[0m=\u001b[1m[\u001b[0m\u001b[1;36m0\u001b[0m, \u001b[1;36m987\u001b[0m\u001b[1m]\u001b[0m,\n",
|
||||
"\u001b[2;32m│ │ │ \u001b[0m\u001b[33mref_s3_data\u001b[0m=\u001b[3;35mNone\u001b[0m\n",
|
||||
"\u001b[2;32m│ │ \u001b[0m\u001b[1m)\u001b[0m\n",
|
||||
"\u001b[2;32m│ \u001b[0m\u001b[1m]\u001b[0m\n",
|
||||
"\u001b[1m)\u001b[0m\n"
|
||||
]
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"from docling_core.types import Document as DLDocument\n",
|
||||
"\n",
|
||||
"dl_docs = [DLDocument.model_validate_json(doc.text) for doc in docs]\n",
|
||||
"for retr_item in query_res.source_nodes:\n",
|
||||
" source_metadata = DocumentMetadata.model_validate(\n",
|
||||
" retr_item.node.source_node.metadata\n",
|
||||
" )\n",
|
||||
" dl_doc_hash = source_metadata.dl_doc_hash\n",
|
||||
" dl_doc = [d for d in dl_docs if d.file_info.document_hash == dl_doc_hash][0]\n",
|
||||
" path = NodeMetadata.model_validate(retr_item.node.metadata).path\n",
|
||||
" native_node = get_native_node(dl_doc=dl_doc, path=path)\n",
|
||||
" console.rule(f\"dl_doc_hash={dl_doc_hash[:7]}...\\n{path=}\")\n",
|
||||
" pprint(native_node)"
|
||||
]
|
||||
},
|
||||
{
|
||||
|
||||
27
poetry.lock
generated
27
poetry.lock
generated
@ -2071,6 +2071,20 @@ files = [
|
||||
[package.dependencies]
|
||||
jsonpointer = ">=1.9"
|
||||
|
||||
[[package]]
|
||||
name = "jsonpath-ng"
|
||||
version = "1.6.1"
|
||||
description = "A final implementation of JSONPath for Python that aims to be standard compliant, including arithmetic and binary comparison operators and providing clear AST for metaprogramming."
|
||||
optional = false
|
||||
python-versions = "*"
|
||||
files = [
|
||||
{file = "jsonpath-ng-1.6.1.tar.gz", hash = "sha256:086c37ba4917304850bd837aeab806670224d3f038fe2833ff593a672ef0a5fa"},
|
||||
{file = "jsonpath_ng-1.6.1-py3-none-any.whl", hash = "sha256:8f22cd8273d7772eea9aaa84d922e0841aa36fdb8a2c6b7f6c3791a16a9bc0be"},
|
||||
]
|
||||
|
||||
[package.dependencies]
|
||||
ply = "*"
|
||||
|
||||
[[package]]
|
||||
name = "jsonpointer"
|
||||
version = "3.0.0"
|
||||
@ -3953,6 +3967,17 @@ files = [
|
||||
dev = ["pre-commit", "tox"]
|
||||
testing = ["pytest", "pytest-benchmark"]
|
||||
|
||||
[[package]]
|
||||
name = "ply"
|
||||
version = "3.11"
|
||||
description = "Python Lex & Yacc"
|
||||
optional = false
|
||||
python-versions = "*"
|
||||
files = [
|
||||
{file = "ply-3.11-py2.py3-none-any.whl", hash = "sha256:096f9b8350b65ebd2fd1346b12452efe5b9607f7482813ffca50c22722a807ce"},
|
||||
{file = "ply-3.11.tar.gz", hash = "sha256:00c7c1aaa88358b9c765b6d3000c6eec0ba42abca5351b095321aef446081da3"},
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pre-commit"
|
||||
version = "3.8.0"
|
||||
@ -7169,4 +7194,4 @@ examples = ["langchain-huggingface", "langchain-milvus", "langchain-text-splitte
|
||||
[metadata]
|
||||
lock-version = "2.0"
|
||||
python-versions = "^3.10"
|
||||
content-hash = "fe428eb26a2f4a9a353a0e04d7e9a6d3740401cf4bbc9e6d6999de1646884884"
|
||||
content-hash = "a1be777babc1c8ab76f083ae3b1c8d36cb40db0269ade8f14828163e0f311f6a"
|
||||
|
||||
@ -81,6 +81,7 @@ nbqa = "^1.9.0"
|
||||
[tool.poetry.group.examples.dependencies]
|
||||
datasets = "^2.21.0"
|
||||
rich = "^13.8.1"
|
||||
jsonpath-ng = "^1.6.1"
|
||||
|
||||
[tool.poetry.extras]
|
||||
examples = [
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user