Project Overview

TRNSL8 is the concept for a product which expands on traditional translation software to provide accuracy, context and cultural nuance. 

It is intended for people who are at least conversationally bilingual, and who have found themselves needing to translate for people in their lives.  TRNSL8 pulls data from around the web to provide emotional, legal, technical, and any other relevant context around suggested translations. Crowd sourcing in the spirit of Captcha and Duolingo provides a level of certainty that doesn't currently exist, and allows the product to grow without a huge investment in paid language experts. I completed this concept project during a 10 week User Experience Design course at General Assembly, under instructor Katherine Hlavac. We were tasked with identifying a problem we had observed in the world around us, but one that had not affected us personally. 

THE PROBLEM

People who are conversationally bilingual are being forced to act as translators for relatives for technical jargon, legal documents, medical communication, etc.

The Solution

Use machine learning & crowd sourcing to provide context around translations for technical or nuanced words. 

Peripheral landscape Research

Combining features & affordances from across our users' experiences  -

Translation tech is spread thin across products, forcing users to hack a solution together. It was important to first understand the limits of current technology before throwing out ideas for features that could never actually be built. Translation software is both a feat of human achievement, and yet so far from being a reliable tool that translation fails are a universally understood joke. My challenge was not to code a better translator, but to take available technology and users' affordances with other products and craft an experience around it that would provide the value people are seeking.

Generative Research

The translation problem is much larger than I ever would have thought  -

My subjects were spending all of the time they had with their relatives translating, and in some cases it was causing a great deal of tension. For this project I completed eight interviews, each 45-60 minutes via Skype. I reached out to friends and family of friends I had grown up with, as well as finding a couple of participants in online forums who were interested enough in my project to give me their time. I spoke with speakers of Spanish, Chinese and Korean but upon continuing to research and iterate I would talk to people from as many different regions and languages as possible to account for cultural nuances. I created this discussion guide as a basis for the conversation with my participants, but often skipped or added questions based on what I learned from them.

Persona Creation

The art of making choices for other people -

Everyone I spoke with had interesting and unique experiences, but the problems they faced overlapped a great deal. For purposes of this project, creating personas was a beneficial exercise because it allowed me to condense and personify sets of needs and behaviors that were not part of my own experience or thought process. I was able to more easily keep the needs of my users' top of mind during the design process.

Yuriana and Justin (visible to the right in this carousel) cover the two main types of users who need this product. Another persona type I considered was someone who was highly fluent in two languages, and needed heavy focus on nuance and context for translating for themselves rather than others.

DESIGN VALUEs from Key Insights

Creating a north star vision  -

To synthesize the information I gathered during the Contextual Inquiry, I wrote down key quotes and observations about peoples' attitudes and behavior while speaking with them. I arranged these into themes, rearranged, and then rearranged again after sharing and discussing with my classmates. For each of these themes, I wrote a Design Value statement to create the framework for how I would approach solving the users' problems through design. The way that I have used Design Values is to be a guidepost to check all features against while designing a product. 

01.

Users need to be able to verify when translations are correct in the context of what they are translating for and give users peace of mind.

02.

The experience should capitalize on users' affordances to combine resources and techniques to create a more quick and efficient tool.

03.

The primary function of the tool should assist users in translation of legal and technical jargon. There is a secondary need of translation for highly nuanced words and phrases that require knowledge of cultural context.  

04.

Users need to know when the system is not 100% confident.

05.

The tool should help facilitate communication between tech-savvy primary users and tech-averse secondary users, even when separated by physical distance. 

06.

The tool must have room to grow as more research is done around nuances of different cultural norms and dialects of languages. 

07.

The tool should leverage machine learning and crowd-sourcing to grow its data and refine accuracy as use increases. 

08.

The tool must still be functional when direct or indirect users desire privacy about any subject.

STRATEGIC CREATIVITY

Goal Narratives, User Flows & Sketches 

Before jumping into wireframing solutions, I broke down ideation into more digestible steps.

