public-apps/tag_from_library.livemd

<!-- livebook:{"app_settings":{"auto_shutdown_ms":5000,"multi_session":true,"show_source":true,"slug":"transcriber"}} -->

# Tag Audio

```elixir
Mix.install(
  [
    {:audio_tagger, path: "/Users/noah/development/audio_tagger"},
    {:kino_bumblebee, "~> 0.4.0"},
    {:exla, ">= 0.0.0"},
    {:explorer, "~> 0.7.0"},
    {:kino_explorer, "~> 0.1.11"}
  ],
  config: [
    nx: [default_backend: EXLA.Backend]
    # exla: [
    #   clients: [
    #     cuda: [
    #       platform: :cuda,
    #       lazy_transfers: :never
    #     ]
    #   ]
    # ]
  ]
)
```

## Step 1: Create Vector Embeddings for ICD-10 Codes

```elixir
# Use sentence-transformers/all-MiniLM-L6-v2 to create vectors for each medical code description
tmpfile = Path.join(System.tmp_dir(), "icd10_vector_tensors.bin")

if File.exists?(tmpfile) do
  IO.puts("Found pre-calculated ICD-10 vector embeddings. Skipping embedding.")
else
  AudioTagger.SampleData.icd10_codes()
  |> AudioTagger.Classifier.SemanticSearch.precalculate_label_vectors(tmpfile)

  IO.inspect(tmpfile, label: "Wrote vector embeddings")
end
```

## Step 2: Transcribe Audio Recording

```elixir
# 1 - Prepare model and choose audio file
featurizer = AudioTagger.Transcriber.prepare_featurizer()
audio_input = Kino.Input.audio("Audio", sampling_rate: featurizer.sampling_rate)
```

```elixir
# 2 - Transcribe audio recording to text (using openai/whisper-tiny)
# Takes 5–6s for about a minute of audio
chosen_audio = Kino.Input.read(audio_input)
file = chosen_audio.file_ref |> Kino.Input.file_path() |> File.read!()

transcription_df =
  AudioTagger.Transcriber.transcribe_audio(featurizer, file, chosen_audio.num_channels)

# Show a sample of rows
transcription_df |> Explorer.DataFrame.head(3)
```

## Step 3: Tag Transcribed Audio

```elixir
labels_df = AudioTagger.SampleData.icd10_codes()

tagged_audio =
  transcription_df
  # |> AudioTagger.Classifier.SemanticSearch.tag(labels_df, tmpfile)
  |> AudioTagger.Classifier.SemanticSearch.tag(
    labels_df,
    "/var/folders/8g/1hrq420n22b05m6k98mxqlrr0000gn/T/icd10_vector_tensors.bin"
  )
```