Structure Interpretation of Text Formats
summary
An extensible program-synthesis system called Unravel framework learns sequences of parametrized parsing rules (Regex-Split, Delimiter-Split, String-Split, JSON rule, Pattern-Partition) to convert mixed and nested text formats into tables; a prototype integrated with Power BI is evaluated on 600+ real files and achieves 90% predictive success and recovers 62% more with user examples, showing multi-rule pipelines are required for about 50% of real inputs.
AI-written summary · about
chapters (14)
- 0:00Start
- 0:28Motivation
- 1:12File example
- 3:37Existing approaches
- 4:31The Unravel approach
- 5:44Parametrized rules & partial structures
- 6:25Black-box rules & extensibility
- 6:47Rule sequence
- 7:00The Unravel algorithm
- 8:03An example (step 1)
- 11:56Interactivity example
- 12:36Implementation & experiments
- 13:24Application in Microsoft Power BI
- 14:40Summary: the Unravel framework
related talks
39mDiagrams as Code 2.0
48mUsing AI in Software Design: How ChatGPT Can Help With Creating a Solution Architecture
10mSporq: An Interactive Environment for Exploring Code Using Query-by-Example
56mInterpreter Pattern In Light of Java 21
43mC4 models as code
41mSRE & Complexification: Where Verbs and Nouns Do Battle
22mMorphir: A Single Language for Business and Technology
27mOperationalizing the Integration of User Interaction Specifications in Modeling Editor Synthesis