Research

Research on Dramatic Compositions and their Sound Recordings, including preservation and preparation of new Derivative Works

3,930
Audio Quality Analyzed
1,391
AssemblyAI Transcripts
1,469
GPT-4 Treatments
1,385
spaCy NLP Analyses

Audio Analysis Pipeline

Multi-provider computational analysis of audio recordings using state-of-the-art libraries for quality assessment, defect detection, and acoustic fingerprinting.

librosa

Python audio analysis library

  • • Spectral analysis (STFT, Mel spectrograms)
  • • Feature extraction (MFCC, chroma, tempo)
  • • Signal-to-noise ratio estimation
  • • Dynamic range measurement
  • • Harmonic/percussive separation

essentia

Music information retrieval library

  • • Audio quality metrics
  • • Loudness normalization analysis
  • • Spectral complexity metrics
  • • Silence detection and trimming
  • • Onset detection and segmentation

pyaudiorestoration

Defect detection and restoration

  • • Click and pop detection
  • • Crackle and hiss identification
  • • Clipping and distortion analysis
  • • Dropouts and gap detection
  • • Quality scoring and grading

AI-Powered Content Generation

AssemblyAI Transcription

Automated speech-to-text transcription with speaker diarization, producing 1,391 full episode transcripts.

  • • Speaker identification and labeling
  • • Word-level timestamps
  • • Confidence scoring per segment
  • • Entity recognition (names, places)
  • • Punctuation and formatting

GPT-4 Treatment Generation

Creative "Tales from the Crypt"-style episode summaries generated from transcripts, totaling 1,469 treatments.

  • • Horror anthology narrative style
  • • Plot summarization and dramatization
  • • Thematic element extraction
  • • Character relationship mapping
  • • Twist ending preservation

Natural Language Processing

spaCy Analysis Pipeline

Industrial-strength NLP analysis of 1,385 transcripts for linguistic research and semantic understanding.

  • • Named entity recognition (NER)
  • • Part-of-speech tagging
  • • Dependency parsing
  • • Sentence segmentation
  • • Lemmatization and tokenization

Research Applications

Linguistic and cultural insights derived from computational analysis of golden age radio dialogue.

  • • Mid-century American English patterns
  • • Regional dialect preservation
  • • Vocabulary evolution tracking
  • • Character network analysis
  • • Thematic and narrative structure

Database Schema

Analysis Tables

  • • librosa_quality_scores: 3,930 rows
  • • essentia_quality_scores: Audio quality metrics
  • • pyaudiorestoration_defects: Defect detection
  • • transcriptions: 1,391 AssemblyAI outputs
  • • treatments: 1,469 GPT-4 summaries
  • • spacy_analyses: 1,385 NLP results

File Organization

  • • R2 Storage: Cloudflare object storage
  • • public/: Processed database content
  • • private/: Complete archive of recordings
  • • Structure: series/episode hierarchy
  • • Formats: JSON, MP3, SQLite exports
  • • Metadata: Cast, crew, quality scores