Upgrading with description and summarization.

spaces_info.py

@@ -1,10 +0,0 @@
-description = """Gradio Demo for exploring Speech Transcription.
-
-Upload an audio file or record yourself to see a transcription.
-The transcription passes through 3 models: transcription, punctuation, and capitalization. 
-All output is given
-
-Tips:
-- Large files will take a while to process.
-- Live recordings is on the second tab.
-"""

src/app.py

@@ -8,11 +8,27 @@ puntuation_model = PunctuationModel()
 # capitalization_model = ("KES/caribe-capitalise")
 # text = "My name is Clara and I live in Berkeley California Ist das eine Frage Frau Müller"
 # print(result)
-from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, AutoModel
+
+description = """
+# Gradio Demo for exploring Speech Transcription.
+
+Upload an audio file or record yourself to see a transcription.
+The transcription passes through 4 models: transcription, punctuation, capitalization, and summarization. 
+All output is given
+
+Tips:
+- Large files will take a while to process.
+- Live recording is on the second tab.
+"""
+
 
 capitalise_tokenizer = AutoTokenizer.from_pretrained("KES/caribe-capitalise")
 capitalise_model = AutoModelForSeq2SeqLM.from_pretrained("KES/caribe-capitalise")
+spell_tokenizer = AutoTokenizer.from_pretrained("murali1996/bert-base-cased-spell-correction")
+spell_model = AutoModel.from_pretrained("murali1996/bert-base-cased-spell-correction")
 
+summarizer = pipeline("summarization")
 
 pipe = pipeline(
     model="facebook/wav2vec2-large-960h", 
@@ -25,26 +41,53 @@ def translate(audio_file):
     text = x['text']
     return text
 
+
 def punctuation(text):
     punctuation = puntuation_model.restore_punctuation(text)
     return punctuation
 
+
 def capitalise(text):
     text = text.lower()
     inputs = capitalise_tokenizer("text:"+text, truncation=True, return_tensors='pt')
     # print(capitalization)
-    output = capitalise_model.generate(inputs['input_ids'], num_beams=4, max_length=512, early_stopping=True)
+    output = capitalise_model.generate(inputs['input_ids'], num_beams=4, max_length=4096, early_stopping=True)
+    # output = capitalise_model.generate(inputs['input_ids'], num_beams=4, max_length=1024, early_stopping=True)
     capitalised_text = capitalise_tokenizer.batch_decode(output, skip_special_tokens=True)
 
     result = ("".join(capitalised_text))
+    return result
+
+
+def spell_check(text):
+    text = text.lower()
+    inputs = spell_tokenizer(text, return_tensors='pt')
+    # print(capitalization)
+    output = spell_model.generate(inputs)
+    spell_text = spell_tokenizer.batch_decode(output, skip_special_tokens=True)
+
+    result = ("".join(spell_text))
 
     return result
 
+
+def summarize(text):
+    results = None
+    length = len(text)
+    while not results:
+        try:
+            results = summarizer(text[:length], min_length=10, max_length=128)
+        except IndexError:
+            print(f"shortening text: {length} -> {length//2}")
+            length = length // 2
+    return results[0]['summary_text']
+
 def all(file):
     trans_text = translate(file).lower()
     punct_text = punctuation(trans_text)
     cap_text = capitalise(punct_text)
-    return trans_text, punct_text, cap_text
+    sum_text = summarize(punct_text)
+    return trans_text, punct_text, cap_text, sum_text
 
 input = gr.Audio(type="filepath")
 live_in = gr.Audio(type="filepath", source="microphone")
@@ -54,21 +97,22 @@ live_in = gr.Audio(type="filepath", source="microphone")
 raw_output = gr.Text(label="Raw Output")
 puncuation_output = gr.Text(label="Punctuation Output")
 capitalization_output = gr.Text(label="Capitalization Output")
+sum_output = gr.Text(label="Summarized Output")
 
-translater = gr.Interface(
-    fn=translate, 
-    inputs=input, 
-    outputs=[raw_output])
+# translater = gr.Interface(
+#     fn=translate, 
+#     inputs=input, 
+#     outputs=raw_output)
 
-punctuation = gr.Interface(
-    fn=punctuation,
-    inputs=raw_output,
-    outputs=[puncuation_output])
+# punctuation = gr.Interface(
+#     fn=punctuation,
+#     inputs=raw_output,
+#     outputs=puncuation_output)
 
-capitalization = gr.Interface(
-    fn=capitalise,
-    inputs=puncuation_output,
-    outputs=[capitalization_output])
+# capitalization = gr.Interface(
+#     fn=capitalise,
+#     inputs=puncuation_output,
+#     outputs=capitalization_output]
 
 
 
