Update README for v2.0

README.md

@@ -1,90 +1,83 @@
 ---
-language:
-- en
-pretty_name: Coq-HoTT Dataset
-tags:
-- theorem-proving
-- formal-methods
-- hott
-- coq
 license: bsd-2-clause
 task_categories:
-- text-generation
-- feature-extraction
-- other
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
-dataset_info:
-  features:
-  - name: fact
-    dtype: string
-  - name: type
-    dtype: string
-  - name: library
-    dtype: string
-  - name: imports
-    list: string
-  - name: filename
-    dtype: string
-  - name: symbolic_name
-    dtype: string
-  - name: docstring
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 3581192
-    num_examples: 11533
-  download_size: 919179
-  dataset_size: 3581192
+  - text-generation
+  - feature-extraction
+language:
+  - en
+tags:
+  - theorem-proving
+  - formal-methods
+  - coq
+  - hott
+  - homotopy-type-theory
+size_categories:
+  - 10K<n<100K
 ---
 
-# Coq-HoTT Dataset
+# Coq-HoTT
+
+Structured dataset of formalizations from the Coq-HoTT library (Homotopy Type Theory in Coq).
 
 ## Dataset Description
 
-The Coq-HoTT Dataset is derived from the Coq-HoTT repository, focusing on the formalization of Homotopy Type Theory in the Coq proof assistant. This dataset processes `.v` files from the theories directory to extract mathematical content in a structured format.
+- **Source:** [HoTT/Coq-HoTT](https://github.com/HoTT/Coq-HoTT)
+- **Entries:** 11,533
+- **Files processed:** 583
+- **License:** BSD 2-Clause
+
+## Schema
+
+| Column | Type | Description |
+|--------|------|-------------|
+| fact | string | Full declaration (name, signature, body) |
+| type | string | Definition, Lemma, Theorem, Class, Instance, etc. |
+| library | string | Sub-library within HoTT (Basics, Types, Spaces, etc.) |
+| imports | list | Require Import statements |
+| filename | string | Source file path |
+| symbolic_name | string | Declaration identifier |
+| docstring | string | Documentation comment (41% coverage) |
 
-This work builds upon the format established by Andreas Florath (@florath) in his Coq Facts, Propositions and Proofs dataset.
+## Statistics
 
-### Dataset Structure
+### By Type
+| Type | Count |
+|------|-------|
+| Definition | 6,726 |
+| Instance | 2,355 |
+| Lemma | 1,308 |
+| Class | 363 |
+| Fixpoint | 141 |
+| Record | 139 |
+| Theorem | 103 |
+| Proposition | 67 |
+| Inductive | 39 |
+| Other | 292 |
 
-The dataset includes the following fields:
-- `fact`: The complete mathematical statement, including type (Definition/Lemma/Theorem), name, and body
-- `imports`: The Require Import statements from the source file
-- `filename`: The source file name within the theories directory
-- `symbolic_name`: The identifier of the mathematical object
-- `__index_level_0__`: Sequential index for the dataset
+### Docstring Coverage
+- **4,756 entries (41%)** have documentation
 
-### Example Row
+## Example Entry
 
-```
-fact: "Definition minimal(n : nat) : Type := forall m : nat, P m -> n <= m."
-imports: "Require Import HoTT.Basics HoTT.Types. Require Import HoTT.Truncations.Core. Require Import HoTT.Spaces.Nat.Core."
-filename: "BoundedSearch.v"
-symbolic_name: "minimal"
-__index_level_0__: 0
-```
 
-### Source Code
 
-The dataset was generated using a custom Python script that processes the Coq-HoTT repository's theories directory, extracting mathematical content while preserving the structure and relationships between definitions, imports, and their source files.
-``` to clean and organize content. `.v` files were processed for mathematical content, while `.md` files were included to retain important documentation and context.
+## Use Cases
 
-### Usage:
+- Retrieval / RAG for Coq theorem proving
+- Homotopy Type Theory research
+- Training embeddings for formal mathematics
+- Documentation generation
 
-This dataset is designed for:
-- Formal Methods Research: Analyzing formal proofs and definitions in Homotopy Type Theory.
-- Machine Learning Applications: Training models on formal verification, code completion, and theorem proving tasks.
-- Educational Purposes: Providing structured examples of Coq formalizations.
+## Citation
 
-### License:
 
-This dataset is distributed under the  BSD 2-clause licence, aligning with the license of the original Coq-HoTT repository.
 
-### Acknowledgments:
+## Changelog
 
-- Original repository: Coq-HoTT (https://github.com/HoTT/Coq-HoTT)
-- Inspiration: Hugging Face user Andreas Florath (@florath) and his dataset on Coq.
+- **v2.0** (Jan 2025): Re-extraction with improved schema
+  - 7,848 -> 11,533 entries (+47%)
+  - Added library and docstring columns
+  - Changed imports from string to list
+  - Removed index_level column
+  - fact now includes full signature
+- **v1.0** (Dec 2024): Initial release