Spaces:
Running
Running
alessandro trinca tornidor
commited on
Commit
•
6777887
1
Parent(s):
ca7e6be
feat: add score DE and EN components
Browse files- aip_trainer/lambdas/app_description.md +1 -1
- app.py +21 -18
aip_trainer/lambdas/app_description.md
CHANGED
@@ -5,4 +5,4 @@ for more details.
|
|
5 |
|
6 |
Selecting a new example invalidates all previous inputs (Text-to-Speech audio and Speech-to-Text recordings): in that case you should recorder again your audio speeceh.
|
7 |
|
8 |
-
Right now this tool uses {} as sample rate value. From empirical tests the best sample rate value is 48000
|
|
|
5 |
|
6 |
Selecting a new example invalidates all previous inputs (Text-to-Speech audio and Speech-to-Text recordings): in that case you should recorder again your audio speeceh.
|
7 |
|
8 |
+
Right now this tool uses {sample_rate_start} as sample rate value. From empirical tests the best sample rate value is 48000.
|
app.py
CHANGED
@@ -2,7 +2,7 @@ import json
|
|
2 |
from pathlib import Path
|
3 |
import gradio as gr
|
4 |
|
5 |
-
from aip_trainer import PROJECT_ROOT_FOLDER, app_logger
|
6 |
from aip_trainer.lambdas import js, lambdaGetSample, lambdaSpeechToScore, lambdaTTS
|
7 |
|
8 |
|
@@ -20,10 +20,7 @@ with gr.Blocks() as gradio_app:
|
|
20 |
project_root_folder = Path(PROJECT_ROOT_FOLDER)
|
21 |
with open(project_root_folder / "aip_trainer" / "lambdas" / "app_description.md", "r", encoding="utf-8") as app_description_src:
|
22 |
app_description = app_description_src.read()
|
23 |
-
gr.Markdown(app_description)
|
24 |
-
with gr.Row():
|
25 |
-
number_score_de = gr.Number(label="Score DE", value=0.0)
|
26 |
-
number_score_en = gr.Number(label="Score EN", value=0.0)
|
27 |
with gr.Row():
|
28 |
with gr.Column(scale=4, min_width=300):
|
29 |
with gr.Row():
|
@@ -64,18 +61,6 @@ with gr.Blocks() as gradio_app:
|
|
64 |
show_download_button=True,
|
65 |
)
|
66 |
with gr.Column(scale=4, min_width=320):
|
67 |
-
examples_text = gr.Examples(
|
68 |
-
examples=[
|
69 |
-
["Hallo, wie geht es dir?", "de", 1],
|
70 |
-
["Hi there, how are you?", "en", 1],
|
71 |
-
["Die König-Ludwig-Eiche ist ein Naturdenkmal im Staatsbad Brückenau.", "de", 2,],
|
72 |
-
["Rome is home to some of the most beautiful monuments in the world.", "en", 2],
|
73 |
-
["Die König-Ludwig-Eiche ist ein Naturdenkmal im Staatsbad Brückenau, einem Ortsteil des drei Kilometer nordöstlich gelegenen Bad Brückenau im Landkreis Bad Kissingen in Bayern.", "de", 3],
|
74 |
-
["Some machine learning models are designed to understand and generate human-like text based on the input they receive.", "en", 3],
|
75 |
-
],
|
76 |
-
inputs=[learner_transcription, language, difficulty],
|
77 |
-
)
|
78 |
-
|
79 |
transcripted_text = gr.Textbox(
|
80 |
lines=2, placeholder=None, label="Transcripted text", visible=False
|
81 |
)
|
@@ -85,7 +70,13 @@ with gr.Blocks() as gradio_app:
|
|
85 |
label="Letters correctness",
|
86 |
visible=False,
|
87 |
)
|
88 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
recording_ipa = gr.Textbox(
|
90 |
lines=1, placeholder=None, label="Learner phonetic transcription"
|
91 |
)
|
@@ -104,6 +95,18 @@ with gr.Blocks() as gradio_app:
|
|
104 |
)
|
105 |
with gr.Row():
