Depending on where a company starts out, buyers of language services — and the language service providers (LSPs) themselves — can have different priorities in terms of their expansion efforts.
But as of late, the Middle East and North Africa, also known as the MENA region, has emerged as an up-and-coming market, and LSPs (and buyers) are investing accordingly.
At SlatorCon Zurich 2023, Slator’s Andrew Smart opened a panel on emerging markets and AI in localization by referring to MENA as “an exciting region often overlooked by the industry as a whole.”
In fact, Smart continued, this high-growth, dynamic region is home to some of the fastest-growing world economies, such as Saudi Arabia, Qatar and Dubai. “From a growth perspective, I see the region continuing very solidly,” commented Nour Al Hassan, founder and CEO of Dubai-based LSP Tarjama.
Trendyol, Turkey’s largest e-commerce platform, is also betting on MENA. The company recently launched an Arabic version for Saudi Arabia and the UAE, building on those countries’ existing interest in Turkish products.
Head of Localization at Trendyol, Beatriz Verdasco, explained the company’s thought process behind establishing a platform in Arabic for the Gulf markets, especially Saudi Arabia.
“We […] prioritize the impact to the customer first and then the speed,” Verdasco said. “So sometimes cost is maybe not going to be the lowest, but if we can generate impact for the customer and be able to deliver that impact fast, then […] those are the main priorities.”
Al Hassan agreed that translation and transcreation are obligatory for client-facing content for the Middle East.
“You look at, how do people consume content in this region? They consume it in their own language. So in order for you to localize SKUs, which are the product description for any product technically to buy, you have to have it in Arabic,” she said, adding, “And it needs to be sometimes in a certain dialect for them to be able to buy.”
Al Hassan pointed out that, as a market, the Levant is significantly more price-sensitive than Gulf Coast countries, so companies expanding into MENA may prefer to focus on the latter region.
Of course, the rollout process varies by company, and even more so between private industry and the public sector.
Blanca Pinero-Canovas, Director of Language and Documentation Services at the Geneva, Switzerland-based World Trade Organization (WTO), spoke to the stringent system of checks and balances that can stretch out timelines for public sector buyers.
However, Pinero-Canovas noted that the advent of large language models (LLMs) has reached a “tipping point” such that even general management culture is now open to experimenting. She herself is piloting certain technologies for some projects, “start[ing] very small, and then if a technology is something where I can build a solid business case, take it forward.”
Strategic partnerships with private sector players will also help WTO respond more quickly, she added.
Some initiatives governments have been piloting in this space include Falcon, from Abu Dhabi; JACE, an LLM specifically for Arabic; and another LLM to be launched by the Saudi government “very soon.”
Al Hassan explained that in these cases, the governments will build the infrastructure — a foundational model — and then let startups build solutions for the government.
“Some of these companies contacted us as well for data, for testing, evaluation, or integrating them in our tools and offering it to the market,” she said, referencing three use cases Tarjama has already done for clients.
In response to the hype surrounding LLMs, the WTO, like governments, is trying to send out requests for information to the market, in hopes that respondents will pitch solid business cases for LLMs. The contractual process has to be gradual, Pinero-Canovas said, because “in order to be authorized to launch the RFP, we have to have a ballpark figure which is approved.”
For Arabic-focused LLMs, Al Hassan predicted, data will remain a challenge, considering that Arabic only accounts for three percent of content online. The best data is “behind walls,” rather than online.
“So until the government reaches a number of agreements or acquisitions of these data sets, it’s going to be still weak from that perspective,” Al Hassan said. Similarly, LLMs are still very expensive for people to use on a daily basis.
One challenge for Trendyol has been working around “a lack of sophisticated language models with Turkish as a source language,” Verdasco said.
“It’s not very efficient to go through English all the time, so we might end up having to build our own model that’s consistent with our domain for those language pairs that we have high volumes,” she explained. “But we are at the beginning of the journey there.”
For those who could not attend in person, SlatorCon Zurich 2023 recordings will be available in due course via our Pro and Enterprise plans.