- complete data collection for two episodes
- make script to extract relevant data from json response in format amenable to manual revision
- add RuPaul commentary as control?
- how specific should the "context" and "episode arc" be? (Must be answered through EDA)
- measure duration of quote? If so, must capture quote end timestamp
- probably should finetune who is speaking to whom and in what context. For example, how to capture difference between Ru making puns about someone versus telling them comments directly?
- "grouping" refers to conversation. Must improve conversation grouping notation
- quotation marks for "comment" / actual quote?
- facial expressions are as good as words -- emojis?
-
record audio and convert to .flac format
-
enable Cloud Speech API in GCP
-
activate Cloud Shell
-
create API key
-
create Speech API request using longrunningrecognize method in request.json:
{ "config": { "encoding":"FLAC", "sampleRateHertz": 16000, //can bump to 44100 but config must match "languageCode": "en-US" }, "audio": { "uri":"gs://{bucket-name}/{file-name}.flac" } }
-
call Speech API:
curl -s -X POST -H "content-Type: application/json" --data-binary @request.json "https://speech.goo gleapis.com/v1/speech:longrunningrecognize?key=${API_KEY}"
-
wait at least 30 seconds, then GET the response by name and pipe to text file
curl -s GET "https://speech.googleapis.com/v1/operations/{operation-name}?key=${API_KEY}" > {file-name}.txt
-
send to storage bucket:
gsutil cp *.txt gs://{bucket-name}