Question 1

Which Saudi dialects does CallScribe handle best — Najdi, Hijazi, or Khaleeji?

Accepted Answer

All three. The Khaleeji acoustic model has Saudi-specific country and region adapters: Najdi (Riyadh and central plateau), Hijazi (Mecca, Medina, Jeddah, Red Sea coast), and East-Province Khaleeji (Dammam, Khobar, Dhahran). You can lock a project to one or let dialect ID auto-route per call. WER on clear telephony audio is 8-12% across all three; mixed-dialect calls (a Riyadh agent serving a Jeddah customer) are common and supported.

Question 2

Does CallScribe support Emirati Khaleeji code-switching with English?

Accepted Answer

Yes — segment-level language identification splits each utterance into Arabic and English regions, with the Khaleeji acoustic model handling the Arabic and a parallel English model handling the English. Mid-utterance switches ("بنعمل reschedule") are decoded across the boundary with a bilingual lexicon. UAE call audio is the most code-switched in our benchmark set; the architecture is built for it.

Question 3

Does CallScribe handle the Qatari dialect specifically, or just generic Gulf Arabic?

Accepted Answer

Qatari is a tagged country variety inside the Khaleeji acoustic model, not an afterthought — the model is fine-tuned on GCC call recordings including Qatari speakers, with a held-out per-country evaluation set so a regression on Qatari audio cannot be hidden by gains elsewhere. WER on clear Qatari telephony audio is in the 8-12% Khaleeji range. Mixed calls (an Egyptian agent serving a Qatari customer) are diarized and dialect-routed per speaker segment.

Question 4

Kuwait has no comprehensive data-protection law — what actually governs call recording and transcription?

Accepted Answer

Three things in practice: CITRA's Data Privacy Protection Regulation (binding telecom and ICT service providers, including cloud), the data-privacy provisions of the Electronic Transactions Law (Law No. 20 of 2014) restricting disclosure of personal data in electronic records, and sector rules — chiefly CBK expectations for banks. CallScribe operates as a processor under a Data Processing Addendum with documented safeguards; for regulated-sector workloads we recommend counsel review of the current CITRA framework, and we support in-region processing where that review lands on data-in-country.

Question 5

Does CallScribe handle both Bahraini dialect communities — ʿArab and Baḥārna?

Accepted Answer

Yes. Bahraini is a tagged country variety inside the Khaleeji model with both communal patterns represented in the training distribution, and a held-out Bahraini evaluation set so regressions cannot hide behind gains elsewhere in the Gulf. Calls mixing both varieties — common on any Manama queue — are diarized and decoded per speaker segment. WER on clear Bahraini telephony audio is in the 8-12% Khaleeji range.

Question 6

Is Omani Arabic really different enough from Khaleeji to need separate handling?

Accepted Answer

Yes — it is the most divergent variety in the GCC. Sedentary Omani speech preserves [q] for MSA qāf where the rest of the Gulf uses [ɡ], retains morphology (including feminine plural verb forms) other Khaleeji varieties have collapsed, and uses high-frequency forms like wēn tsīr that Saudi- or Emirati-tuned models mistranscribe. CallScribe trains on Omani audio and holds out a per-country Omani evaluation set so regressions cannot hide behind Gulf-average gains.

Arabic transcription across the GCC

Saudi Arabia (KSA)

United Arab Emirates

Qatar

Kuwait

Bahrain

Oman