Pashayi Speech to Text: Preserving a Language Through AI

Introduction

Pashayi is a Dardic language spoken by the Pashai people in the remote valleys of eastern Afghanistan and parts of Pakistan. With an estimated 400,000 speakers, it is a minority language that faces pressure from dominant languages like Dari and Pashto. Despite its relatively small speaker population, Pashayi has a rich oral tradition of folk tales, songs, and poetry. Until recently, there were no reliable speech-to-text tools for Pashayi, leaving its speakers without digital access to transcription and subtitling services. Speechyou changes that by offering accurate AI-powered transcription for Pashayi in Latin script.

Why Accurate Pashayi Transcription Matters

Transcription of spoken Pashayi is crucial for several reasons:

• Preservation: Oral histories and traditional knowledge can be documented and archived in a searchable text format.
• Education: Teachers can create subtitled videos for Pashayi-speaking students, improving literacy in the native language.
• Accessibility: Deaf and hard-of-hearing individuals who read Pashayi can access video content through subtitles.
• Media: Community radio and video producers can reach wider audiences by adding subtitles to their programs.

Without a dedicated ASR system, these tasks were nearly impossible or required manual transcription by a handful of native speakers.

Challenges in Transcribing Pashayi

1. Script and Orthography

Pashayi is traditionally written in a modified Arabic script, but there is also a Latin orthography developed by linguists. Speechyou uses the Latin script for output, which is easier for most users to type and edit. The challenge is that the Latin orthography is not standardized across all dialects; our system adopts a consistent representation based on the IPA-based transcription common in linguistic literature.

2. Dialectal Variation

As mentioned, Pashayi has several dialects (Kata, Kurangali, Warduji, etc.) that differ in pronunciation, vocabulary, and even grammar. Our model is trained on data from multiple dialects, and users can select the dialect that best matches their audio for improved accuracy.

3. Tonal and Prosodic Features

Some dialects of Pashayi use pitch to distinguish meanings. For example, the word "kata" can mean 'house' or 'shoulder' depending on the tone. Our ASR model includes a pitch detection module that helps differentiate these minimal pairs.

Use Cases in Detail

• Podcast Subtitling: A Pashayi-language podcast can be automatically transcribed and subtitled, making it accessible to a global audience.
• Oral History Projects: NGOs and universities can transcribe interviews with elders, creating searchable archives of cultural knowledge.
• Local News: Community journalists can produce written news articles from their audio reports, reaching both literate and non-literate audiences.
• Language Learning: Students of Pashayi can use transcriptions to study vocabulary and pronunciation.

How Speechyou Helps

Speechyou is built to handle low-resource languages like Pashayi. Key features include:

• Dialect selection: Choose your dialect for better accuracy.
• Noise reduction: Clean up field recordings for clearer transcription.
• Multiple output formats: Get plain text, SRT, or VTT subtitles.
• Continuous learning: The model improves as you use it.

Conclusion

Pashayi may be a small language, but its speakers deserve modern tools to preserve and promote their linguistic heritage. Speechyou's Pashayi speech-to-text capability is a step forward in bridging the digital divide. Whether you are a researcher, educator, or community member, you can now transcribe Pashayi audio with confidence. Try it today and help keep the Pashayi language alive in the digital world.