Voice technology has come a long way, but for smaller language communities, especially dialects it can still be a challenge. This is exactly what the Aseto AI is working to improve. In a recent SigmaLive feature, the Cyprus-based company shared how it developed speech recognition technology designed specifically for the Greek language and the Cypriot dialect, where many international systems still struggle. Aseto shares the journey, challenges, and progress behind building this speech recognition technology.
Why Is Aseto AI’s Work a Big Step for Greek and Cypriot Voice Technology?
According to the article, the idea did not begin as a grand vision, but more along the lines of frustration. Aseto AI found that voice systems which performed well in English could not handle the language patterns used in Greek and Cypriot. Some common issues included numbers, addresses, local pronunciations, and dialect words. George uses the example of the Cypriot way of saying “twenty two,” which completely confuses international models. This showed that if speech is not transcribed correctly from the start, the whole workflow breaks down.
So to solve that, Aseto AI built its own approach using targeted research, carefully prepared datasets, and fine-tuning of open-source models. The company says this work helped achieve speech transcription accuracy above 95 percent, while widely used international systems can drop below 80 percent in real-life conditions for Greek and especially for Cypriot speech. That difference matters because accurate transcription is the foundation for any useful voice system.
One of the most important parts of the project was building the right dataset. Aseto explains that simply collecting voice samples was not enough, but rather, the team had to identify specific weaknesses in existing models and train the system with data that directly addressed those weaknesses. This included local expressions, Cypriot pronunciations, even a full base of Cypriot addresses and location names so the system could understand requests the way people naturally say them.
The article also points to real-world results, with one recent implementation already handling around 120 calls per day. Early results reported by the company include faster service, lower operational pressure, and less burden on customer support teams. Aseto says its technology can answer frequent questions, connect with CRM or ERP systems, provide real-time information, and record requests more efficiently. The company sees this as especially valuable not only for businesses but also for public services, where missed calls and slow responses are common frustrations.
In regards to the future of Aseto AI, its next step is to continue improving its Greek-language solutions and develop a new high-quality Greek text-to-speech system before expanding further into other markets. For Cyprus and Greece to move forward in AI, local innovation is everything. They need stronger local technology, institutional support, and confidence in homegrown solutions. Features like this one in SigmaLive have helped Aseto gain stronger visibility as a local leader in practical AI and hopefully boost conversations toward creating meaningful solutions for the future. To read the full interview, click here.
