Businesses are expanding their investments in natural language processing (NLP) technology, whether to power translation or document summarization. According to a poll conducted by John Snow Labs and Gradient Flow in 2021, 60 percent of IT executives claimed their NLP budgets increased by at least 10% from 2020, with a third saying spending increased by more than 30%.


It's a market that's really competitive. There's a new generation of vendors offering NLP services on top of open source AI models, including well-funded startups like OpenAI, Cohere, AI21 Labs, and Hugging Face, as well as IT heavyweights like Google, Microsoft, and Amazon. Yochai Levi, on the other hand, remains undeterred. He's a co-founder of One AI, an NLP platform that raised $8 million from Ariel Maislos, Tech Aviv, Sentinel One CEO Tomer Wiengarten, and other undisclosed venture companies and angel investors today.


"Advanced NLP is still employed mostly by specialist academics, Big Tech, and governments, despite the market's rapid growth," Levi told TechCrunch via email. "We feel the technology is reaching maturity, and after constructing NLP from the ground up numerous times, we thought it was time to productize it and make it available to all developers."


Levi lays out what he believes square measure the most important challenges plaguing information processing development. It’s typically tough to rector open supply models, he argues, as a result of they need to be matched each to the correct domain and task. for instance, a text-generating model trained to classify medical records would be a poor appropriate associate degree app designed to form advertisements. Moreover, models got to be perpetually retrained with new knowledge — lest they become “stale.” Case in purpose, OpenAI’s GPT-3 responds to the question “Who’s the president of the U.S.?” with the solution “Donald Trump” as a result of it absolutely was trained on knowledge from before the 2020 election.


Levi believes the answer could be a package of information processing models trained for specific business use cases — in alternative words, One AI’s product. He teamed up with corporate executive Amit mountain, CPO Aviv Dror and CSO Asi Sheffer in 2021 to pursue the thought. mountain antecedently was the pinnacle of AI at LogMeIn when the corporate noninheritable  his second startup, Nanorep, associate degree AI and chatbot merchandiser. Dror helped to co-found Nanorep and served as a platform product manager at Wix. Sheffer, a former knowledge mortal at Nanorep, was the principal knowledge mortal at LogMeIn. As for Apostle, he was the VP of on-line selling at LivePerson and therefore the head of promoting at WeWork.


One AI offers a group of models which will be mixed and matched in an exceedingly pipeline to method text via one API decision. every model is chosen and trained for its pertinency to the enterprise, Levi said, and mechanically matched by the platform to a customer’s task and domain (e.g., spoken communication account, sales insights, topic detection and proofreading). One AI’s models also can be combined with open supply and proprietary models to increase One AI’s capabilities.



The platform’s API accepts text, voice and video inputs of assorted formats. With One AI’s language studio, users will experiment with the genus Apis and generate calls to use in code.



“With the maturation of Language AI technologies, it's finally time for machines to start out adapting to USA,” mountain told TechCrunch via email. “The adoption of language comprehension tools by the broader developer community is that the thanks to get there.”


One AI costs its information processing service across many tiers, as well as a free tier that has process for up to 1 million words a month. The stackable “growth tier” adds one hundred,000 words for $1.


The hurdle One AI can got to overcome is convincing customers that its services square measure additional enticing than what’s already out there. In April, OpenAI aforementioned that tens of thousands of developers were victimization GPT-3 via its API to get words for over three hundred apps. But, with Fortune Business Insights pegging the information processing market at $16.53 billion in 2020, it can be argued that there’s an oversized enough slice of the pie for newcomers.


Added TehAviv founder and managing partner Yaron Samid: “Language is that the Most worthy untapped resource, on the far side the reach of most merchandise and firms. this can be faithful most industries and domains and ends up in negative outcomes that embody everything from lost sales to lower levels of user engagement and loyalty to name harm. Unleashing Language AI permits USA to harness the facility of this unstructured knowledge and switch it into helpful info and insights.”


One AI says that some of the seed spherical take are place toward increasing its 22-person team, which has ten information processing knowledge scientists.