Lexoid

Lexoid is an efficient document parsing library that supports both LLM-based and non-LLM-based (static) PDF document parsing.

Dev

1. make setup
2. source .venv/bin/activate
3. To build .whl for testing: poetry build

Usage

Example Notebook

Here's a quick example to parse documents using Lexoid:

from lexoid.api import parse
from lexoid.api import ParserType

parsed_md = parse("https://www.justice.gov/eoir/immigration-law-advisor", parser_type="LLM_PARSE", raw=True)
# or
pdf_path = "path/to/immigration-law-advisor.pdf"
parsed_md = parse(pdf_path, parser_type="LLM_PARSE", raw=True)

print(parsed_md)

Parameters

• path (str): The file path or URL.
• parser_type (str, optional): The type of parser to use ("LLM_PARSE" or "STATIC_PARSE"). Defaults to "LLM_PARSE".
• raw (bool, optional): Whether to return raw text or structured data. Defaults to False.
• pages_per_split (int, optional): Number of pages per split for chunking. Defaults to 4.
• max_threads (int, optional): Maximum number of threads for parallel processing. Defaults to 4.
• **kwargs: Additional arguments for the parser.

Benchmark

Initial results (more updates soon)

Rank	Model/Framework	Similarity	Time (s)
1	gpt-4o	0.799	21.77
2	gemini-1.5-pro	0.742	15.77
3	gpt-4o-mini	0.721	14.86
4	gemini-1.5-flash	0.702	4.56