AskVic WhatsApp Chatbot

"Victoria's multicultural communities have been disproportionately affected by the pandemic. As a result of the fast pace of information change and the ever-evolving Covid-19 situation, information is inconsistently translated into their native languages. In Victoria, 26% of the population speak a language other than English at home and 266,000 Victorians are not fluent in English resulting in critical information not reaching several specific communities. The problem that THINK HQ proactively aimed to solve was clear: how do we remove the language barrier for all Victorians to easily access relevant, updated and translated information allowing equal participation in everyday life?"

We created the AskVic WhatsApp ChatBot in 8 languages respecting the cultural and linguistic nuances, and accessible in What'sApp, a familiar channel already used by CALD communities. The WhatsApp ChatBot covers all essential topics for CALD communities to equally engage in everyday life: COVID-19 and vaccinations, Jobs and careers, Business, Housing, Transport, Social support, Education, Law and justice. Accessing the ChatBot is incredibly straightforward, users simply text 0480 032 278, saying "Hello" in their native language. The AskVic ChatBot can respond in 8 languages: Hindi, Arabic, Vietnamese, Turkish, Punjabi, Spanish, and Filipino (Tagalog), as well as English.

The simplicity of accessing the ChatBot via the WhatsApp application means users can ask any question with one touch of the app, any time of the day, any day of the year. In the following months since the Chatbot's launch, the feedback and testing from participants have been highly favourable, with a significant majority finding the ChatBot either useful or extremely useful (70%+). Beyond user delight, the long lasting impact resides in the new opportunities for the Victorian government in connecting people throughout Victoria's multicultural communities, increasing confidence in navigating a post-pandemic Victoria with reliable information using NAATI translations.

One of the key features aims at ensuring users could interact with the bot without any words of English to serve users with the lowest English proficiency. The Artificial Intelligence model was architected to recognise and set the conversational language on first interaction. On first interaction, users say "hello" or its equivalent in the 7 other languages available and the bot would immediately respond in the requested language. We knew that machine learning (ML) translation could not accurately utilise the idioms and finesse of each language to confidently tackle some sensitive topics such as homelessness or health support. The content architecture, using Botfront.io as a dialogue engine, ensured we could utilise NAATI translated content and respect the cultural and linguistic nuances of each language. Other key features included a training model built on Botfront.io and using the Rasa AI so we could adjust the 120+ user intents with appropriate bot responses building up more than 200 dialogues available in each language. Through user and community testing, we were able to ensure the dialogues were tested and adjusted until truly solving essential questions for vulnerable users.