Leveraging Large Language Models for Generating Labeled Mineral Site Record Linkage Data CCS: Information systems Entity resolution CCS: Information systems Geographic information systems
Jiyoon Pyo, Yao-Yi Chiang · 2024
Read Leveraging Large Language Models for Generating Labeled Mineral Site Record Linkage Data CCS: Information systems Entity resolution CCS: Information systems Geographic information systems by Jiyoon Pyo, Yao-Yi Chiang (2024) 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-2412-03575.txt.
Description
Jiyoon Pyo and Yao-Yi Chiang use a generative model to manufacture the labeled examples that a smaller, faster model needs to link mineral site records. Record linkage decides which database rows describe the same deposit, and for mineral sites the rows mix locations with descriptive fields, at scale and in messy formats. Training the efficient discriminative models demands hand-checked ground truth, which is slow and pricey to build. Writing in twenty twenty-four, their recipe has the large model generate labeled linkage data, then fine-tunes the smaller model on the synthetic labels.
The hybrid kept the small model's speed while closing the data gap. The authors report an improvement of over forty-five percent in F1 score over small-model approaches. The pattern, big model as teacher-data factory and small model as worker, is pitched for other linkage problems where gold data are missing.
Catalog record
| CREATOR | Jiyoon Pyo, Yao-Yi Chiang |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-03575