Getting started with Yandex SpeechSense

Yandex SpeechSense allows you to analyze your business communication channels based on dialog audio recordings or text chat messages and supports integration with your PBX and CRM systems. SpeechSense utilizes Yandex SpeechKit voice technologies to transcribe speech and perform statistical and qualitative analysis of recorded audio dialogs.

SpeechSense can analyze audio and text dialogs in Russian.

You can upload your data to SpeechSense or use a demo recording synthesized using SpeechKit.

Getting startedGetting started

1. Go to the management console and log in to Yandex Cloud or sign up if not signed up yet. For information on how to get started with Yandex Cloud, see Getting started with Yandex Cloud.
2. Accept the user agreement.
3. In Yandex Cloud Billing, make sure you have a billing account linked and its status is ACTIVE or TRIAL_ACTIVE. If you do not have a billing account yet, create one.
4. Make sure that your account has the speech-sense.spaces.creator role assigned.
5. Open the SpeechSense home page.
6. Select the organization to work with SpeechSense in or create a new one.

Configure the environmentConfigure the environment

1. Create a space where all your projects will be stored: select Create space, enter a name, add a description (optional), and click Create.

2. Link a billing account to the space. This account will be debited for the use of SpeechSense.

   Tip

   You can only manage a billing account if you have a Yandex account. If using Yandex Cloud through an identity federation, contact support.

3. Go to the Connections tab and create a connection based on metainformation of your audio recordings or text chats:

   1. Enter the Connection name.
   2. Select the Two-channel audio or Chat data type.
   3. Set the connection parameters. For more information, see Creating a connection.
   4. Click Create connection.
   Example of metadata.json for loading audio data

   {
      "direction_outgoung": "true",
      "client_id": "456",
      "client_name": "John Doe",
      "date": "2023-09-29T09:08:38.958Z",
      "date_to": "2023-09-29T09:15:07.897Z",
      "language": "RU",
      "operator_id": "123",
      "operator_name": "Jane Doe"
   }
   

4. Create a project: on the space page, click Create project, enter a project name, and add a connection to the project. For each connection, you can select filtering rules based on its metainformation and the YandexGPT API criteria. With filters, you can ensure that only the dialogs you need are added to the project. Once you are done adding connections and configuring filters, click Create project.

Upload your dataUpload your data

SpeechSense uses the gRPC API to upload data.

To upload data:

1. Create a service account.

2. Add the service account to the namespace with the speech-sense.data.editor role. To learn more about the roles available in the service, see Access management in SpeechSense.

3. Create an API key or IAM token for the service account to authenticate with the API. Learn more about authentication in the SpeechSense API.

4. Upload your audio data or chat history data using a Python data transfer script.

   Use one message for the whole audio.

   SpeechSense supports the following audio file formats:

   • LPCM: AUDIO_ENCODING_LINEAR16_PCM
   • WAV: CONTAINER_AUDIO_TYPE_WAV
   • OggOpus: CONTAINER_AUDIO_TYPE_OGG_OPUS
   • MP3: CONTAINER_AUDIO_TYPE_MP3