The Pashto language is available in most computer programmes, and there is also a facility of its translation into Urdu, English and vice versa but these translations are not standard and reliable.

“The computer translated ‘Influential community elders’ as ‘Buzurg ke hamil ilaqa mashran’ and ‘Respectable community member’ as ‘Mohtaram community kay member’,” says Abdul Majid Khan, an expert in communication and digital content from Peshawar, who was a part of a United Nations project in the former tribal areas.

Speaking to Lok Sujag, Abdul Majid says that he often needs translations of Urdu and English written material into Pashto, for which he has tried ChatGPT, Google and Gemini. “They provide the basic draft but lack the ability to understand the context and translate difficult or unfamiliar words into easily understandable common language. They are not very helpful in translation tasks”.

Questions over tech and biases against certain languages

Sohail Abid, a technologist and writer, thinks the quality of AI translation cannot be judged as poor from just this aspect of technology. “It is important to understand the reason for its ‘poor performance’. The more a language is used in the digital space, the better the AI proficiency in that language will be. In fact, AI programmes keep on learning constantly; that’s why such languages as are used less in digital space have provided fewer opportunities to these programmes to gain proficiency in them,” he says.
The Pashtun region in Pakistan and Afghanistan has been a battlefield between global powers for the past half-century. The Soviet Union was present here for nearly 10 years while the US military was stationed here for 20 years. These powers conducted research on every aspect of the Pashtun society and produced a plethora of written and audiovisual material but that’s not enough to address issues related to the Pashto language in the digital space.

Pamir Sahil, a journalist and teacher associated with Mashal Radio, believes that all this data was created from a specific angle to fulfill specific needs. No matter how broad these resources are, they can only cater to one aspect of the language and cannot encompass its depth, which is necessary for improving Pashto in digital space.

Weak link between Pashto and digital world 
Shafeeq Gigiani, who has been active in the social and political circles of Peshawar for the past two decades, understands this weak link between Pashto and the digital world very well. He runs an organisation, The Enlight Lab, which is working to bring about social change through participation of the youth and use of technology.

Speaking to Lok Sujag, he laments: “Pashto, like all other languages of Pakistan, faces difficulties due to availability of limited written resources in digital form. In the last two decades, use of Pashto on social media platforms has increased manifold but the material, in literature and other sciences, is in the form of books, which, even if available online, are in scanned file format, that’s unreadable for AI. Similarly, there are many other sources of Pashto that are not usable for computers.”

Ambitious roadmap for Pashto in digital space 

Shafeeq Gigiani has taken an important step to bring all these resources into the reach of computers. He has launched an online platform, called Decentralized Autonomous Organization (DAO), which aims to bring together the experts interested in contributing to development of Pashto language in the digital space.

Shafeeq is inspired by the 16th century Roshani Movement in the Pashtun region and considers his work a continuation of it. The Roshani Movement was founded by Sufi reformer Bayazid Roshan (Bayazid Ansari). In addition to social and religious reforms, one of his notable achievements was determining a script for Pashto, in which he introduced new alphabets for unique sounds of Pashto, in addition to letters of Arabic and Persian, which are still used today with some modifications.
Pashto DAO’s website has also created an ambitious roadmap with a goal for step-to-step development of Pashto in the digital world.

 “Under Pashto DAO, on the International Mother Language Day (Feb 21), we will officially launch a new initiative related to Pashto sounds. This initiative will eventually help AI models like ChatGPT, DeepSeek, Google Gemini to speak Pashto, listen to it and understand it,” says Shafeeq.

He has named this initiative as ‘PashtoPal’.

Gigiani is involved in this project as a low resource language processing expert along with Hanifur Rehman, who resides in the UK. The latter is serving as a senior developer and expert in multilingual natural language processing models.
Incorporating Pashto into AI 

Rehman has been a part of efforts for the past three years to make Pashto an AI language. He says that PashtoPal is a part of Mozilla Common Voice, a global initiative. The objective of this volunteer effort is to collect a database of voice samples from all languages to make them available to developers working on any aspect of language computerisation.

“We are part of efforts to include Pashto in every activity in the field of technology. We completed a project a year-and-a-half ago that contained about 6,000 questions and answers (prompt and completion). We are also working on the Pashto AI model, which is complete and about to be launched. At the same time, we are launching the first project of Pashto voice-to-text transcription, whose flaws, if any, we will be gradually removed.  All these activities are a part of our 2023 plan,” says Hanifur Rehman.

While commending these efforts, Pamir Sahel points out the unfortunate aspect is that these efforts are being made by individuals or small group motivated young people on their own, and they do not have any institutional or governmental support in Pakistan or Afghanistan.

“In KP as well as Balochistan, there are institutions, called Pashto Academy, have been active in research and publishing work for decades but their work is related to history, poetry and literature, and unfortunately, there is hardly any work being done by them on modern challenges of the language, linguistics and its different aspects, such as philology. With this approach (of the institutions) how can Pashto be ready to face modern challenges?” asks Pamir.

Shafeeq Gigiani remains determined and optimistic: “If our efforts under DAO to build digital resources (corpus) are successful, these problems can certainly be overcome. We also plan to hold monthly meetings and online sessions for the promotion of Pashto AI,” he says.