Rust Research Paper Parser (rsrpp)

RuSt Research Paper Parser (rsrpp)

The rsrpp library provides a set of tools for parsing research papers.

Quick Start

Pre-requirements

• Poppler: sudo apt install poppler-utils
• OpenCV: sudo apt install libopencv-dev clang libclang-dev

Installation

To start using the rsrpp library, add it to your project's dependencies in the Cargo.toml file:

cargo add rsrpp

Then, import the necessary modules in your code:

extern crate rsrpp;
use rsrpp::parser;

Examples

Here is a simple example of how to use the parser module:

let mut config = ParserConfig::new();
let url = "https://arxiv.org/pdf/1706.03762";
let pages = parse(url, &mut config).await.unwrap(); // Vec<Page>
let sections = Section::from_pages(&pages); // Vec<Section>
let json = serde_json::to_string(&sections).unwrap(); // String

Tests

The library includes a set of tests to ensure its functionality. To run the tests, use the following command:

cargo test

License: MIT

Releases

• LLM-enhanced processing is now enabled by default (ParserConfig::new() sets use_llm: true)
  • If OPENAI_API_KEY is not set, LLM is automatically disabled at runtime
  • Use config.use_llm = false to explicitly disable
• Fixed LLM section validation discarding sections from pages the LLM hadn't examined
  • merge_sections() now uses page-range-aware logic

• Fixed body text loss in Nature-format and non-standard papers:
  • Added section detection fallback for papers without "Abstract" heading using anchor-word matching
  • Added text area degenerate detection to prevent filtering out all blocks
  • Capped table detection regions at 50% of page area to reject false positives from chart gridlines
  • Exempted known section titles from table-region filtering
• Improved math extraction accuracy:
  • Fixed critical bug where LLM-extracted math text was discarded; added trigram-based block alignment
  • Reduced false positives: dates, statistics, section/figure references
  • Added detection for multi-char math functions, ASCII exponents/subscripts, letter fractions, norm notation
  • Unified math output to LaTeX format inside <math> tags
  • Added context-based validation for structure-only pattern matches

• Fixed panic-causing unwrap() calls with proper error handling.

• Fixed Poppler 25.12.0 compatibility on macOS.

• Refactored fix_suffix_hyphens to support 31 compound word suffixes:
  • -based, -driven, -oriented, -aware, -agnostic, -independent, -dependent, -first, -native, -centric, -intensive, -bound, -safe, -free, -proof, -efficient, -optimized, -enabled, -powered, -ready, -capable, -compatible, -compliant, -level, -scale, -wide, -specific, -friendly, -facing, -like, -style
• Added unit tests for suffix hyphenation functionality.

• updated how to extract section titles from PDF.

• restructured rsrpp.parser.
• updated how to extract section titles from PDF.
• updated tests.

• removed init_logger form rsrpp.

• fixed typo.
• introdeced tracing logger.

• Updated rsrpp version for rsrpp-cli.

• Updated dependencies.
• removed build.sh because it requires sudo when installing the crate.

• Fixed a bug: remove unused println!.

• Fixed a bug in xml loop to finish when the file reaches to end.

• Added verbose mode.
• Fixed a bug in the process extracting page number.

• Updated: implemented new errors to handle invalid URLs.

• Updated: The max retry time for saving PDF files has been increased.

• Fix bugs: After converting to PDF, the program now waits until processing is complete.

• Fixed bugs in get_pdf_info.
• Made minor improvements.

• Added cli -> rsrpp-cli.

• Updated the Section module. content: String was replaced by content: Vec<TextBlock>.