Chatbots can revolutionize communications with consumers and business clients. But some international companies run into problems when implementing a business strategy for multilingual communication. What are the best strategies for immediately identifying the language of each engaged user? How do you communicate best with users who don’t speak one of your currently supported chatbot languages? How do you handle supporting documentation in multiple languages? We’ll look at some best practices, tips and tricks for improving the multilingual capabilities of your chatbot so your users and bots don’t get lost in translation.
A Business Strategy for Multilingual Chatbot Linguistic Enhancement
A business strategy for implementing chatbots on multilingual websites and social media aims to reach additional markets for a low incremental cost. Anyone who has implemented a chatbot system knows that the first part is the hardest: getting a good system up and running for a single language. The incremental cost for each additional language is relatively low.
If you’re already exploring or working with Chatfuel, you probably know that chatbots are a fast-growing niche. According to Statista, the size of the chatbot market is expected to reach $1.25 billion in 2025, a huge leap from 2016’s market size of $190M. A factor in the dramatic increase is the global benefits of chatbots. The multilingual capability expands a bot’s reach from one to many markets.
Which languages should you include? Obviously, the choice will depend on the specific products or services you offer, and the interests of a specific linguistic target market in those offerings. Still, some guidance can be found by considering the most spoken languages in the world to gauge the size and market for each language.
Obviously, Mandarin and English lead the list, but there may be some surprises in the top 10 that can inform and guide marketing decisions as well as technical considerations. For example, Arabic is in 5th place but implementing this language for your chatbot would require right to left capabilities, a significant technical requirement.
But multilingual chatbots can also be fine-tuned for countries like India and Nigeria where many different languages are spoken within the same country.
Chatfuel Basics are Simple to Implement
The basics about multilingual chatbots have been well-covered by Chatfuel in this primer. The basic questions about the subject — what are chatbots? What are chatbots used for?– we assume this audience doesn’t need the answers.
How do chatbots work? As detailed in the above-linked article, Chatfuel’s flow concept makes things easier for you to develop a unified multilingual solution. And the interface is much easier for your users as well.
One bot, multiple language flows is shorthand for Chatfuel’s elegant multilingual solution. The basic steps are easy as 1-2-3:
- Insert content into the bot’s blocks in the language you intend. Multiple languages are supported. This does not address the possibility that the user does not speak any of the supported language but it does channel the user to the best available alternatives.
- Set up the Persistent menu in the languages of your choice. Open the Configure tab and select Localization.
- Then you need to direct your users to the language they desire. One approach is to apply the {{locale}} attribute so that the user’s default Facebook language is auto-detected. But a better way is to adjust the bot’s Welcome Message so users are prompted to select their desired language with buttons or quick replies.
After that selection is made, unless or until the user makes a change, each message from your bot to that user will be in the selected language. Bear in mind that the default phrases for plugins you are using, such as Live Chat, Share Location, or User Input, can be in only one language.
In the following sections, we’ll consider more advanced topics which build on these basics including multilingual chatbot best practices, tips and tricks.
AI Enhancements to Help Make Your Bots Translate Better
Each language has its nuances, as do regional versions of a language. Consider the significant differences between American and British English. Between languages the differences can be vast. Be wary of relying on single trigger words, because they can be misunderstood easily.
AI can help a bot know how to handle slang, idioms, typos, misspellings, abbreviations, even emoticons, even in another language. The goal is to help your bot grasp the intent behind the user’s message so it can respond more accurately and helpfully. To the rescue comes AI-driven machine language assistants that can be integrated to your Chatfuel bot to help them understand and perform more human-style conversation.
- Google’s Dialogflow is a conversational AI tool to help your bot understand the context of user messages better.
- Janis.AI is a framework for developing Dialogflow-enhanced Chatfuel bots. A Slack assistant helps monitor and manage learning and responses to users.
Setting up this pair is not rocket science, and you can follow these instructions. Clearly, Dialogflow is a powerful Natural Language Programming tool, but managing its Intents can be tricky, requiring some creative strategies, some good one contained in this article. Bear in mind that the general language setting should be for a single language.
Another useful tool for extending the capabilities of your Chatbot can be Zapier, a handy tool for connecting various services and tools into a coherent flow. While an in-depth treatment of its capabilities is beyond the scope of this article, you consult this resource to add flow-based text-to-voice, voice-to-text, or voice-to-voice transcriptions and translations
All the AI in the world, however, is no substitute for high quality professional translation of the specific language pairs you need to support. Otherwise, you risk a classic, and messy, GIGO situation: garbage in, garbage out. With respect to the various AI tools available, whether NLP or machine translation, they’re improving but they’re still no match for an expert human translator.
The Value of Professional Translation for Optimizing Multilingual Chatbots
No matter which languages you choose, you’ll need help with professional translation services. These can be found through localization services, translation companies, freelance translators, and machine translation software. The logical place to start, among them, is with a translation company, because you should be able to obtain good strategic and practical advance… for free. Free is always a good way to start.
Let’s take our simplest multilingual bot instance. We seek to translate from English to Spanish. There are many ways from here to there. The easiest (and most expensive) is to have a translation agency or localization company do the translation for us, then just copy and paste into our Chatfuel boxes. That way we can be confident of the translation quality.
Bear in mind that the need to localize content involves more than language translation. You need to convert numerical and date formats, measurement and currency units. More significantly, there may be cultural issues that you need to consider. All these issues should be easily handled by a competent agency specializing in multilingual communication, especially in websites, apps, social media, and other digital uses.
Professional translation agencies can also help efficiently with the translation of documentation, including help system, whether or not rendered as a chatbot.
Freelance and Machine Language Translation Services as an Option
An alternative to working with translation agencies would be to hire a freelancer for each language you need. The advantage is lower cost and more control. Go to a freelance marketplace like Upwork, Freelancer, or Fiverr. Search with keywords like translator and the languages you seek to translate to and from. You’ll get plenty of offers, together with the freelancer’s profile, ratings, rates, and reviews. Shortlist 3-5 and engage them to get the best deal.
Make sure the chosen agency or translator has experience with chatbots. Typically translation work deals with longer text that are used in chatbots. This demands a clear understanding of how bots work and the increased urgency of grasping the intent of short conversational snippets found in user messages.
You may well be tempted to use machine translation software like Google Translate or Microsoft Translator. Admit it: you’ve given in to the temptation. In terms of translation speed, online machine translation is obviously the fastest way to go. Just copy your text, paste it into the box of Google Translate, select your language, and out will come a rough but usually decent translation. It’s adequate for testing purposes, but not quite ready for primetime in terms of linguistic quality. You may get away with machine translation 90% of the time but the remaining 10% can cost you dearly. Embarrassing mistakes can easily creep in.
At the very least, hire a professional – or better yet, a pair of human linguists — to proofread your machine translation. That’s likely to save you money, and give you a layer of CYA protection. If there’s an embarrassing mistake, at least you’ll have someone to blame. Otherwise, you may find that your user, if not your bot, blames you!
