Southern Balochi Speech-to-Text: Preserving Language and Culture with AI

Southern Balochi (بلوچی جنوبی) is a member of the Iranian branch of the Indo-European language family, spoken primarily in the southeastern regions of Iran (Sistan and Baluchestan province), southwestern Pakistan (Balochistan province), and among diaspora communities in the Gulf states and Europe. With an estimated 5 to 8 million speakers, it is a language of rich oral traditions, including epic poetry, folk tales, and a vibrant musical heritage. Despite its significance, written materials in Balochi are relatively scarce, and the use of the Arabic script (with additional characters) adds a layer of complexity to transcription and text processing.

Why Accurate Balochi Speech-to-Text Matters

Accurate speech-to-text for Southern Balochi is crucial for several reasons. First, it enables the preservation of oral history. Many elders possess unique knowledge of traditional medicine, local history, and folklore that exists only in spoken form. Transcribing these recordings creates a permanent record that can be studied and shared. Second, the Balochi-speaking community increasingly uses social media and video platforms to share content. Automatic transcription and subtitling can help these creators reach a wider audience, including those who do not speak Balochi but are interested in the culture. Third, for linguists and researchers, a reliable ASR system provides a tool for analyzing dialectal variation, phonology, and language change over time.

Specific Challenges in Transcribing Balochi

Balochi presents several challenges for automatic speech recognition:

• Phonemic inventory: Southern Balochi has a set of retroflex consonants (ṭ, ḍ, ṇ, ṝ) that are not found in standard Arabic. These sounds are often misrecognized by generic Arabic speech models.
• Vowel system: The language distinguishes between long and short vowels, but the Arabic script typically does not mark short vowels. This leads to ambiguity; for example, the word 'گل' could be read as /gul/ (flower) or /gal/ (throat).
• Dialect diversity: The main dialects (Rakhshani, Makrani, Sarawani, Lashari) differ in pronunciation, intonation, and even some grammatical endings. A model trained on one dialect may perform poorly on another.
• Code-switching: Many Balochi speakers are bilingual in Persian, Urdu, or Arabic, and code-switching is common in everyday speech. The ASR system must handle multiple languages within a single utterance.

How Speechyou Addresses These Issues

Speechyou's Southern Balochi model is built on a corpus of over 500 hours of dialectal speech, carefully annotated by native speakers. The acoustic model is fine-tuned to recognize retroflex consonants and to distinguish vowel length. For the script challenge, we use a two-pass approach: first, the raw phoneme sequence is generated, then a language model predicts the most likely textual representation, including optional diacritics. Users can choose to output with or without harakat. Additionally, we offer dialect-specific profiles that adjust the acoustic and language models to better match the target variety.

Use Cases in Practice

• Podcasters and YouTubers: Balochi-language podcasters can upload their episodes and receive full transcriptions and SRT subtitles within minutes. This makes their content searchable and accessible to hearing-impaired viewers.
• Researchers and anthropologists: Field recordings of oral narratives can be transcribed quickly, allowing for more time to be spent on analysis. The tool's timestamp feature helps in aligning text with audio.
• Educators: Teachers preparing Balochi-language materials can transcribe spoken lessons and create bilingual subtitles for instructional videos.
• Cultural preservation organizations: NGOs working to document Balochi heritage can batch-process large audio archives, generating searchable text indices.

Conclusion

Southern Balochi is a language with a deep cultural heritage, but it faces the risk of digital marginalization. Accurate speech-to-text technology is a bridge between the spoken word and the written record, empowering communities to preserve their language and share it with the world. Speechyou's dedicated support for Southern Balochi (Arabic script) ensures that native speakers, researchers, and content creators have a reliable, easy-to-use tool for transcription and subtitle generation. Whether you are transcribing a single interview or building a large archive, Speechyou provides the accuracy and flexibility needed for this unique and important language.