CanadaPublic sector2021 – ongoingStatistics Canada
fastText coding of census write-in responses
Statistics Canada put fastText into production to reduce manual census coding
Problem
Census write-in responses that failed reference-file matching required extensive manual coding. Statistics Canada hired temporary workers for a process lasting months.
Approach
Statistics Canada developed fastText models to replace or supplement manual coding for the 2021 Census. It retained reference-file matching and specialist correction. Models entered production for all but 1 variable considered after subject-matter experts reviewed their accuracy.
Results
No published figures
Takeaway
Specialist review remained part of production. Evaluation also required care because historical human-coded answers contained errors and some classification systems had changed.
Categories
- Size
- Large enterprise
- Industry
- Public sector
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Statistics Canada described model development, evaluation and production deployment for census coding. Expert-coded evaluation samples helped determine which models could proceed. Quality-assurance work continued during production.
Full details are in the original sources.
