CASE STUDY
Translate Together
NLP & MACHINE LEARNING
A gamified tool to assist translators and the general public with translating into the 4 national languages of Singapore.
This tool is powered by a Natural Language Processing engine (NLP) and artificial intelligence (AI). It uses machine learning to improve the accuracy of translation and train the NLP.
I lead a small team in ways to train and NLP translation system that would bring in users to train a system by submitting copy for machine translation and then rate the translations.
My Role
Design Lead
Design Researcher
A gamified tool to assist translators and the general public with translating into the 4 national languages of Singapore.
This tool is powered by a Natural Language Processing engine (NLP) and artificial intelligence (AI). It uses machine learning to improve the accuracy of translation and train the NLP.
I lead a small team in ways to train and NLP translation system that would bring in users to train a system by submitting copy for machine translation and then rate the translations.
My Role
Design Lead
Design Researcher
CASE STUDY
Translate Together
NLP & Machine Learning
A gamified tool to assist translators and the general public with translating into the 4 national languages of Singapore.
This tool is powered by a Natural Language Processing engine (NLP) and artificial intelligence (AI). It uses machine learning to improve the accuracy of translation and train the NLP.
I lead a small team in ways to train and NLP translation system that would bring in users to train a system by submitting copy for machine translation and then rate the translations.
My Role
Design Lead
Design Researcher
My Team
Jessica- Lead & Interaction Designer
Jason- Visual Designer
Our Plan
This tool is powered by a Natural Language Processing engine (NLP), a combination of computer science, information engineering, and artificial intelligence (AI). It uses machine learning to improve the accuracy of translation and train the NLP.
Upon meeting with MCI they had a vague understanding of what they wanted. They had an NLP they needed to “train”. Many people on the team work with translators on a daily basis and wanted to use their brains to improve/ train the NLP engine.
The team wanted to gamify the process of submitting translations. Our role was to figure out the best way to do this. We started off by doing a competitive analysis and contextual inquiries.
At the end of the day, we brought our findings back to the users. They key finding was that users were not motivated by gamification, feared that a leader board would penalize them career wise, and that they really wanted tools and a community of other translators to share with.
Competitive Analysis
We looked at over 6 in-depth tools and 60 inspirational products to bring feature ideas to the table. We also wanted the team to have a better idea of tools already in the translation landscape. We also wanted to identify opportunities to focus on for quick wins and long-term plans.
Contextual Inquiries
We set up a mixture of business interviews and contextual inquiries. The business users had one concept of the life and needs of translators, but the contextual inquiries showed different things.
Our Plan
This tool is powered by a Natural Language Processing engine (NLP), a combination of computer science, information engineering, and artificial intelligence (AI). It uses machine learning to improve the accuracy of translation and train the NLP.
The team wanted to gamify the process of submitting translations. Our role was to figure out the best way to do this. We started off by doing a competitive analysis and contextual inquiries.
Competitive Analysis
We looked at over 6 in-depth tools and 60 inspirational products to bring feature ideas to the table. We also wanted the team to have a better idea of tools already in the translation landscape. We also wanted to identify opportunities to focus on for quick wins and long-term plans.
Contextual Inquiries
We set up a mixture of business interviews and contextual inquiries. The business users had one concept of the life and needs of translators, but the contextual inquiries showed different things.
Features & Sprints
We then developed a prioritized rough features list with the team and broke things into 4 design sprints. We took 2 sprints just experimenting on the layout and features in the translation tool.
Each sprint had a kickoff followed by wireframes that we turned into a prototype to take to users. We conducted think aloud sessions that allowed us to greatly improve the designs.
Design Choices
After creating 2 wireframe concepts and blending the best performing features for optimal results, we started working on additional compelling new features like translator tools and a translation community. The idea was to develop additional offerings that would motivate translator to return and submit translations into the NLP tool.
We took the wireframe into visual design and utilized an illustration style to bring a playfully, polished look to the tool. We focused on simple bright colors as accents and brought the requested side-by side translation into the tool.
Design Choices
We then developed a prioritized rough features list with the team and broke things into 4 design sprints. We took 2 sprints just experimenting on the layout and features in the translation tool.
Each sprint had a kickoff followed by wireframes that we turned into a prototype to take to users. We conducted think aloud sessions that allowed us to greatly improve the designs.
The Results
The primary success measurements on this project were:
- Create a concept that could be built later in the year
- Number of translations submitted for NPL training
- Number of translators signing up
- Usage by translators
The SG Translate machine translation engine was first rolled out for use by Government agencies in 2019, and has since generated more than 300,000 translations. Our firm was charged with the design portion but not development portion.
300,000+
Translations
The Results
The primary success measurements on this project were:
- Create a concept that could be built later in the year
- Number of translations submitted for NPL training
- Number of translators signing up
- Usage by translators
The SG Translate was first rolled out for use by Government agencies in 2019, and has since generated more than 300,000 translations. Our firm was charged with the design portion but not development portion.
300,000 +
Translations