How to build (real) multilingual Oracle Chatbots
A few months back we at BotSupply were selected for the Oracle Scaleup Ecosystem which is an Oracle program that helps scale-ups grow their businesses and well … scale. This partnership gives us access to Oracle Cloud technology, co-sell opportunities with Oracle customers and more.
For us, this meant premium access to the Oracle Bots Platform and the opportunity to develop a BotSupply Natural Language Processing (NLP) SDK for Oracle Bots. The SDK enables Oracle customers to build bots in a wide range of local languages.
Read on to get a bit of context and learn what this SDK thing means in practice.
What is the Oracle Bots Platform?
Oracle Bots Platform is a powerful end-2-end bot framework & platform built into one and is part of the Oracle Autonomous Mobile Cloud Enterprise. The ‘autonomous’ indicates that it’s an Oracle-managed service, which means that Oracle takes care of all the underlying services like channel management, databases, storage, scalability and so on. All you need to worry about is building your bot and integrations. Super cool!
Oracle Bots: Natural Language Processing & Translation Services
Out of the box, Oracle Bots Platform comes with English Natural Language Processing (NLP). Even though NLP support is in English, you can still add multi-language support for your bot. Using translation services from Google and Microsoft, your bot can automatically translate the user’s messages that it receives and your bots own prompts and replies to and from English.
Using translation services has pros and cons.
Cons of using translation services
Recently I overheard a conversation where someone mentioned ‘How does your brand sound?’ This is what conversational design is all about. Conversational design enables companies to create new and potentially exceptional customer experiences with the use of conversations. When designing for conversations, the tone, voice, and wording are fundamental. Even though translation services have come along way, most people can still see if a text has been through translation services, or not. So the question to ask yourself is, do you care if your brand voice sounds like Google Translate? If I should highlight another con of using translation services that we’ve encountered frequently. Big enterprises and government institutions that are not loving the idea of sending their data through translation services.
Pros of using translation services
For example, a company with customers in 50 countries but with a centralized support team that doesn’t speak all 50 languages anyway. Chances are that these companies are already using translation services in their support workflows.
Another example could be a local tourist organization that gets visitors from all over the world. Firstly, tourist organizations often don’t have the resources to create and manage a bot in 50 different languages, and this is where the translation services come in handy. Secondly, if a visitor is looking for the beach, nearest Italian restaurant or stadium, it probably doesn’t matter much if the phrasing and wording of the conversation is spot on and aligned to ‘brand voice’ as long as they get the info they need.
Extending the Oracle Bots language capabilities with the BotSupply Natural Language Processing (NLP)
As you may know, we’ve spent the last few years building our award-winning NLP technology from the ground up with world-class AI scientists from around the world. Our state of the art NLP is amongst others utilizing various machine learning (ML) algorithms to detect user intents and extract entities in multiple languages. Our newest NLP release from June 2018 also comes packed with many other cool features like autoML, synthetic data generation and more that you should check out.
But also, the last few months we’ve been developing an SDK for Oracle Bots. The SDK enables fast and simple integration of our NLP into Oracle Bots so Oracle customers seamlessly can build bots in a wide range of local languages.
Integrate our SDK and start building Oracle Bots in the following languages
Tier One Languages
Intents and fifteen (15) general entities out of the box
Tier Two Languages
Intents and three (3) general entities out of the box
- and 34 other languages
I want to mention that besides the general entities you can also add and train custom entities. Furthermore, we’re continuously upgrading Tier Two Languages to Tier One Languages based on client demand.
There you have it, building (real) multilingual Oracle Bots has never been more straightforward.