Rider

Rider is a search engine for research papers that I built to support my own research workflow.

🔗 Live Demo: http://rider.cs.niu.edu/

Motivation

Existing academic search platforms such as Google Scholar and Semantic Scholar are extremely useful, but they often introduce friction in daily research workflows:

  • Noisy or inconsistent metadata
  • Limited control over conference-level filtering
  • Difficulty focusing on high-quality venues
  • Mixed archival vs. non-archival content
  • Ranking not always aligned with research needs

To address these issues, Rider focuses on clean, curated, and conference-aware search for research papers.

Key Features

Rider is designed around practical research usage:

Users can filter papers by specific venues, including top AI/ML conferences, making it easier to focus on high-quality research output.

Clean and curated dataset

The system emphasizes data quality and consistency, aiming to reduce noise compared to general-purpose academic search engines.

Fast keyword-based retrieval

The current backend is powered by Meilisearch, providing efficient BM25-style ranking for keyword search.

Ongoing semantic search integration

We are currently working on integrating semantic search capabilities with a student collaborator to improve retrieval beyond keyword matching.

System Design (Current)

  • Backend search engine: Meilisearch
  • Ranking model: BM25-style lexical retrieval
  • Data model: curated conference-based paper collection
  • Planned extension: semantic embeddings for hybrid search

Future Work

Rider is actively evolving. Planned improvements include:

  • Semantic search for paper retrieval
  • Better ranking signals for research relevance
  • Expanded conference coverage
  • Improved metadata normalization
  • Potential recommendation features

Status

Active development. Feedback and suggestions are welcome.