⬆️How to Ingest Transcripts

Getting data into Junior is the main way of interacting with Junior's output. When you ingest a call into Junior, Junior will create a very high fidelity transcript and then clean it to generate a set of client-ready notes. This also forms the basis of Junior's data architecture, and powers everything else downstream.

Permission to Record

Disclaimer: As part of our Terms and Conditions, Junior does not accept liability for the source of audio recordings on our platform. You must acquire consent to record if that is legally required in your jurisdiction.

A common introduction to the call might be:

"Hello, thanks for taking the time to speak with us. Would you be comfortable if we recorded this conversation for note-taking purposes? We will delete the audio file and your data will be anonymised and aggregated when we conduct analysis. We use note-taking software to allow us to focus on the conversation at hand."

How? aka Input Methods

There are three main methods to get an interview onto the platform:

1. You can upload a transcript of the call into the platform

   • Junior can take in audio files, docx, or PDF files

2. You can send a white labeled recording bot to the call. The bot will automatically generate a transcript at the end of the call and upload it to Junior

3. You can record the call directly in your browser window

We recommend these steps to input data into Junior, listed from highest to lowest quality:

1. White labeled recording bot: this produces the highest quality transcript

2. Audio from the network: we have built a custom transcription tool that generates very high quality transcripts, near human grade.

3. In-browser recording: similar to above, it can generate a very high quality transcript, the audio quality, however, is a little lower than (2).

4. docx & PDF: these are often lower quality transcripts.

5. Typed notes: see here.

Multi-lingual Support

🗺️ Junior is fairly multi-lingual. You don't need to do anything to set this up, just drop in the conversation as described above. Junior can handle multiple languages in a single conversation, so if you're switching from French to Spanish to English in a single call, don't worry! Junior will convert all languages to an English transcript.

Unfortunately, Junior can't speak all languages to the same degree. In order to ensure you are getting the highest quality transcriptions, we suggest you limit to the following languages for now: 🇪🇸Spanish, 🇮🇹Italian, 🇵🇹Portuguese, 🇩🇪German, 🇯🇵Japanese, 🇵🇱Polish, 🇷🇺Russian, 🇳🇱Dutch, 🇮🇩Indonesian, 🇫🇷French, 🇹🇷Turkish, 🇸🇪Swedish, 🇺🇦Ukrainian, 🇲🇾Malay, 🇳🇴Norwegian, 🇫🇮Finnish.

Junior will always yield an English language transcript.

What happens after uploading?

After you drop in an audio file, Junior does a few things:

1. We compress the audio file and convert it to a text transcript.

2. We then extract question and answer pairs from the transcript, clean them for grammatical mistakes, and enhance the questions and answers to make them readable without any extra context.

3. We extract Proper Nouns from every question answer pair.

4. We generate Key Takeaways for the call

5. We match question-answer pairs to any matching interview guide if there is one

6. Extract data points to fill in any relevant quant trackers.

Junior sometimes encounters errors downstream with transcript processing. We will automatically retry generating the transcript from the audio file up to 6 times before failing.

Note: We do our best to notify users when there is an incident happening that might delay processing transcripts.