sven-nm commited on
Commit
71e2845
·
1 Parent(s): 7af5112

Update readme

Browse files
Files changed (1) hide show
  1. README.md +9 -2
README.md CHANGED
@@ -1,5 +1,5 @@
1
  ### Model and entities
2
- `roberta_classics_ner` is a domain-specific RoBERTa-base for named entity recognition in the domain of classical studies. The entities it can recognise are the following:
3
 
4
 
5
  | id | label | desciption | Example |
@@ -19,6 +19,7 @@
19
 
20
  ### Example
21
 
 
22
  ```
23
  B-AAUTHOR B-AWORK B-REFSCOPE
24
  Homer 's Iliad opens with an invocation to the muse ( Il. I. 1).
@@ -28,7 +29,8 @@ Homer 's Iliad opens with an invocation to the muse ( Il. I. 1).
28
 
29
  ### Dataset
30
 
31
- `roberta_classics_ner` was fine-tuned and evaluated on `EpiBau`, a dataset which has not been released publicly yet.
 
32
  Entity counts of the `Epibau` dataset are the following:
33
 
34
  | | train-set | dev-set | test-set |
@@ -43,8 +45,10 @@ Entity counts of the `Epibau` dataset are the following:
43
 
44
 
45
  ### Results
 
46
  Trained and tested on `EpiBau` with a 85-15 split, the model yields a general F1 score of **.82** (micro-averages). Detailed scores are displayed below. Evaluation was performed with the [CLEF-HIPE-scorer](https://github.com/impresso/CLEF-HIPE-2020-scorer), in strict mode)
47
 
 
48
  | metric | AAUTHOR | AWORK | REFSCOPE | REFAUWORK |
49
  | --------- | ------- | ----- | -------- | --------- |
50
  | F1 | .819 | .796 | .863 | .756 |
@@ -53,3 +57,6 @@ Trained and tested on `EpiBau` with a 85-15 split, the model yields a general F1
53
 
54
 
55
 
 
 
 
 
1
  ### Model and entities
2
+ `roberta_classics_ner` is a domain-specific RoBERTa-based model for named entity recognition in Classical Studies. It recognises bibliographical entities, such as:
3
 
4
 
5
  | id | label | desciption | Example |
 
19
 
20
  ### Example
21
 
22
+
23
  ```
24
  B-AAUTHOR B-AWORK B-REFSCOPE
25
  Homer 's Iliad opens with an invocation to the muse ( Il. I. 1).
 
29
 
30
  ### Dataset
31
 
32
+
33
+ `roberta_classics_ner` was fine-tuned and evaluated on `EpiBau`, a dataset which has not been released publicly yet. It is composed of four volumes of **Structures of Epic Poetry**, a compendium on the narrative patterns and structural elements in ancient epic.
34
  Entity counts of the `Epibau` dataset are the following:
35
 
36
  | | train-set | dev-set | test-set |
 
45
 
46
 
47
  ### Results
48
+
49
  Trained and tested on `EpiBau` with a 85-15 split, the model yields a general F1 score of **.82** (micro-averages). Detailed scores are displayed below. Evaluation was performed with the [CLEF-HIPE-scorer](https://github.com/impresso/CLEF-HIPE-2020-scorer), in strict mode)
50
 
51
+
52
  | metric | AAUTHOR | AWORK | REFSCOPE | REFAUWORK |
53
  | --------- | ------- | ----- | -------- | --------- |
54
  | F1 | .819 | .796 | .863 | .756 |
 
57
 
58
 
59
 
60
+
61
+ Questions, remarks, help or contribution ? Get in touch [here](https://github.com/AjaxMultiCommentary), we'll be happy to chat !
62
+