#language technology

This month’s linguistics jobs interview is with Brent Woo, a Language Engineer for a voice assistant. Brent has an MA from Eastern Michigan University and a PhD in Linguistics from the University of Washington. In this interview, Brent describes how his career provides him with work-life balance and how his linguistics training ties into his work developing technology we use every day, but might not realize requires language data to run smoothly. 

You can find Brent online on LinkedIn. 

What did you study at university? 

Linguistics and more linguistics! I entered undergrad not really knowing what I wanted to do. I took some finance and economics with the vague idea of becoming an accountant, but I didn’t have an interest in the material. I took some political science and cultural studies classes with some interest in “international affairs” but it still wasn’t quite right for me. I took an intro linguistics class and knew immediately this was it. This was the mix between analytical thinking and interesting human issues that I was looking for. I went on to finish BAs in Linguistics and Russian Studies, and then an MA and PhD in Linguistics. What is your job? 

Language Engineers work on the language designs for voice assistants. When you say something funny like “let’s raise the roof” and the voice assistant can interpret that to raise its volume to the maximum level, that was likely due to an LE language design. LE look into common variations on requests, understand the failures in human-machine communication, and adjust recognition coverage for them. We work on the scale of millions of utterances a day, so this process also has to be automated to some degree. How does your linguistics training help you in your job? 

I’m lucky to be considered a subject matter expert in language, and can usually make compelling arguments for utterance coverage based on language and human arguments alone. A common scenario is product or business stakeholders will request to have utterance X go to feature Y, and an LE will know from their expertise that no sane human customer would utter X to trigger Y. The LE has to present the case why that wouldn’t be a good design, and propose alternatives, and overall help the feature get better language coverage. This involves classic syntax for figuring out sentence structures and variations, and semantics for the annotation schema. Sometimes phonetics comes up when there are speech recognition conflicts or misfires! What was the transition from university to work like for you? 

Graduate school was great preparation for moving into this position. I already had experience running quantitative experiments and analyzing the results, managing different stakeholders (advisor, students, lab, external collaborators), and—most importantly—dealing with difficult customers. Two things were new to me at the beginning: having conversations at the business and product level, and the pace of work. I wasn’t familiar how to navigate conversations about project budget and headcount, but I had a good friend as a mentor who helped me out with these early on. The pace of work differs from grad school: work comes and goes with the product release cycles, there are periods of crunch time and periods of relative calm and catch-up. Do you have any advice you wish someone had given to you about linguistics/careers/university? If you’re interested in exploring a non-academic path, talk to linguists in those paths, watch the panels and webinars we’ve run on these topics, ask for mock interviews, and read Superlinguo! You’ll get a sense of what background you need and what projects you should be working on. 

Any other thoughts or comments?

The most important thing for me when finding a job after graduate school was to find a work-life balance. While I’m happy and lucky to have an interesting job related to language, I was really thrilled to discover that it was very flexible and I can live a rich life outside of work responsibilities. There were other jobs that may have been a better fit, but were more demanding on my lifestyle or diplomatic capacity (for example, managing people). I have many colleagues at this and other companies who have taken on more intense positions and end up burning out or quitting due to the stress. I find the work-life balance in this position almost ideal and I am very happy.

Related interviews:

Interview with a Natural Language Annotation Lead

Interview with a Software Engineer

Interview with a Data Scientist

Interview with a User Experience (UX) Researcher

Interview with a Computational Linguist

Recent interviews:

Interview with a Data Manager & Digital Archivist

Interview with a Natural Language Annotation Lead

Interview with an EMLS/Linguistics instructor & mother of four

Interview with a Performing Artiste and Freelance Editor

Interview with a Hawaiian and Tahitian language Instructor, Translator & Radio Host

Resources:

The full Linguist Jobs Interview List

The Linguist Jobs tag for the most recent interviews

The Linguistics Jobs slide deck (overview, resources and activities)

An article in Slate about language revitalization in the digital era. Excerpt: 

At a recent conference called Breath of Life 2.0, held at Miami University of Ohio, participants explored the possibilities and pitfalls of archival databases. For the people involved in language revitalization work, whether they’re awakening a dormant language (one that no longer has fluent speakers) or trying to prevent an existing language from losing all its speakers, one of the major obstacles is a lack of digital resources.

