pitehu
/

T5_NER_CONLL_ENTITYREPLACE

Text2Text Generation

text-generation-inference

Inference Endpoints

Model card Files Files and versions Community

Tiancheng Hu commited on Jan 28, 2022

Commit

13fff88

•

1 Parent(s): 4991afa

Update README.md

Files changed (1) hide show

README.md +2 -2

README.md CHANGED Viewed

@@ -30,10 +30,10 @@ There are some outputs (212/3453 or 6.14% that does not have the same length as
 F1 score on testing set of those with matching length :
 | Model | This Model | T5_NER_CONLL_OUTPUTLIST | BERTbase
-| :---: | :---: | :---: |
 | F1| 0.8901 | 0.8691| 0.9240
-*Caveat: The testing set of these aren't the same, due to matching length issue...
 T5_NER_CONLL_OUTPUTLIST only has 27/3453 missing length (only 0.78%); The BERT number is directly from their paper (https://arxiv.org/pdf/1810.04805.pdf)

 F1 score on testing set of those with matching length :
 | Model | This Model | T5_NER_CONLL_OUTPUTLIST | BERTbase
+| :---: | :---: | :---: | :---: |
 | F1| 0.8901 | 0.8691| 0.9240
+**Caveat: The testing set of these aren't the same, due to matching length issue...
 T5_NER_CONLL_OUTPUTLIST only has 27/3453 missing length (only 0.78%); The BERT number is directly from their paper (https://arxiv.org/pdf/1810.04805.pdf)