Update README: add training as primary use case

README.md

@@ -4,8 +4,6 @@ tags:
   - theorem-proving
   - formal-methods
   - coq
-  - information-theory
-  - coding-theory
 task_categories:
   - text-generation
   - feature-extraction
@@ -17,13 +15,28 @@ size_categories:
 
 # Coq-Infotheo
 
-Structured dataset from **Infotheo** - formalization of information theory and error-correcting codes in Coq.
+Structured dataset from **Infotheo** — Information theory and error-correcting codes.
+
+**4,766 declarations** extracted from Coq source files.
+
+## Applications
+
+- Training language models on formal proofs
+- Fine-tuning theorem provers
+- Retrieval-augmented generation for proof assistants
+- Learning proof embeddings and representations
 
 ## Source
 
 - **Repository:** https://github.com/affeldt-aist/infotheo
-- **License:** LGPL-2.1
 
-## Statistics
+## Schema
 
-- **Total entries:** 4766
+| Column | Type | Description |
+|--------|------|-------------|
+| fact | string | Declaration body |
+| type | string | Lemma, Definition, Theorem, etc. |
+| library | string | Source module |
+| imports | list | Required imports |
+| filename | string | Source file path |
+| symbolic_name | string | Identifier |