Mike Zhang
commited on
Commit
·
34d8ba6
1
Parent(s):
76af73d
Update README.md
Browse files
README.md
CHANGED
|
@@ -4,4 +4,16 @@ language:
|
|
| 4 |
tags:
|
| 5 |
- retrained
|
| 6 |
- SpanBERT
|
| 7 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
tags:
|
| 5 |
- retrained
|
| 6 |
- SpanBERT
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
|
| 10 |
+
SpanBERT
|
| 11 |
+
|
| 12 |
+
This is the SpanBERT model from:
|
| 13 |
+
|
| 14 |
+
Mike Zhang, Kristian Nørgaard Jensen, Sif Dam Sonniks, and Barbara Plank. __SkillSpan: Hard and Soft Skill Extraction from Job Postings__. To appear at the 2022 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL). 2022.
|
| 15 |
+
|
| 16 |
+
This model is pre-trained from scratch on the BookCorpus and WikiData. To pre-train from scratch we use the code from Splinter: https://github.com/oriram/splinter.
|
| 17 |
+
On our job posting dataset, we found that our `spanbert-base-cased` model works better than the original.
|
| 18 |
+
|
| 19 |
+
More information can be found in the paper (which should be released when the NAACL proceedings are online).
|