init repository

Author: xzyaoi

.github/workflows/aid-ci.yml

@@ -0,0 +1,20 @@
+# This workflow will install Python dependencies, run tests and lint with a single version of Python
+# For more information see: https://help.github.com/actions/language-and-framework-guides/using-python-with-github-actions
+
+name: Testing
+
+on:
+  push:
+    branches: [ master ]
+  pull_request:
+    branches: [ master ]
+
+jobs:
+  build:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v2
+    - name: Test AID Model
+      uses: autoai-org/[email protected]

.gitignore

@@ -0,0 +1,4 @@
+docker*
+pretrained/
+runner*
+*.pyc

README.md

@@ -0,0 +1,12 @@
+# eth-library-lab/speech
+
+![AID Testing](https://github.com/eth-library-lab/speech/actions/workflows/aid-ci.yml/badge.svg)
+
+> Speech Analysis Package
+
+This project is based on the AID Project. Common topics are discussed in the [AID docs](https://aid.autoai.org). Make sure to read it!
+
+## Usage
+
+
+## Reference

aid.toml

@@ -0,0 +1,8 @@
+[package]
+name="speech"
+vendor="eth-library-lab"
+tagline="Speech Analysis Package"
+
+[[solvers]] 
+name="speechSolver"
+class="speech/solver/speechSolver"

prepip.sh

@@ -0,0 +1,2 @@
+echo "Installing pip-required dependencies"
+apt-get update && apt-get install sox -y

pretrained.toml

@@ -0,0 +1 @@
+

requirements.txt

@@ -0,0 +1,3 @@
+mlpm
+textblob
+deepspeech

setup.sh

@@ -0,0 +1,2 @@
+echo "Finished Build..."
+python -m textblob.download_corpora

speech/bundle.py

@@ -0,0 +1,64 @@
+from os import fsdecode
+from mlpm.solver import Solver
+from deepspeech import Model
+import wave
+import shlex
+import subprocess
+import numpy as np
+from textblob import TextBlob
+import textblob
+try:
+    from shhlex import quote
+except ImportError:
+    from pipes import quote
+
+def convert_sample_rate(audio_path, desired_sample_rate):
+    sox_cmd = 'sox {} --type raw --bits 16 --channels 1 --rate {} --encoding signed-integer --endian little --compression 0.0 --no-dither - '.format(quote(audio_path), desired_sample_rate)
+    try:
+        output = subprocess.check_output(shlex.split(sox_cmd), stderr=subprocess.PIPE)
+    except subprocess.CalledProcessError as e:
+        raise RuntimeError('SoX returned non-zero status: {}'.format(e.stderr))
+    except OSError as e:
+        raise OSError(e.errno, 'SoX not found, use {}hz files or install it: {}'.format(desired_sample_rate, e.strerror))
+
+    return desired_sample_rate, np.frombuffer(output, np.int16)
+
+class speechSolver(Solver):
+    def __init__(self, toml_file=None):
+        super().__init__(toml_file)
+        # Do you Init Work here
+        self.ds = Model("pretrained/deepspeech-0.9.3-models.pbmm")
+        self.scorepath = ("pretrained/deepspeech-0.9.3-models.scorer")
+        self.ds.enableExternalScorer(self.scorepath)
+        self.desired_sample_rate = self.ds.sampleRate()
+        self.ready()
+    def infer(self, data):
+        # if you need to get file uploaded, get the path from input_file_path in data
+        # First convert audio file to wav format
+        fin = wave.open(data['input_file_path'], 'rb')
+        fs_orig = fin.getframerate()
+        resampled = False
+        if fs_orig != self.desired_sample_rate:
+            resampled = True
+            fs_new, audio = convert_sample_rate(data['input_file_path'], self.desired_sample_rate)
+        else:
+            audio = np.frombuffer(fin.readframes(fin.getnframes()), np.int16)
+        
+        audio_length = fin.getnframes() * (1/fs_orig)
+        fin.close()
+        
+        result = self.ds.stt(audio)
+        textblob_analyzer = TextBlob(result)
+        sentiment = []
+        for sentence in textblob_analyzer.sentences:
+            sentiment.append({
+                'sentence': str(sentence),
+                'polarity': sentence.sentiment.polarity,
+                'subjectivity': sentence.sentiment.subjectivity
+            })
+        return {
+            "transcript": result,
+            "audio_length": audio_length,
+            "resampled": resampled,
+            "sentiment": sentiment
+        }