tee-oh-double-dee
commited on
Commit
•
21b0b81
1
Parent(s):
fc77f38
Update README.md
Browse files
README.md
CHANGED
@@ -12,7 +12,7 @@ metrics:
|
|
12 |
- accuracy
|
13 |
pipeline_tag: text-classification
|
14 |
widget:
|
15 |
-
- text: "Speaker 1: This is really terrific work
|
16 |
- text: "Speaker 1: These edits are terrible. Please review my comments above again.</s>Speaker 2: I reviewed your comments, which were not helpful. Roll up your sleeves and do some work."
|
17 |
---
|
18 |
|
@@ -93,9 +93,9 @@ where \\(w_c = \frac{N}{N_c}\\), where \\(N\\) denotes the number of examples in
|
|
93 |
We evaluate accuracy at the individual utterance level and report the following results:
|
94 |
| Split | Accuracy |
|
95 |
|---|---|
|
96 |
-
| Train | % |
|
97 |
-
| Validation | % |
|
98 |
-
| Test | % |
|
99 |
|
100 |
Without loss weighting, it is possible to achieve an accuracy of 45%.
|
101 |
|
|
|
12 |
- accuracy
|
13 |
pipeline_tag: text-classification
|
14 |
widget:
|
15 |
+
- text: "Speaker 1: This is really terrific work!</s>Speaker 2: Thanks!!"
|
16 |
- text: "Speaker 1: These edits are terrible. Please review my comments above again.</s>Speaker 2: I reviewed your comments, which were not helpful. Roll up your sleeves and do some work."
|
17 |
---
|
18 |
|
|
|
93 |
We evaluate accuracy at the individual utterance level and report the following results:
|
94 |
| Split | Accuracy |
|
95 |
|---|---|
|
96 |
+
| Train | 39.21% |
|
97 |
+
| Validation | 35.04% |
|
98 |
+
| Test | 37.25% |
|
99 |
|
100 |
Without loss weighting, it is possible to achieve an accuracy of 45%.
|
101 |
|