A Semantic Search Engine for Mathlib4
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866929696241876992 |
|---|---|
| author | Gao, Guoxiong Ju, Haocheng Jiang, Jiedong Qin, Zihan Dong, Bin |
| author_facet | Gao, Guoxiong Ju, Haocheng Jiang, Jiedong Qin, Zihan Dong, Bin |
| contents | The interactive theorem prover Lean enables the verification of formal mathematical proofs and is backed by an expanding community. Central to this ecosystem is its mathematical library, mathlib4, which lays the groundwork for the formalization of an expanding range of mathematical theories. However, searching for theorems in mathlib4 can be challenging. To successfully search in mathlib4, users often need to be familiar with its naming conventions or documentation strings. Therefore, creating a semantic search engine that can be used easily by individuals with varying familiarity with mathlib4 is very important. In this paper, we present a semantic search engine (https://leansearch.net/) for mathlib4 that accepts informal queries and finds the relevant theorems. We also establish a benchmark for assessing the performance of various search engines for mathlib4. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_13310 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | A Semantic Search Engine for Mathlib4 Gao, Guoxiong Ju, Haocheng Jiang, Jiedong Qin, Zihan Dong, Bin Information Retrieval Machine Learning Logic in Computer Science The interactive theorem prover Lean enables the verification of formal mathematical proofs and is backed by an expanding community. Central to this ecosystem is its mathematical library, mathlib4, which lays the groundwork for the formalization of an expanding range of mathematical theories. However, searching for theorems in mathlib4 can be challenging. To successfully search in mathlib4, users often need to be familiar with its naming conventions or documentation strings. Therefore, creating a semantic search engine that can be used easily by individuals with varying familiarity with mathlib4 is very important. In this paper, we present a semantic search engine (https://leansearch.net/) for mathlib4 that accepts informal queries and finds the relevant theorems. We also establish a benchmark for assessing the performance of various search engines for mathlib4. |
| title | A Semantic Search Engine for Mathlib4 |
| topic | Information Retrieval Machine Learning Logic in Computer Science |
| url | https://arxiv.org/abs/2403.13310 |