My learnings from bot devs and first-hand localizer’s advice (part 2: collab & benefits)
This post was originally published in Chatbots Magazine.
I hope you enjoyed the first part of this article. In this part, I share tips on collaboration and choosing the right partners for your multilingual bot project, and why bother localizing your bot at all — benefits of localization.
# COLLABORATE, IT’S A TEAM EFFORT
Designers will need more than just design tools to do their jobs. […] The reality is: this person doesn’t need to be you, but work with you on that.
— Fabricio Teixeira and Caio Braga, source: uxdesign.cc
Releasing a scripted polyglot bot is a team effort that takes more than professional dev skills. Your work might expand into such areas as AI, NLP training for other languages, script writing and possibly developing different bot personas for different countries, cultural appropriateness, visuals and text l10n, i18n, and more.
Collaborate with Writers
Building conversational interfaces sparked the demand for a new profession — conversational scriptwriting, and conversational or chatbot writers are increasingly being hired by some of the largest tech companies.
From Google hiring writers from Pixar and The Onion to help build Google Assistant to Microsoft’s Tay bot (not the best example, as it was phased out for its racist attitudes) to Microsoft’s Cortana team, which includes a poet, a novelist, a playwright, and a former TV writer…
Obviously, writing for AI is becoming a quite sought-after job in Silicon Valley.
Elizabeth Dwoskin wrote about this for Washington Post in a more extended piece titled “The next hot job in Silicon Valley is for poets.”
Not only tech giants but also indie devs are following the suit. For example, Humani: Jessie’s Story chatbot was created by Hollywood screenwriters.
It’s clear to me that designing the conversational interfaces is not the property of only design or development teams
To humanize their dialog-based AI/autonomous email assistant, x.ai created an entirely new team role —an AI interaction designer. Diane Kim, a former folklore major from Harvard, who used to direct and star in plays and now an AI interaction designer, says:
Think of an interaction designer for a visual interface — they’re interested in how different arrangements of visual elements (such as typography, color, motion, etc.) may influence users’ actions or behavior. Similarly, I’m just as concerned with how the precise word choice, punctuation, and tone of Amy’s dialog will impact our users’ behavior. […] It’s through these subtle means (whether to use an adverb or not, whether to end a sentence with an exclamation point or a period) that we are able to humanize our autonomous agents.
That’s plenty of examples!
And those examples align with the above premise that conversational bots are a team effort and emphasize that skilled foreign language linguists (or comedians, actors etc.) should have a prominent seat at the table.
Invite “Verbal Designers” To The Dev Team
[In 2017] the definition of “design” will loosen up
Cross disciplinary teams will thrive
— 99U, source
Such developments present chatbot writers with job opportunities in tech. And for brands and bot devs who are in conversational commerce this is an opportunity to engage users on a deeper, more personal level.
Engaging means converting.
Any brand wants to turn conversations into conversions. The converting conversations are not just working but also engaging and actionable. And such conversations between brands and customers need to be designed by writers in order to be engaging.
Devs who care about the top quality and have the budget for it, should team up with chatbot writers who will craft engaging conversations for their bots.
The Gyant bot team hired writers who had written the voices of toys before and partly their team has roots in the game industry. And their users show good engagement. Their dev recounts that 89% “actually go from the first to the last question — that’s 15–20 questions!”
Partner With Translators
Designers, developers and translators. A wonder team.
It comes down to dollars. If money is not an issue then a conversational writer who lives in the country in question and speaks the bot’s original language fluently is ideal.
But not everyone can afford hiring pro scriptwriters for all languages. If you’re on a budget, good specialist translators should do a decent job too.
When you already have that text by a writer that mirrors the bot’s personality, you maybe just need a good translator that can transfer the personality to the new language. That doesn’t need to be a writer. Personality of the bot character is already there and chat text is often modern, simple, and uses youth language, and nothing complicated.
Not just any translator will suit such needs, though. Look for ones skilled in creative types of translation (localization, marketing, literary) and those who are willing to become a long-term member of your (remote) team.
Below are just a few cases when having dedicated linguistic partners will prove useful.
Translating User Data and Feedback
You’ll want to understand the data you capture from your users and use it in your analytics.
There will be cases when you’ll want to get feedback and be able to understand it. Conversations are filled with feedback: Users will give it whether they realize they’re doing it or not. Your linguists will help you analyze interactions and learn from them in order to improve the conversation flow and targeting and to train your bot.
Help With User Intent and Trigger Phrases
Linguists with special skills like SEO could help collecting all possible intent phrases that foreign users might type.
For example, using Keyword Planner.
Help with building the vocabulary is obvious.
Also, your linguists can help planning for non-standard language.
Preparing for Non-Standard Language
They’ll help to prepare for sexts, swearing, or gibberish that users will be sending to your bot. They can help with generating or finding obscenities lists in their languages, like these for English, for using those later to detect and properly deal with such messages.
They’ll advise you on text-speak and what words are most likely to be misspelled and how etc.
For typos, they can use an SEO typo generator or similar tool to generate common misspellings and errors
Help in Word Sense Disambiguation (WSD)
Today, AI is having a tough time interpreting the context of a human discourse. The same word or phrase in different contexts may acquire different meanings.
When the machine encounters ambiguous phrase, it essentially is playing a statistical guessing game. Even the state-of-the-art in the Supervised Machine Learning field is achieving only about 90% accuracy.
Working with dedicated linguists, you’ll be able to handle those cases when your AI fails to understand the input and keep retraining it.
Researching For Recommendations
If your bot recommends resources from the Web, they must be in the user’s native language (e.g. a chatbot giving movie recommendations).
