MDS Spotlight: Meet Toshiko Shibano, MDS Computational Linguistics, Class of 2020
As a professional translator, Toshiko Shibano has always been interested in machine translation. In particular, she wanted to know how linguistic data were processed in deep learning models and wanted to be able to test out her hypotheses about how to improve current machine translation models.
Prior to joining the UBC MDS Computational Linguistics (CL) program, Shibano ran a translation business in Japan that specialized in financial accounting for almost 20 years.
“I picked up programming to automate routine processes so that I could spend more time on translation. As my business grew to become one of the top translation companies, I expanded my SQL-based business management system to streamline operational processes, to collect data, and to monitor Key Performance Indicators (KPIs),” Shibano said.
Her desire to pursue a data science career was furthered where she saw the drastic impacts of data-driven decision-making in her translation business.
After handing over her translation business, Shibano moved to Canada and earned a Bachelor of Arts in Psychology from UBC. Shibano then spent a year studying computer science at UBC before deciding to switch gears and apply for the UBC MDS-CL program.
In fact, Shibano was also accepted into the UBC Master of Data Science (MDS) program but ultimately chose the Computational Linguistics program as she knew she would be able to make a more meaningful contribution to language modeling and linguistic feature engineering.
“There are people who are truly gifted with mathematics and statistics. My focus is to find my own niche. Half of the curriculums of MDS-CL and MDS overlap, so I thought as long as I master the core curriculum, I would be able to collaborate with mathematicians and statisticians,” she explained.
To help her prepare for the MDS-CL program, Shibano reviewed calculus and probability, and took matrix algebra at UBC. She also recommended studying the basics of the Bayes theorem and reading as much as possible on Natural Language Processing (NLP).
One of the most important things that Shibano will take away from the program and bring into her career as a data scientist is the importance of identifying methodology that is statistically grounded and establishing appropriate metrics to evaluate the results.
Shibano listed the hours of structured and guided coding practice she received every week in helping her prepare for the job market. “These intensive, guided coding assignments definitely prepared me for working from concept through to execution.”
With the program completed, Shibano hopes to translate her degree where she can be involved in data science projects as an NLP engineer.
“Taking advantage of my experience in business, translation, and psychological studies, I would like to design models and metrics that are enriched by linguistic features that capture human cognition and sentiment,” she added.
Finally, if there was one thing she would tell potential students looking at the MDS-CL program is this:
“If you are like me and enjoy devising solutions to real world problems, have a desire to transfer knowledge and skills from a previous job to the digitized world, or are passionate about language, this program is worth the money and time. CL has seen advancement at a rapid pace. Learning the state-of-the-art techniques in an academic setting with supportive faculty would definitely give you an edge in the job market!”
Toshiko’s Top 3 Tips on Succeeding in the MDS Computational Linguistics Program:
- Don’t be afraid of asking questions! The faculty understands the diverse nature of student backgrounds. As a student with an arts background, I asked questions like “Here is my hypothesis and experimental setup. Does this statistically make sense?”, “Why is entropy computed this way?”
- Always be kind. Stay away from unnecessary rivalry!
- Do not be discouraged by grades. The admission to this program is competitive, and your classmates are highly competent. It’s only natural your grades sometimes suffer.