Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -6,26 +6,30 @@ size_categories:
|
|
6 |
- 10M<n<100M
|
7 |
---
|
8 |
|
9 |
-
|
10 |
-
4.3 million unique words/phrases (English Wikipedia titles or their parts) occurring in 33.8 million English Wikipedia paragraphs.
|
11 |
|
12 |
-
|
13 |
-
26 million phrase pairs in the corrupted phrase inventory, as recognized by different ASR models
|
14 |
|
15 |
-
|
16 |
-
|
17 |
|
18 |
-
|
19 |
-
|
20 |
|
21 |
-
|
22 |
-
|
23 |
|
24 |
-
|
25 |
-
|
26 |
-
gives pairs of initial phrase and its recognition result.
|
27 |
-
Does not include .wav files, but these can be reproduced by feeding g2p to tts
|
28 |
|
29 |
-
|
30 |
-
|
31 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
- 10M<n<100M
|
7 |
---
|
8 |
|
9 |
+
This is the dataset presented in my [ASRU-2023 paper](https://arxiv.org/abs/2309.17267).
|
|
|
10 |
|
11 |
+
It consists of multiple files:
|
|
|
12 |
|
13 |
+
Keys2Paragraphs.txt (internal name in scripts: yago_wiki.txt):
|
14 |
+
4.3 million unique words/phrases (English Wikipedia titles or their parts) occurring in 33.8 million English Wikipedia paragraphs.
|
15 |
|
16 |
+
Keys2Corruptions.txt (internal name in scripts: sub_misspells.txt):
|
17 |
+
26 million phrase pairs in the corrupted phrase inventory, as recognized by different ASR models
|
18 |
|
19 |
+
Keys2Related.txt (internal name in scripts: related_phrases.txt):
|
20 |
+
62.7 million phrase pairs in the related phrase inventory
|
21 |
|
22 |
+
FalsePositives.txt (internal name in scripts: false_positives.txt):
|
23 |
+
449 thousand phrase pairs in the false positive phrase inventory
|
|
|
|
|
24 |
|
25 |
+
NgramMappings (internal name in scripts: replacement_vocab_filt.txt):
|
26 |
+
5.5 million character n-gram mappings dictionary
|
27 |
+
|
28 |
+
asr
|
29 |
+
outputs of g2p+tts+asr using 4 different ASR systems (conformer ctc was used twice),
|
30 |
+
gives pairs of initial phrase and its recognition result.
|
31 |
+
Does not include .wav files, but these can be reproduced by feeding g2p to tts
|
32 |
+
|
33 |
+
giza
|
34 |
+
raw outputs of GIZA++ alignments for each corpus,
|
35 |
+
from these we get NgramMappings.txt and Keys2Corruptions.txt
|