But if you don’t speak that language, how are you going to find such resources?
You’ll need collaborators who’ll research and find content in their languages.
A long story short: Those are not your translator’s regular services, so you’ll need to find more flexible and dedicated language specialists willing to work with you on such non-standard project.
# INSIDER TIPS ON CHOOSING RIGHT PARTNERS
Onboarding the right people upfront is crucial.
Because your bot’s language skills is it’s core UX you can’t get this wrong: User sentiments and the bot’s value, popularity, and retention will depend on it.
Since I’m from the translation industry and currently managing an indie localization team I know a thing or two about choosing translators. I’ll quickly share insider tips.
Here are your options with my comments:
- Community, crowdsource — free, but it’s a make-it-or-break-it gamble
- Bilingual friends or coworkers — affordable, but hardly the skilled specialists you need for the above tasks
- Professional freelancers — starting to eat your budget; it’s a hit or miss; plus, it requires time and effort to search job boards and review applications, and they might not be available next time you need them
- Translation agency — a big pool of translators, but more expensive: they will charge extra on top of the price for localization because most often they are a company with some staff, and where does the money for their salaries come from if not from the clients/you? Overall, the agency business model works like so: Quote you the highest price and then find the cheapest translators so that their profit is the biggest while trying to deliver quality work. Plus, you almost never know who does the actual job (whether localizers care about your project or more about $, and if they are localizers at all)
- Dedicated translation/localization team — many vetted specialists, so there will be someone available when you need translation again. Such teams usually don’t have extra staff (CEOs, co-founders, marketers, accountants, cleaning ladies) that drives the pricing up which means no additional costs for you as you’ll be working directly with your translators. And you can pick best-suited linguists for your project one by one. In my view, the best, modern, and agile solution.
Let’s not even attempt to put MT on that list, for obvious reasons
No MT shortcuts should be taken
I share more tips in this post I wrote for IndieWatch.
Here’s Max Lobanov’s take on MT (Localization Manager at Google):
Machine translation, crowdsourcing, and all other horizontal activities had the same goal — to reduce the cost of localization. […] The only problem with this approach is that there are no levels of quality. […] Even user feedback at the travel sites needs to be at least grammatically correct as when machine translated they often say the opposite of the original message, just because they were not post edited. So any localization shall be of the “first freshness”.
— Max Lobanov
His article on pitfalls of taking shortcuts when localizing digital products can be found here.
As previously mentioned, not just any translator will do this kind of job. Look for ones skilled in creative types of translation: website/video game localizers, marketing, audiovisual (with subtitling experience), literary translators and such.
# BENEFITS OF LOCALIZATION
Localization is UX
“Copywriting is Interface Design”. So is translation.
Last but not least.
Adding languages to a bot is sure a challenge, but it’s a benefit for the same reason: You can reach a wider audience, jump ahead of the competition, and take your share of the foreign market before others do it.
People like using things in their native languages, and business owners know this! So, the more languages you can teach your bot, the more clients will be interested in buying your services as a bot dev.
While bots are in their early stages of usefulness, a lot of your potential clients are already realizing that bots is a handy tool and new channel for making money by reaching a lot of people at places they frequent the most — messaging platforms: Bot$.
Website owners (online businesses, e-commerce stores) are realizing that serving customers from around the world your existing content in multiple languages is not enough — you need to provide support to those people in the same language they were browsing your website in. Think multilingual customer support bots.
Marketing agencies are realizing bots give them the abilities similar to email marketing — to reach potential customers directly in a personalized way (one-on-one messages, segmentation, and addressing by the first name). Since most online businesses have international customers, it’s a good idea to start creating and offering them such bots before others catch up to this opportunity. Think conversational commerce bots.
This is the tip of the potential bot devs can tap into if they can build polyglot bots.
It’s not only me.
Makers of PokerBot advise to “translate your bot to different languages” to expand your bot’s reach and list localization as one of the hacks to help your bot get to the top of StoreBot.me. PokerBot is translated into 14 languages.
The ContentRobin bot’s team shares that having dual language interface (Russian and English) helped them expand the bot’s audience. They are not going to stop and plan on adding new languages (Spanish and German first) and forecast it could increase their audience by another 3,000–4,000 active users monthly.
If your bot’s first language is not English, localization is almost a necessity and should pay off better.
Thinking ahead, there are AI devices that speak rather than send text messages (e.g. Google Home), but today they speak only 1–2 languages (Amazon Alexa supports English and Echo also speaks German). There’s a huge potential for teams who’ll make such bots speaking foreign languages.
When localizing such vocal conversational UIs, you’ll need to onboard voice-over artists and the sound design team, but even then, you’ll still need to first write or translate their scripts with linguists’ help.
# FINAL THOUGHTS
Bot localization is in its infancy. No best practices or guidelines exist yet, so now is the time for creative teams to start developing those best practices for multilingual bot development and bot localization.
Brands are just getting started on the conversational bot trend, so this is the perfect time to start offering them the multilingual tools they need.
Start building your polyglot today.
Below are the only few online resources on multilingual bots I’ve been able to find:
- Internationalize your Messenger Chatbot by Klemens Zleptnig
- Reach a wider audience with multilingual chatbots by Inbenta
- How to build a multilingual chatbot for billions of users by Gal Opprnheimer, Built.io
Do you know other resources? Or, do you know other tips on building polyglot bots? Leave them in a comment and I will add them to this article.
Hope you liked this article! If you did, please click 💚 below. And if you digged it and would like to be notified about my next one, please subscribe. Also, join my brand-new Polyglot Bots Facebook group here.