PhilipMay commited on
Commit
358841c
1 Parent(s): 2d03f2b

deascribe Processing Steps

Browse files
Files changed (1) hide show
  1. README.md +16 -0
README.md CHANGED
@@ -11,6 +11,22 @@ size_categories:
11
  - n<1K
12
  ---
13
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
  ## Copyright
15
  Copyright (c) the authors of [xliuhw/NLU-Evaluation-Data](https://github.com/xliuhw/NLU-Evaluation-Data)\
16
  Copyright (c) 2022 [Philip May](https://may.la/), [Deutsche Telekom AG](https://www.telekom.com/)
 
11
  - n<1K
12
  ---
13
 
14
+ # NLU Few-shot Benchmark - English and German
15
+
16
+ ## Processing Steps
17
+ - drop `NaN` values
18
+ - drop duplicates in `answer_de`
19
+ - add `char_count` column
20
+ - delete all rows with `char_count` > 70
21
+ - add column `label`: `df["label"] = df["scenario"] + "_" + df["intent"]`
22
+ - remove classes (`label`) with less than 25 samples:
23
+ - `audio_volume_other`
24
+ - `cooking_query`
25
+ - `general_greet`
26
+ - `music_dislikeness`
27
+ - random selection for train set - exactly 20 samples for each class (`label`)
28
+ - rest for test set
29
+
30
  ## Copyright
31
  Copyright (c) the authors of [xliuhw/NLU-Evaluation-Data](https://github.com/xliuhw/NLU-Evaluation-Data)\
32
  Copyright (c) 2022 [Philip May](https://may.la/), [Deutsche Telekom AG](https://www.telekom.com/)