skills.rest
xlinh2301
xlinh2301COMMUNITY
Explore all skills by this author →
GitHub

predictingthepast

Restore, attribute, date, and contextualize ancient Latin and Greek inscriptions using Aeneas and Ithaca models.

Updated Sep 9, 2026
Search Tags:#epigraphy#latin#ancient greek#aeneas#text restoration#ithaca#ancient inscriptions
One-click install
npx skills add https://github.com/xlinh2301/EditCTC --skill predictingthepast-xlinh2301
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: predictingthepast
Source: https://github.com/xlinh2301/EditCTC/tree/main/.agents/skills/predictingthepast
Command: npx skills add https://github.com/xlinh2301/EditCTC --skill predictingthepast-xlinh2301

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires jax, predictingthepast, and includes scripts (resource) and references (resource) components.

What problem does it solve? Damaged ancient inscriptions are difficult to restore, date, and locate geographically without deep epigraphic expertise. This Skill runs Google's Aeneas (Latin) and Ithaca (Ancient Greek) neural models locally to restore missing characters, predict geographical and chronological origins, retrieve parallel inscriptions, and generate text embeddings. ## Core Features & Use Cases - Text Restoration: Fill known-length gaps (?) and unknown-length gaps (#) in damaged inscriptions via beam search, with ranked candidate predictions and saliency scores. - Attribution & Dating: Predict the top geographical regions and most probable year (800 BCE to 800 CE) for an inscription, with per-character saliency. - Contextualization & Embedding: Retrieve the top-K parallel inscriptions with full texts, Trismegistos IDs, and similarity scores, plus a 384-dimensional text embedding vector. - Use Case: A historian has a fragmentary Latin dedication like donat in ??????????rtis. The Skill preprocesses the text, restores it as donat in templum martis, attributes it to Britannia around 295 CE, and produces an interactive HTML dashboard of the results. ## Quick Start Ask the assistant to restore and date your ancient inscription, providing the text with ? marking each missing character, for example: restore and attribute the Latin text "donat in ??????????rtis".

Frequently Asked Questions about predictingthepast

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I restore missing characters in a Latin inscription?▼

Mark each missing character with `?` for known-length gaps or `#` for unknown-length gaps, then run inference with the --restore flag. The model returns ranked beam-search candidates with scores, and the top prediction is shown with restored characters highlighted.

What is the difference between Aeneas and Ithaca models?▼

Aeneas handles Latin inscriptions and Ithaca handles Ancient Greek inscriptions; both perform restoration, attribution, dating, and contextualization. Select the model with the --language flag set to latin or greek.

Can Ithaca or Aeneas date an ancient text?▼

Yes, the --attribute flag produces chronological attribution covering 800 BCE to 800 CE in 10-year bins. Output includes the top predicted year, a probability-weighted average year, and the full year distribution.

What are the input length limits for inscription restoration?▼

Input must be at least 25 characters (pad with dashes if shorter) and texts over 750 characters are split into overlapping windows processed independently. Restoration with more than 10 `?` markers becomes slow, roughly 10 seconds per additional character on a high-end CPU.

Why does restoration fail when I use # and ? together?▼

The markers cannot be adjacent: consecutive `##`, or adjacent `?#` and `#?` combinations are rejected. Use a single `#` per unknown-length gap and separate it from fixed-length `?` gaps with other characters.

Related Skills

Explore More Education & Research Skills→
lucifertrjlucifertrj
17

nltk-nlp-text-cleaning

Clean, tokenize, and normalize English text with NLTK for NLP pipelines.

Community
Advanced
LAF-USLAF-US

huggingface-paper-publisher

Index arXiv papers and publish linked research articles on Hugging Face Hub.

Official
Advanced
Best6668Best6668
1

model-validation

Automate end-to-end predictive model validation with holdout and cross-validation.

Community
Advanced
levelheadslevelheads

linear-a-decipherer

Analyze Linear A sign sequences across Luwian, Semitic, Pre-Greek, and Proto-Greek frameworks.

Official
Advanced