Concept: Wyoming’s startup Language I/O has developed AI-powered software for real-time, company-specific language translations. It claims to provide text-to-text translation in over 100 languages for customer care departments of companies in multiple industries globally. The software is focused on translation across all e-support channels including email, articles, chat, and social messaging among others.

Nature of Disruption: Language I/O leverages natural language processing techniques to engineer unique features from the data, which power the machine learning models. It selects neural machine learning models for requests and adopts preferred translations for product names, misspellings, acronyms, industry jargon, and slang. It also uses a special type of unsupervised neural network ‘self-organizing map’ to automatically detect and flag anomalous content before a person even sees it. A second model uses this data to identify potential glossary terms and the external translation quality feedback allows it to adjust and improve over time. Language I/O offers an API that allows clients to get company-specific translations and connects with customer relationships management systems such as Zendesk, Oracle, and Salesforce. These systems benefit from the feedback provided by agents and the professional linguists that Language I/O works with to make adjustments to its core technology.

Outlook: The impact of translating information into other languages on enterprises is significant. Language I/O claims to address the language-translation issue that seeks to perform more accurate, personalized translations via an engine. Customers tell Language I/O which words they want in their dictionary, which in turn enables the models to improve over time across more than 100 languages. The startup has raised $5m in a recent round of funding which it plans to use toward customer acquisition as it expands the size of its workforce.

This article was originally published in