1. Ideate on features and solutions without any assigned priority, then map those back to the Design Values to see what holds up.
2. Writing goal narratives for my users. I wanted to think through my users' interactions in an organic, narrative framework.
3. Creating simple user flows based on those scenarios that start to break down necessary interactions within the context of a digital product.

Creating a Prototype

Understanding that I don't know what I don't know  -

 This feedback is reflected in the wireframes in the following section. It was important to me that even though this was a class project, to make this experience truly human-centered. The challenge of creating a prototype for this project was that the value is so heavily reliant on user input and feedback from the system. For a lot of users I interviewed, translation is a much more emotional experience than I initially realized. In an ideal world, with funding, I would test this as a basic but coded, working prototype. I would also continue to do generative research and talk to people from a broader range of cultures, speaking different languages with different nuances.

Findings from User Feedback

01.

The features of the app need to be more connected overall to give users the most efficient experience. The translator chat should be smarter and identify if a user is translating several words in the same category or from the same public document.

02.

Add filtering and searching when viewing the real world examples pulled in for context. This feature is what users got the most excited about.

03.

This app could also be very useful for someone who is highly fluent in two languages but ocassionally needs to translate jargon and highly connotative words for themselves.

04.

Hidden functionality (such as tapping a comment to interact with it) could be revealed through subtle motion design.

05.

It’s too text heavy and could use graphics to break up the content.

06.

Users want to multi-select content to share.

FLUIDITY & FLEXIBILITY

Translation when you're already bilingual  - 

Language is so nuanced and so evolutionary that when someone needs to translate something more advanced than basic survival conversation, it becomes really tricky. A huge thing I learned during user testing is also that translation for a bilingual person does not flow only one way like it would for someone learning a language. Peoples' thinking can start in either of the two languages, and being able to swap easily back and forth while maintaining consistent feedback is key. Users can tap "translate" to swap the starting language, while maintaining the same translation "conversation." 

TAILORING TO EXISTING BEHAVIOR

How can technical become conversational?

Using a chatbot interface for translation allows a user to think more naturally about language questions. If their concern is a single word, it's still an easy format that they are used to from other translation tools. However, if someone wants to ask a question, it's a great opportunity for the system to learn very basic answers, or to direct a user to another resource like documents or forums.

CONTEXT IS KEY

Turning users' hacks into features  - 

I had already heard from multiple research participants that they had a hack in place - to use Google Translate, then do Google searches to see it used in sentences. The quickest way to confirm if a translation is correct is to see it in context. TRNSL8 searches the web for examples of a word being used in "real life". It breaks down usage into categories that show formal and informal use, and literature versus journalism versus technical industries like medicine. The system will know if this word is often used in metaphors or puns, and its easy to see if a word is not being used at all for something like insurance. 

The human element

Humans and robots, working together  -

Translation between languages is not one for one. We have puns, slang, metaphor, and connotations that are understood differently by different groups. People are the ones making and changing the rules about language, so the ability to connect people together is a necessary piece of this product. TRNSL8 provides a forum that values every single question and comment, and encourages participation. A system of upvoting and downvoting gamifies the experience and incentivizes high quality contributions. Users are able to share any topic or comment, so if they find something that perfectly translates or explains a question, there is no need to try to re-articulate this to someone else.

GROWING A COMMUNITY

Incentivizing use without alienation  -

TRNSL8 provides a community that is not just a backup for when the translation system fails, but also serves to contribute to crowd-sourced data. 

To ensure that the community is a living, breathing source of information, I know that people need to be incentivized to participate. With limited time on their hands, people will be tempted to just get their question answered and not contribute. The solution is to require a certain number of points to ask questions - and users will be awarded points by interacting by commenting, voting, answering questions and completing activities. 

EFFICIENCY

Working together to make less work  -

Once one user completes and verifies a translation, no other user should have to do this again. TRNSL8 features a workflow for taking an entire document and translating it all at once. 

The system identifies words that either could not be scanned, or that have been commonly flagged as controversial or having multiple meanings. Users can read through, view words to see discussions around them, make adjustments, and ultimately submit this form to a public database that can be verified through crowd sourcing. Once a document has been verified, it will be available for anyone to download.