Intelligent Voice publishes a range of connectors for call recorders, online meeting services, and chat archives. These are designed to take a wide range of disparate source data, standardise it and queue it for ingestion into Intelligent Voice.
Data flow
Processes
1.0 Publish or export
The call recorder, online meeting service or chat archive makes the data available. This is usually a separate step after capturing the recording, and may require a different application, service, or license.
2.0 IV Connector decrypt, filter, convert, normalise
Depending on the source, the IV Connector may have the following options:
-
Decrypting exported data using a provided key
-
Filtering input data using preset rules
-
Lossless compression and identification of silent records
-
Converting media formats
-
Normalising metadata such as timestamps or participant identifiers
Not all features are available in all connectors - consult the documentation for each connector or contact IV support for more details.
The output from this process is standardised in the Call Recorder Connector Interface Technical Specification or Chat Connector Interface Technical Specification.
3.0 Match custodians, apply rules, import
This process is implemented by the RecordingImportWorker or ChatImportWorker supplied with IV. It accepts the standardised output from the IV Connector, and optionally matches it to custodian metadata in a range of formats. It can then apply different rules or processing instructions depending on the metadata or the matched custodian, such as using different models depending on the languages spoken by the custodian, or requesting speaker separation.
4.0 Speech detection, transcription, Topic extraction, optional analytics
Intelligent Voice will process audio or video data, detecting speech, transcribing speech and extracting topics, and optionally running other analytics such as speaker separation (diarization), biometric speaker identification, summarisation, sentiment analysis and more.
Data stores
X1 Export location
The contains data exported from the call recorder, online meeting service, chat archive etc. The method of access depends on the source, and could involve the connector fetching data from file share, API or SFTP server, or listening for API requests.
Q1 Import queue
This contains the jobs for ingestion into IV. This could be actual queuing system (supported queues include Gearman, Azure Service Bus, Azure Queue Storage and AWS SQS) or a filesystem.
F1 File storage
This contains the audio, video or chat transcript, and the metadata file.
D1 Intelligent Voice Database
This is where the processing instructions and the outputs are stored.