@@ -76,11 +120,17 @@ capitalization = gr.Interface(
 live_demo = gr.Interface(
     fn=all,
     inputs=live_in,
-    outputs=[raw_output, puncuation_output, capitalization_output])
+    outputs=[raw_output, puncuation_output, capitalization_output, sum_output],
+    description=description)
 demo = gr.Interface(
     fn=all,
     inputs=input,
-    outputs=[raw_output, puncuation_output, capitalization_output])
+    outputs=[raw_output, puncuation_output, capitalization_output, sum_output],
+    description=description)
 
-# demo.launch(share=True)
-gr.TabbedInterface([demo, live_demo], tab_names=["Upload File", "Record Self"]).launch()
+# interface = gr.Series(
+#     gr.Textbox(value=description, show_label=False, interactive=False),
+#     gr.TabbedInterface([demo, live_demo], tab_names=["Upload File", "Record Self"])
+# )
+interface = gr.TabbedInterface([demo, live_demo], tab_names=["Upload File", "Record Self"])
+interface.launch()

src/requirements.in

@@ -13,4 +13,6 @@ numpy
 pandas
 icecream
 jupyter
-gradio
+gradio
+deepmultilingualpunctuation
+sentencepiece

src/requirements.txt

@@ -9,9 +9,17 @@
 absl-py==1.2.0
     # via tensorboard
 aiohttp==3.8.1
-    # via fsspec
+    # via
+    #   fsspec
+    #   gradio
 aiosignal==1.2.0
     # via aiohttp
+analytics-python==1.4.0
+    # via gradio
+anyio==3.6.1
+    # via
+    #   httpcore
+    #   starlette
 argon2-cffi==21.3.0
     # via notebook
 argon2-cffi-bindings==21.2.0
@@ -28,6 +36,10 @@ attrs==21.4.0
     #   jsonschema
 backcall==0.2.0
     # via ipython
+backoff==1.10.0
+    # via analytics-python
+bcrypt==3.2.2
+    # via paramiko
 beautifulsoup4==4.11.1
     # via nbconvert
 bleach==5.0.1
@@ -37,23 +49,36 @@ build==0.8.0
 cachetools==5.2.0
     # via google-auth
 certifi==2022.6.15
-    # via requests
+    # via
+    #   httpcore
+    #   httpx
+    #   requests
 cffi==1.15.1
-    # via argon2-cffi-bindings
+    # via
+    #   argon2-cffi-bindings
+    #   bcrypt
+    #   cryptography
+    #   pynacl
 charset-normalizer==2.1.0
     # via
     #   aiohttp
     #   requests
 click==8.1.3
-    # via pip-tools
+    # via
+    #   pip-tools
+    #   uvicorn
 colorama==0.4.5
     # via icecream
+cryptography==37.0.4
+    # via paramiko
 cycler==0.11.0
     # via matplotlib
 debugpy==1.6.2
     # via ipykernel
 decorator==5.1.1
     # via ipython
+deepmultilingualpunctuation==1.0.1
+    # via -r requirements.in
 defusedxml==0.7.1
     # via nbconvert
 entrypoints==0.4
@@ -64,8 +89,12 @@ executing==0.8.3
     # via
     #   icecream
     #   stack-data
+fastapi==0.79.0
+    # via gradio
 fastjsonschema==2.16.1
     # via nbformat
+ffmpy==0.3.0
+    # via gradio
 filelock==3.7.1
     # via
     #   huggingface-hub
@@ -77,22 +106,37 @@ frozenlist==1.3.0
     #   aiohttp
     #   aiosignal
 fsspec[http]==2022.5.0
-    # via pytorch-lightning
+    # via
+    #   gradio
+    #   pytorch-lightning
 google-auth==2.9.1
     # via
     #   google-auth-oauthlib
     #   tensorboard
 google-auth-oauthlib==0.4.6
     # via tensorboard
+gradio==3.1.1
+    # via -r requirements.in
 grpcio==1.47.0
     # via tensorboard
+h11==0.12.0
+    # via
+    #   gradio
+    #   httpcore
+    #   uvicorn
+httpcore==0.15.0
+    # via httpx
+httpx==0.23.0
+    # via gradio
 huggingface-hub==0.8.1
     # via transformers
 icecream==2.1.3
     # via -r requirements.in
 idna==3.3
     # via
+    #   anyio
     #   requests
+    #   rfc3986
     #   yarl
 importlib-metadata==4.12.0
     # via markdown
@@ -121,6 +165,7 @@ jedi==0.18.1
     # via ipython
 jinja2==3.1.2
     # via
+    #   gradio
     #   nbconvert
     #   notebook
 jsonschema==4.7.2
@@ -149,8 +194,14 @@ jupyterlab-widgets==1.1.1
     # via ipywidgets
 kiwisolver==1.4.4
     # via matplotlib
+linkify-it-py==1.0.3
+    # via markdown-it-py
 markdown==3.4.1
     # via tensorboard