|
106 |
btn = gr.Button(value="Recognize speech accuracy")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
107 |
|
108 |
def get_updated_score_by_language(text: str, audio_rec: str | Path, lang: str, score_de: float, score_en: float):
|
109 |
_transcripted_text, _letter_correctness, _pronunciation_accuracy, _recording_ipa, _ideal_ipa, _res = lambdaSpeechToScore.get_speech_to_score_tuple(text, audio_rec, lang)
|
|
|
2 |
from pathlib import Path
|
3 |
import gradio as gr
|
4 |
|
5 |
+
from aip_trainer import PROJECT_ROOT_FOLDER, app_logger, sample_rate_start
|
6 |
from aip_trainer.lambdas import js, lambdaGetSample, lambdaSpeechToScore, lambdaTTS
|
7 |
|
8 |
|
|
|
20 |
project_root_folder = Path(PROJECT_ROOT_FOLDER)
|
21 |
with open(project_root_folder / "aip_trainer" / "lambdas" / "app_description.md", "r", encoding="utf-8") as app_description_src:
|
22 |
app_description = app_description_src.read()
|
23 |
+
gr.Markdown(app_description.format(sample_rate_start=sample_rate_start))
|
|
|
|
|
|
|
24 |
with gr.Row():
|
25 |
with gr.Column(scale=4, min_width=300):
|
26 |
with gr.Row():
|
|
|
61 |
show_download_button=True,
|
62 |
)
|
63 |
with gr.Column(scale=4, min_width=320):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
64 |
transcripted_text = gr.Textbox(
|
65 |
lines=2, placeholder=None, label="Transcripted text", visible=False
|
66 |
)
|
|
|
70 |
label="Letters correctness",
|
71 |
visible=False,
|
72 |
)
|
73 |
+
with gr.Row():
|
74 |
+
with gr.Column(scale=3, min_width=100):
|
75 |
+
pronunciation_accuracy = gr.Number(label="Current pronunciation accuracy %")
|
76 |
+
with gr.Column(scale=2, min_width=100):
|
77 |
+
number_score_de = gr.Number(label="Score DE", value=0)
|
78 |
+
with gr.Column(scale=2, min_width=100):
|
79 |
+
number_score_en = gr.Number(label="Score EN", value=0)
|
80 |
recording_ipa = gr.Textbox(
|
81 |
lines=1, placeholder=None, label="Learner phonetic transcription"
|
82 |
)
|
|
|
95 |
)
|
96 |
with gr.Row():
|
97 |
btn = gr.Button(value="Recognize speech accuracy")
|
98 |
+
with gr.Accordion("Click here to expand the table examples", open=False):
|
99 |
+
examples_text = gr.Examples(
|
100 |
+
examples=[
|
101 |
+
["Hallo, wie geht es dir?", "de", 1],
|
102 |
+
["Hi there, how are you?", "en", 1],
|
103 |
+
["Die König-Ludwig-Eiche ist ein Naturdenkmal im Staatsbad Brückenau.", "de", 2,],
|
104 |
+
["Rome is home to some of the most beautiful monuments in the world.", "en", 2],
|
105 |
+
["Die König-Ludwig-Eiche ist ein Naturdenkmal im Staatsbad Brückenau, einem Ortsteil des drei Kilometer nordöstlich gelegenen Bad Brückenau im Landkreis Bad Kissingen in Bayern.", "de", 3],
|
106 |
+
["Some machine learning models are designed to understand and generate human-like text based on the input they receive.", "en", 3],
|
107 |
+
],
|
108 |
+
inputs=[learner_transcription, language, difficulty],
|
109 |
+
)
|
110 |
|
111 |
def get_updated_score_by_language(text: str, audio_rec: str | Path, lang: str, score_de: float, score_en: float):
|
112 |
_transcripted_text, _letter_correctness, _pronunciation_accuracy, _recording_ipa, _ideal_ipa, _res = lambdaSpeechToScore.get_speech_to_score_tuple(text, audio_rec, lang)
|