A Semantic Parsing Framework for End-to-End Time Normalization

Xin Su, Sungduk Yu, Phillip Howard, Steven Bethard · 2025

Read A Semantic Parsing Framework for End-to-End Time Normalization by Xin Su, Sungduk Yu, Phillip Howard, Steven Bethard (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-06450.txt.

Description

Xin Su and co-authors recast time reading as code writing, so software can pin phrases like next Friday to exact dates, in a 2025 paper. Time normalization feeds search, question answering, and medical tools. The old schema for time tags strains at nested, event-tied, or split phrases. Their frame, SCATE, builds time sense from small operators that stack.

The team wrote a working SCATE library in Python. Large language models, they show, can write code in that frame for real phrases. A pipeline then lets those models mint large training sets, checked by running the code. Small local models trained on that data beat their giant teachers on the task. The authors claim a path to time tools that are accurate, open to inspection, and cheap to run on site.

Catalog record

CREATORXin Su, Sungduk Yu, Phillip Howard, Steven Bethard
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE34 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-06450

Browse the full library · About the Tower of Babel