Christopher Johnson

I'm excited to be joining the Human Computation team at Twitter as an Analytical Linguist! I'll be creating language annotation workflows.

DefinedCrowd uses its crowdsourcing platform Neevo to collect and annotate data for clients' AI/machine learning projects. As a member of the UX Design team I created guidelines for speech and NLP data workflow products (including speech collection, transcription, transcription validation & correction, named entity tagging, and quality of experience studies), and I wrote FAQs for the Neevo app about speech transcription and transcription validation. I also participated in the design process of the whole UX team, attending twice-weekly design review meetings and reviewing, revising, and writing copy for buttons, checkboxes, tooltips, error notifications, pop-ups, and other UI elements.

As a member of the Project Execution team I wrote guidelines for crowdsourced data collection and annotation projects in speech, NLP, and vision. These included speech collection and transcription, transcription validation & correction, document classification, named entity tagging, image collection, and image tagging.

My teammates and I were responsible for the language data used to train and evaluate machine learning models that added new features to Alexa. Most of my work focused on improving Alexa's ability to interpret contextual references (pronouns and other expressions with anaphoric or deictic interpretations). My role involved data collection, creation, enrichment, and analysis. I developed annotation workflows and conventions, wrote and revised annotation guidelines, did gold standard annotations, trained annotators, and verified their work. I also identified new use cases, devised data sampling strategies, wrote prompts for data collection, and did linguistic analyses of the things people say to Alexa. The ultimate aim of the work was to define Alexa's linguistic ground truth.

Worked half-time as a Creative Strategist contractor at Microsoft and spent the other half doing brand strategy, messaging, copywriting, and naming for other Northbound clients.

Created a distinctive online voice and brand with my blog The Name Inspector and built a naming business around it. The blog about language and naming earned media coverage in The New York Times, The Wall Street Journal, The Washington Post, Newsweek, Forbes, Entrepreneur, Inc., and other publications, allowing me to build a client base with no advertising. For my clients I created many company names, and also did product naming and copywriting. My best-known name is Zulily, for the Seattle-based online retailer. It's featured as the sponsor name on the MLS soccer jerseys for the Seattle Sounders FC.

As the first Professor of Practice in the UW College of Arts and Sciences, I taught half-time in the Department of Communication about language in branding, social media, and popular discourse, drawing on my experience as a verbal branding consultant and my background in linguistics.

My book Microstyle: The Art of Writing Little was published by W.W. Norton & Co. in July 2011. It's a guide to the linguistic techniques that make very short messages interesting, memorable, and effective, and it got supportive reviews in The New York Times, The Wall Street Journal, The Washington Post, NPR Books, USA Today, and The Huffington Post, among other media outlets. Microstyle has been translated into simplified Chinese (for mainland China), traditional Chinese (for Taiwan), Korean, Turkish, and Bulgarian, and rights have been sold for Brazilian Portuguese as well. I did several radio interviews to promote the book, including a live hour-long call-in show on Wisconsin Public Radio.

Taught and conducted research in linguistics and child language acquisition in the Department of Comparative Human Development.

Designed and implemented a software component for representing verbs and their argument structures in WordsEye, a system that converts text descriptions of scenes into 3D visual representations. Applying frame semantics and object-oriented principles, implemented the component in Common Lisp.

Worked on natural language understanding technology for a text-based e-commerce chatbot. Oversaw development of general English lexicon and grammar, re-structured lexicon according to object-oriented principles, managed two developers, coded in Smalltalk, led development of regression test corpora, initiated data mining of customer interactions to identify needed new features.