• Fast and effective math search, built on top of a state-of-the-art formula retrieval model.
  • Being able to cache on-disk index and specify the memory usage limit.
  • Compressed and compact on-disk math index. Both in-memory and on-disk indices use specially designed compression schema and apply skip-list to boost search performance.
  • Spceialized dynamic pruning strategy for amazing math-aware structural search efficency.
  • Chinese tokeniser available, multi-bytes awareness from the beginning of design.
  • Fulltext search is based on Indri project (C++), mature and fast.
  • Memory leakage checked, search daemon has zero unfree.
  • Robust TeX parser to convert user created math content to search-optimized representation.
  • Math commutative rules awareness.
  • Math symbol alpha-equivalence awareness.
  • Term proximity awareness.
  • Search results highlight.
  • Scalable to multiple nodes across many machines.

Special thanks to people who have contributed to this project:

  • yzhan018 who submitted the initial FOR-delta implementation.
  • Sil who helped to create the cralwer script for AoPS website and fixed many issues regarding to query input and TeX rendering.