zolekode
commited on
Commit
•
b31c62a
1
Parent(s):
1b465c9
t5-small wav2vec2 grammar fixer model and tokenizer
Browse files
README.md
ADDED
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# flexudy-pipe-question-generation-v2
|
2 |
+
After transcribing your audio with Wav2Vec2, you might be interested in a post processor.
|
3 |
+
|
4 |
+
I trained it with only 42K paragraphs from the SQUAD dataset. All paragraphs had at most 128 tokens (separated by white spaces)
|
5 |
+
|
6 |
+
```python
|
7 |
+
from transformers import T5Tokenizer, T5ForConditionalGeneration
|
8 |
+
|
9 |
+
model_name = "flexudy/t5-small-wav2vec2-grammar-fixer"
|
10 |
+
|
11 |
+
tokenizer = T5Tokenizer.from_pretrained(model_name)
|
12 |
+
|
13 |
+
model = T5ForConditionalGeneration.from_pretrained(model_name)
|
14 |
+
|
15 |
+
sent = """GOING ALONG SLUSHY COUNTRY ROADS AND SPEAKING TO DAMP AUDIENCES IN DRAUGHTY SCHOOL ROOMS DAY AFTER DAY FOR A FORTNIGHT HE'LL HAVE TO PUT IN AN APPEARANCE AT SOME PLACE OF WORSHIP ON SUNDAY MORNING AND HE CAN COME TO US IMMEDIATELY AFTERWARDS"""
|
16 |
+
|
17 |
+
input_text = "fix: { " + sent + " } </s>"
|
18 |
+
|
19 |
+
input_ids = tokenizer.encode(input_text, return_tensors="pt", max_length=256, truncation=True, add_special_tokens=True)
|
20 |
+
|
21 |
+
outputs = model.generate(
|
22 |
+
input_ids=input_ids,
|
23 |
+
max_length=256,
|
24 |
+
num_beams=4,
|
25 |
+
repetition_penalty=1.0,
|
26 |
+
length_penalty=1.0,
|
27 |
+
early_stopping=True
|
28 |
+
)
|
29 |
+
|
30 |
+
sentence = tokenizer.decode(outputs[0], skip_special_tokens=True, clean_up_tokenization_spaces=True)
|
31 |
+
|
32 |
+
print(f"{sentence}")
|
33 |
+
```
|
34 |
+
|
35 |
+
INPUT 1:
|
36 |
+
```
|
37 |
+
BEFORE HE HAD TIME TO ANSWER A MUCH ENCUMBERED VERA BURST INTO THE ROOM WITH THE QUESTION I SAY CAN I LEAVE THESE HERE IN TWO THOUSAND AND TWO THESE WERE A SMALL BLACK PIG AND A LUSTY SPECIMEN OF BLACK RED GAME COCK
|
38 |
+
```
|
39 |
+
OUTPUT 1:
|
40 |
+
```
|
41 |
+
Before he had time to answer a much-enumbered era burst into the room with the question, I say, "Can I leave these here?" In 2002, these were a small black pig and a dusty specimen of black red game cock.
|
42 |
+
```
|
43 |
+
|
44 |
+
INPUT 2:
|
45 |
+
```
|
46 |
+
GOING ALONG SLUSHY COUNTRY ROADS AND SPEAKING TO DAMP AUDIENCES IN DRAUGHTY SCHOOL ROOMS DAY AFTER DAY FOR A FORTNIGHT HE'LL HAVE TO PUT IN AN APPEARANCE AT SOME PLACE OF WORSHIP ON SUNDAY MORNING AND HE CAN COME TO US IMMEDIATELY AFTERWARDS
|
47 |
+
```
|
48 |
+
|
49 |
+
OUTPUT 2:
|
50 |
+
```
|
51 |
+
Going along Slushy Country Roads and speaking to damp audiences in Droughty School rooms day after day for a fortnight, he'll have to put in an appearance at some place of worship on Sunday morning and he can come to us immediately afterwards.
|
52 |
+
```
|
53 |
+
I strongly recommend improving the performance via further fine-tuning or by training more examples.
|
54 |
+
- Possible Quick Rule based improvements: Align the transcribed version and the generated version. If the similarity of two words (case-insensitive) vary by more than some threshold based on some similarity metric (e.g. Levenshtein), then keep the transcribed word.
|