Arabic Dialect Speech-to-Text

Hamsa's models are built to handle spoken Arabic dialects rather than only the formal written standard, which matters for call centers and voice apps serving Gulf, Levantine, or Egyptian speakers. A Beam agent can read a completed transcript from a customer call once Hamsa has processed the audio, then apply the approved tagging rule to label the call by topic, the same categories a human agent would choose from. It updates the ticket with that tag. Where the confidence score is low, or the dialect is not one the account has configured for, the agent routes the audio to a human instead.

Arabic Voice Generation

Hamsa also produces spoken Arabic audio in a chosen dialect, which teams use for voice prompts, IVR messages, or read-back confirmations. A Beam agent can read an approved script, request the audio in the dialect the account has set as default, and attach the resulting file to the relevant record, such as an order confirmation or support ticket. It applies only scripts already approved for use; it does not write new wording. If a script has a placeholder that was not filled in, or the requested dialect is not supported, the agent stops and flags the request for a person to fix first.

Dialect-Specific Intent Tagging

Because Arabic dialects can differ enough that a phrase in one region reads differently in another, tagging customer intent accurately depends on the model actually understanding the dialect spoken, not just standard Arabic. A Beam agent reading Hamsa's transcript and its detected dialect can apply the account's intent categories, such as billing question or delivery issue, and update the linked ticket with that label. It notifies the queue owner once tagged. When the detected dialect is uncertain or the phrasing does not clearly match an approved category, the agent leaves the ticket untagged and routes it to a human agent fluent in that dialect.