+markdown-it-py[linkify,plugins]==2.1.0
+    # via
+    #   gradio
+    #   mdit-py-plugins
 markupsafe==2.1.1
     # via
     #   jinja2
@@ -158,13 +209,20 @@ markupsafe==2.1.1
 matplotlib==3.5.2
     # via
     #   -r requirements.in
+    #   gradio
     #   seaborn
 matplotlib-inline==0.1.3
     # via
     #   ipykernel
     #   ipython
+mdit-py-plugins==0.3.0
+    # via markdown-it-py
+mdurl==0.1.1
+    # via markdown-it-py
 mistune==0.8.4
     # via nbconvert
+monotonic==1.6
+    # via analytics-python
 multidict==6.0.2
     # via
     #   aiohttp
@@ -193,6 +251,7 @@ notebook==6.4.12
 numpy==1.23.1
     # via
     #   -r requirements.in
+    #   gradio
     #   matplotlib
     #   pandas
     #   pytorch-lightning
@@ -204,6 +263,8 @@ numpy==1.23.1
     #   transformers
 oauthlib==3.2.0
     # via requests-oauthlib
+orjson==3.7.8
+    # via gradio
 packaging==21.3
     # via
     #   build
@@ -218,9 +279,12 @@ packaging==21.3
 pandas==1.4.3
     # via
     #   -r requirements.in
+    #   gradio
     #   seaborn
 pandocfilters==1.5.0
     # via nbconvert
+paramiko==2.11.0
+    # via gradio
 parso==0.8.3
     # via jedi
 pep517==0.12.0
@@ -231,6 +295,7 @@ pickleshare==0.7.5
     # via ipython
 pillow==9.2.0
     # via
+    #   gradio
     #   matplotlib
     #   torchvision
 pip-tools==6.8.0
@@ -261,8 +326,16 @@ pyasn1-modules==0.2.8
     # via google-auth
 pycparser==2.21
     # via cffi
+pycryptodome==3.15.0
+    # via gradio
+pydantic==1.9.1
+    # via
+    #   fastapi
+    #   gradio
 pydeprecate==0.3.2
     # via pytorch-lightning
+pydub==0.25.1
+    # via gradio
 pygments==2.12.0
     # via
     #   icecream
@@ -270,6 +343,8 @@ pygments==2.12.0
     #   jupyter-console
     #   nbconvert
     #   qtconsole
+pynacl==1.5.0
+    # via paramiko
 pyparsing==3.0.9
     # via
     #   matplotlib
@@ -278,9 +353,12 @@ pyrsistent==0.18.1
     # via jsonschema
 python-dateutil==2.8.2
     # via
+    #   analytics-python
     #   jupyter-client
     #   matplotlib
     #   pandas
+python-multipart==0.0.5
+    # via gradio
 pytorch-lightning==1.6.5
     # via -r requirements.in
 pytz==2022.1
@@ -304,7 +382,9 @@ regex==2022.7.9
     # via transformers
 requests==2.28.1
     # via
+    #   analytics-python
     #   fsspec
+    #   gradio
     #   huggingface-hub
     #   requests-oauthlib
     #   tensorboard
@@ -312,6 +392,8 @@ requests==2.28.1
     #   transformers
 requests-oauthlib==1.3.1
     # via google-auth-oauthlib
+rfc3986[idna2008]==1.5.0
+    # via httpx
 rsa==4.9
     # via google-auth
 scipy==1.8.1
@@ -322,15 +404,25 @@ send2trash==1.8.0
     # via notebook
 six==1.16.0
     # via
+    #   analytics-python
     #   asttokens
     #   bleach
     #   google-auth
     #   grpcio
+    #   paramiko
     #   python-dateutil
+    #   python-multipart
+sniffio==1.2.0
+    # via
+    #   anyio
+    #   httpcore
+    #   httpx
 soupsieve==2.3.2.post1
     # via beautifulsoup4
 stack-data==0.3.0
     # via ipython
+starlette==0.19.1
+    # via fastapi
 tensorboard==2.9.1
     # via pytorch-lightning
 tensorboard-data-server==0.6.1
@@ -350,6 +442,7 @@ tomli==2.0.1
 torch==1.12.0+cu116
     # via
     #   -r requirements.in
+    #   deepmultilingualpunctuation
     #   pytorch-lightning
     #   torchaudio
     #   torchmetrics
@@ -385,15 +478,23 @@ traitlets==5.3.0
     #   notebook
     #   qtconsole
 transformers==4.20.1
-    # via -r requirements.in
+    # via
+    #   -r requirements.in
+    #   deepmultilingualpunctuation
 typing-extensions==4.3.0
     # via
     #   huggingface-hub
+    #   pydantic
     #   pytorch-lightning
+    #   starlette
     #   torch
     #   torchvision
+uc-micro-py==1.0.1
+    # via linkify-it-py
 urllib3==1.26.10
     # via requests
+uvicorn==0.18.2
+    # via gradio
 wcwidth==0.2.5
     # via prompt-toolkit
 webencodings==0.5.1