Jerome Viles experienced that firsthand. An enrolled member of the Confederated Tribes of Siletz Indians and part of the Southwest Oregon Dene Languages Project team, Viles helps organize archival materials produced by linguists and community members over the past 100 years. First, his community had to findthe documents, which were scattered in institutions across the country. But the other problem was organizing and reconciling them. For instance, some records of language may have been written by French Jesuits who came to North America to convert Native American people. But even if three Jesuits worked with the same tribe, they could all have different spelling systems. Then there are audio recordings from the 1800s and 1900s, plus ethnographic material collected by anthropologists.

What Viles’ group and many others needed was a way to collect and compare thousands of words across multiple generations and several dialects. And that’s where the Indigenous Language Digital Archive came in. ILDA is a database built by engineers at Miami University in collaboration with the Miami Tribe of Oklahoma (which built an earlier form of the database specifically for its own language revitalization work). At the Breath of Life 2.0 conference, participants learned how to navigate the software and move their documents into one big digital space. By the end of the five days, a few surprising discoveries had emerged. Community members from the Northern Paiute, working on the Numu language, were tickled to find seven different ways to say “our husband.”

“I was getting a kick out of that,” said Nicholas Cortez. “I was like, where’s this ‘our’ coming from?”

Emily Bender on the coining of the Bender Rule and why it’s important to name the languages we make technology for. Excerpt: 

Though English and Mandarin Chinese are widely spoken, both as first and second languages, clearly a world in which advanced language technology exists only for these two languages is undesirable. The promise of language technology includes pro-social applications spanning a wide range from biomedical applications (e.g. matching patients to research studies or automatically flagging patients for time-sensitive tests based on physician notes), through machine translation of documents available on the web, to interactive tutors for language learning and other learning scenarios and more. These benefits should be available to all. 

Furthermore, the existence of even the most basic language technology (keyboards or input systems supporting the writing system, spell checkers, web search) builds up the value of a language which can be an important factor in self-esteem and educational outcomes for speakers of minoritized languages and can contribute to the maintenance of languages under threat of displacement by local majority languages (see e.g. Bamgbose 2011).

And yet, the field of NLP is caught in a negative feedback loop that hinders the expansion of the languages we work on. Work on languages other than English is often considered “language specific” and thus reviewed as less important than equivalent work on English. Reviewers for NLP conferences often mistake the state of the art on a given task as the state of the art for English on that task, and if a paper doesn’t compare to that then they can’t tell whether it’s “worthy”.

I believe that a key underlying factor here is the misconception that English is a sufficiently representative language and that therefore work on (just) English isn’t also language-specific. This misconception is abetted by the habit of failing to name the language studied when it is English.

A paper from the Workshop on the Use of Computational Methods in the Study of Endangered Languages (ComputEL-3) about making better tech tools for linguistic annotation, featuring many ideas that I wish I’d had in grad school! 

Abstract: 

To reduce the annotation burden placed on linguistic fieldworkers, freeing up time for deeper linguistic analysis and descriptive work, the language documentation community has been working with machine learning researchers to investigate what role technology can play, with promising early results. This paper describes a number of potential follow-up technical projects that we believe would be worthwhile and straightforward to do. We provide examples of the annotation tasks for computer scientists; descriptions of the technological challenges involved and the estimated level of complexity; and pointers to relevant literature. We hope providing a clear overview of what the needs are and what annotation challenges exist will help facilitate the dialogue and collaboration between computer scientists and fieldwork linguists.