San Francisco, California (PRWEB) January 05, 2017
Until now the market of AI has been driven by the technological industry and has been observed with growing interest by conservative industries like banking, healthcare, and retail. But these newcomers in the AI market are customer-centric industries, and they are finding that the existing AI technologies are not able to effectively understand the needs expressed by their clients. The Bitext Training Framework for deep learning is aimed to improve the text understanding capacity of AI systems.
The machine learning and deep learning algorithms that power the AI projects have been around for years and have suffered little changes. If the technology has been developed, why is it not possible to freely communicate with a machine? The problem is in the data − there is still a missing link.
On the data side, the old approach was to use just plain raw data, the data hand-tagged by humans and then used as input for the machine learning and deep learning projects. Thus large amounts of data, manual labor, time and processing power were needed to obtain relevant results, and these results were far from perfect. The answer provided by the old approach was simple − add more data and more resources, with the hope that the system will improve. The text analysis industry is the perfect example where the use of the old methods failed to achieve a human-like understanding of language, and never will. Relevant information needed by the machine learning and deep learning algorithms to be able to accurately understand all the complexity of language is not being added to the input texts.
To take the interaction between humans and AI to the next level, Bitext has developed a new Training Framework for projects using machine learning and deep learning algorithms. The new Deep Learning Trainer can transform any kind of unstructured text into a high-quality input that has been automatically annotated and disambiguated.
Bitext leverages all the linguistic knowledge available for a given language to enrich the input texts with lexical, morphological, grammatical, and semantical information. All these features are used by humans at an unconscious level to better understand each other. To mimic humans, the AI systems need to understand the rules that govern communication.
And as people do not use the same language for every aspect of their lives, the Bitext Trainer takes into account the context of every input text and enriches it accordingly. It does not matter what the original source is, as the Bitext Trainer can work with both high and low-quality sources like newspapers, social media comments, or research papers.
The structured and enriched data provided by the Bitext Framework speeds up the training process for machine learning and deep learning projects, saves costs as the better quality of the data means that you need less of it to provide relevant results, and above all, makes an AI distinctly better at understanding natural language as the AI will take into account key information previously unavailable to it.
At Bitext, the belief is that human-like quality when understanding language is key to seeing a general adoption of AI-powered technologies.That will only be possible if the machines are provided with the right data.
Further information on Bitext products, technology, and current clients can be found at bitext.com.
Bitext develops multilingual analytics technology in 30 languages. The company takes a linguistic approach to text analysis, leveraging the knowledge of computational linguists to build internally all the software behind its platform. These solutions can be leveraged through a REST API, the Bitext CX SaaS platform, or licensed On-premise.