Artificial intelligence is transforming our ability to process and understand data at scale in all sorts of ways, and one of its most interesting applications lies in a computer’s analysis of human speech and writing.
When humans communicate through language, a surprising amount of meaning is merely implied — or even hidden — by the actual words we use. To a computer attempting to understand and interact with humans, we must come across as a terribly vague and contradictory species, and rightly so. If we communicated with one another using the incomplete, nuance-free, keyword-heavy phrases we regularly punch into Google, for example, society would be the poorer for it.
So instead of asking us to dumb down our communication, tech companies are making computers smarter. These four Seattle companies are using a variety of business challenges and opportunities to push the limits of how much a computer can actually understand of our infamously vague methods of communication.
Founded: 2014
Funding: $29.5 million
Details: Textio analyzes job listings as recruiters write them, and makes recommendations in real time to help attract the right kinds of candidates. The results can be pretty impressive: by using Textio’s platform to tweak the language in its job listings, the company helped HR teams at Australian software maker Atlassian source and hire far more female applicants than they had ever been able to find before. And by analyzing the language used in job listings, Textio has also been able to build a unique window on the culture inside various tech giants.
Founded: 2015
Funding: $13.1 million
Details: Seattle’s DefinedCrowd improves the quality and speed of machine learning algorithms by grooming data for their consumption. The company starts by integrating information from a client’s own database with data crowdsourced from a worldwide network of contributors, then unleashes its own artificial intelligence algorithms and quality filters to format the data in a way that makes it ripe for reading by other algorithms. DefinedCrowd’s tools can automatically crawl text-based data tailored to specific requirements, and can learn to understand the sentiment of texts and individual sentences. In the natural language processing space, the technology has applications with text analysis, chatbots and monitoring customer reviews and satisfaction.
Founded: 2014
Funding: $27.3 million
Details: Mighty AI provides groundwork, known as “training data,” upon which tech companies can build their flashy and fantastical artificial intelligence applications. The Seattle company uses a combination of human labor and its own AI tools to train algorithms to behave in the way their creators intended. In addition to its main line of work — which centers around road signal recognition and anticipation of pedestrian movement for autonomous vehicles — the company has done some impressive work in the natural language processing space as well, contributing to the development of chatbots that understand the broader context and arc of an entire conversation, rather than responding to individual messages.
Founded: 2014
Funding: $3 million
Details: Vettd starts with the assumption that no leader of a large enterprise can possibly know and understand every one of their employees. The company produces specialized deep learning predictive models for talent evaluation and classification, helping HR teams leverage their company’s best people to solve business challenges — and do so at scale. Headquartered in Bellevue, the company says it has made advances in the extraction and capture of patterns in natural language, and claims to have sped up model training tenfold while retaining accuracy rates above 90 percent.
.jpeg)
