Here is a new company-specific speech model of AI: Aiola jargonic claims that the best rivals in your business lingo

Speech recognition models have become more accurate in recent years. However, they can be built and compared under ideal conditions-Quetal rooms, clear audio and general-purpose dictionary. For businesses, however, the audio in the real world is far softer.

This is the challenge Ayola aims to deal with LaunchHis new automatic speech recognition (ASR), established specifically for the use of the enterprise. The Israeli startup reveals jargons today.

Jargonic is a new model of speech to text designed to deal with specialized jargon, background noise and various accents without extensive retraining or fine setting.

“Our model focuses on three key challenges in recognizing speech: jargon, background noise and accents,” says Jill Hetz, Vice President of AIOLA of AI. “We have built a model that understands a specific industrial jargon in zero, processes noisy environments and maintains a wide range of accents.”

It is now available via API of Enterprise Platform of Aiela, Jargonic is positioned as a production of ASR production solution for enterprises in industries such as production, logistics, financial services and health care.

From first to ai-first

Starting Jargonic is a focus change for Iyola itself. According to the company management, the team redefines its approach to prioritizing AI research and implementation.

“When I arrived here, I saw an incredible product company that had invested strongly in advanced AI capabilities, but was most known for helping people fill out forms,” ​​says Asaf Asbag, chief technology and product of Aiola. “We have shifted the prospect and became an AI company with a great product instead of a product company with AI opportunities.”

“We decided to open our opportunities for the world,” Asbag added. “Instead of serving our model only for businesses in our product, we have developed API and now we start it to make our model of corporate class, Bulletproof the model of all.”

Recognition of jargon, zero adaptation

One of the hallmarks of Jargonic is his approach to the specialized dictionary. Speech recognition systems usually fight when confronted with a domain -specific jargon that does not appear in standard training data. The jargon deals with this challenge with its own keyword notice system that allows zero impact adaptation-can simply provide a list of conditions without further retraining.

In comparison tests, jargon demonstrates a 5.91% average word error (Wer) In four leading English academic data sets, competitors such as eleven laboratories, the AI ​​Assembly, Whisper of Openai and Deepgram Nova-3.

However, the company has not yet revealed comparisons of productivity specifically against larger multimodal transcription models such as Openai’s GPT-4o-Transcribe, Which came nine days ago, boasting the top performance of indicators such as Wer, only 2.46% in English. Aiola claims that his model is still better in choosing a specific business jargon.

The jargon also achieved 89.3% Call percentage under specialized financial conditions and consistently superior to others in multilingual recognition of jargon, reaching 95% accuracy in five languages.

“Once you have severe jargon, the recognition accuracy usually drops by 20%,” Asbag explained. “But with our zero impact approach, where you just list important keywords, accuracy jumps back to 95%. It’s unique to us.”

This ability is designed to remove the retraining process, which is usually required to adapt ASR systems for specific industries.

Optimized for the corporate environment

Jargonic’s development was informed by years of experience in building solutions for corporate clients. The model was trained for over one million hours of transcribed speech, including significant data from industrial and business environments, providing stable in noisy real life.

“What sets us apart is that we have spent years solving problems with real -world businesses,” Hetz said. “We were optimized for speed, accuracy and the ability to deal with complex environments-not just podcasts or videos, but noisy, messy, real-life jobs.”

The architecture of the model integrates to notice keywords directly in the transcription process, which allows jargons to maintain accuracy even under unpredictable audio conditions.

The first future of the voice

For the management of Aiola Jargonic, it is a step towards a broader change in how people interact with technology. The company sees the recognition of speech not only as a business instrument, but also as an essential interface for the future of human -computer interaction.

“Our vision is that every machine interface will soon be the first vowel,” Hetz said. “You will be able to talk to your fridge, your vacuum cleaner, every machine – and it will act and do whatever you want. This is the future we build.”

Asbag sounds this mood, adding: “The AI ​​conversation will become the new web browser. The machines are beginning to understand us and now we have a reason to interact with them naturally.”

So far, Aiola’s focus remains on the enterprise. Jargonic is available immediately for corporate clients through API, which allows them to integrate the ability to recognize the model speech into their own workflows, applications or services aimed at customers.