SERVICES

Arabic is not one dataset.

Build around the dialect, speakers, linguistic task, and intended model use. Our service catalog is a starting point for a project-specific feasibility review.

PROPOSED PILOTS

Start with the use case.

Volume, timing and cost follow rights, feasibility and reviewer availability checks.

01 / ASR

ASR data & error analysis

Input
Authorized recordings, model output and an agreed transcription reference.
Output
Orthographic transcripts and an error register with explicit normalization rules.
Quality check
Adjudicate reference disagreements; document evaluation units and denominators.
Scope this pilot ↗
02 / TTS

TTS & prosody

Input
Audio samples and their texts, with language variety and context specified.
Output
Agreed annotation tiers and observations on pronunciation, stress and intonation.
Quality check
Listening and acoustic inspection; interpret measurements with their settings.
See the lab example ↗
03 / EVALUATION

Arabic model evaluation

Input
Model responses and a defined task sample without personal or confidential data.
Output
A rubric, reasoned ratings and a register of unresolved cases.
Quality check
Reviewer calibration, adjudication and explicit sampling limitations.
Discuss the rubric ↗

ARABIC PHONOLAB · FROM SPEECH TO ANNOTATION

Public lab example

Hear the speech.
Explore the annotation.

From a short Hijazi recording to the layers that describe speech. Explore the waveform, spectrogram, stress and syllable weight in one workspace.

رامي مر لينا أمس

Text as displayed in the sample’s Arabic tier.

Opens the separate lab in a new tab. A research demonstration, not a client delivery or an accuracy guarantee.

HijaziSample.wavHijazi Arabic

Actual public Hijazi sample in PhonoLab: waveform, spectrogram and annotation tiers. A playhead tracks playback over the image.
1.165 sRecording duration44.1 kHzMono recording8Annotation tiers
Public example captured 17 November 2026. Listen here; open the lab for editing and measurements.
01Listen

Play the recording and inspect a selected interval.

02Inspect tiers

Connect the text with its boundaries and labels.

03Explore measurements

Review analysis settings before interpreting results.

DOWNLOADABLE ILLUSTRATIVE DELIVERY PACK

A useful deliverable includes its context.

JSON and CSV records, a bilingual data dictionary, structural checks report, and SHA-256 manifest. This is a synthetic delivery-structure example; it contains no lab recording or lab-derived measurements.

Download example pack ZIP ↓ Read annotation conventions ↗

Questions before a project

Which delivery formats?

Agree TextGrid, JSON or CSV according to required tiers and relationships. Formats are not equivalent; client-tool import checks belong in the agreed scope.

Who controls data rights?

Consent, licensing and permitted uses are reviewed before transfer. The contract defines rights; a public example does not grant redistribution rights to participant recordings.

Which dialects can be supported?

Specify variety, social context, speaker requirements and reviewer availability for each project. We do not claim operational coverage or ready inventory for all dialects.

How is quality accepted?

Agree sampling, rubrics, thresholds and adjudication in advance. Structural checks are separate from linguistic correctness; no accuracy rate is claimed without documented measurement.

How is work priced?

Recording effort, accepted audio duration, annotation labor hours, records, consulting and QA are separate units. An audio hour is not a labor hour. Prices are quoted after scope review.

WHAT WE CAN BUILD TOGETHER

From speech collection to model evaluation.

These are proposed commercial offerings. Availability, staffing, timelines, and deliverable volumes are confirmed only after assessment.

Project-scoped offering
01

Speech & corpus development

Dialect-focused recording design, participant criteria, metadata, and collection protocols. Recruitment and recording capacity are established for each approved project.

Collection · Corpus design

Explore the annotation example ↗

Project-scoped offering
02

Transcription & alignment

Orthographic and phonetic transcription, word and phone alignment, and syllable segmentation with explicit conventions and expert review.

Words · Phones · Syllables

Explore the annotation example ↗

Project-scoped offering
03

Prosody & linguistic annotation

Stress, intonation, semantic and pragmatic annotation designed around a specified research question or AI use case.

Stress · Intonation · Meaning

Explore the annotation example ↗

Project-scoped offering
04

Arabic dialect analysis

Dialect identification and classification with regional and speaker metadata, transparent criteria, and uncertainty labels.

Saudi · Gulf · Arabic varieties

Explore the annotation example ↗

Project-scoped offering
05

AI speech & language evaluation

ASR error analysis, TTS listening protocols, and linguistic evaluation of Arabic language models with task-specific rubrics.

ASR · TTS · Arabic LLMs

Explore the annotation example ↗

Project-scoped offering
06

Dataset QA & consulting

Specification reviews, sampling plans, annotation adjudication, metadata checks, and expert linguistic consulting.

Quality · Documentation · Advice

Explore the annotation example ↗

From method to your first pilot

Professional annotation guide

Transcription conventions, acoustic evidence, quality checks, and an illustrative data structure.

Read the guide ↗

For AI teams

Define a pilot, its deliverables and acceptance criteria before a commercial commitment.

Plan your pilot ↗

Rights & project review

How data rights, privacy and commercial terms are assessed before project acceptance.

View the review process ↗

Explicit scope, measurable acceptance

We define the unit of work, permitted uses, speaker criteria, transcription conventions, sampling method, and error thresholds before delivery. Phone boundaries and prosodic labels require appropriate evidence and review.

Formats that fit your workflow

Depending on the approved specification, deliverables may include WAV audio, TextGrid tiers, CSV or JSON annotations, a data dictionary, provenance records, and a quality report. No proprietary client data is included in demonstrations.

A WELL-DEFINED FIRST STEP

Plan a focused pilot

Share your dialect, task, and acceptance criteria. We will assess the linguistic requirements and practical delivery scope before proposing a commitment.

Discuss your project