United StatesPublic sector2020 – ongoingCity of San Jose
Google AutoML translation for SJ311 service requests
San Jose officials reported 90% accuracy for automated Spanish translations in SJ311
Problem
San Jose's initial website and app translation service produced inaccurate language. Residents needed understandable access to non-emergency city services in Spanish and Vietnamese.
Approach
The city worked with SpringML and Google Cloud to introduce AutoML Translation. Community members helped evaluate translations, and staff combined the model with a glossary. Machine-learning translation was integrated with the customer relationship management system in late 2020.
Results
Takeaway
Officials reported a Spanish accuracy figure while saying Vietnamese translations still needed improvement. They did not explain how the accuracy figure was calculated.
Categories
- Size
- Large enterprise
- Industry
- Public sector
- Business function
- Customer service
- Innovation type
- Process innovation
- Adoption
- Partner co-development
- Evidence
- Vendor case study
Sources
Source summaries
San Jose technology officials described replacing an inadequate translation service with AutoML and SpringML integration work. They reported 90% Spanish translation accuracy and continuing improvements to Vietnamese translation.
Full details are in the original sources.