VICA is GovTech’s next-generation citizen assistant platform that leverages Artificial Intelligence (AI) to develop chatbots that are efficient for government agencies to operate, and useful for end users seeking assistance.
As a platform for agencies to build and train chatbots, VICA is engine-agnostic and leverages the latest natural language processing (NLP) technologies like Google’s Dialogflow to achieve better performance and accuracy. VICA’s automated processes help agencies onboard and “train” their chatbots more easily.
NLP is a machine learning technology that involves the analysis of natural language, such as text and voice messages from humans, to help chatbots gain a better and fuller understanding of users’ inputs, and therefore provide more useful and relevant responses.
Some examples of chatbots built using VICA include:
- OneService Kaki Chatbot
- Gov.sg on WhatsApp
- COVID-19 Info Bot
- COVID-19 Chat for Biz
Easier chatbot training process
VICA makes the training of chatbots easier by automatically generating variations of questions-response pairs for you, and you only need to validate these variations. Previously, many chatbots were difficult to train because they require manually generated question-response pairs and their variants.
VICA can also detect potential conflicts between different questions and flag them before the chatbot goes live.
Reduction of “unknown” questions
VICA can cluster unknown questions into related categories, which is one way to reduce the common problem of ‘unknown questions’ that chatbots face – that is, situations where the chatbot fails to understand the user’s question and is unable to provide an informative response. Having too many unknown questions makes your chatbot less useful, as users are unable to find the information they need.
Without VICA, it will be taxing to manually generate the appropriate response each time your chatbot faces such questions.
Greater automation in importing existing datasets for greater efficiency
VICA improves the importing process by generating the question-response pairs for you and suggesting relevant questions for your factual datasets. Most existing chatbots allow you to import content from existing datasets, such as FAQs or information pages. However, this importing process is largely manual and inefficient. You will often be required to manually label the question-response pairs and generate your own questions.
VICA will automate the importing process, and also compile multiple datasets for bulk import, so that you do not have to manually compile them or import them one at a time.
With VICA’s automated processes and a unified chat frontend for the Whole of Government (WOG), agencies can lower the overheads required to build and maintain their chatbots.
These benefits will ultimately translate to a better chatbot, and thus a better experience for your users.
- VICA-powered chatbots will provide quicker and more direct answers to queries.
- By interacting with these chatbots, citizens will save time waiting for replies or navigating websites to get the information they need. This leads to greater satisfaction with government services and increased trust in government agencies.
OneService Kaki Chatbot
The VICA Team collaborated with the GovText team – from GovTech’s Data Science and Artificial Intelligence Division (DSAID) – to develop the conversational OneService Kaki Chatbot on WhatsApp and Telegram, where members of the public can provide feedback on municipal issues by conversing with the chatbot on these channels.
The chatbot uses NLP and Deep Learning to analyse the feedback and can prompt for more information by asking for additional relevant information that is required to resolve the issue. The chatbot is then able to direct the reported case to the relevant agency that the AI model has deemed as the most suitable to manage the case.
The volatile and rapidly evolving situation during the COVID-19 pandemic meant that individuals and businesses found it difficult to keep up with the latest information. VICA-based COVID-19 ChatBots were instrumental in resolving this information gap. Find out more about the COVID-19 ChatBots here.
The VICA team is looking to improve the automation processes, so that fewer human interventions are required in reviewing the chatbot’s responses and in monitoring whether its responses are relevant and useful to users.
The team is also developing ways to better understand user behaviour. Currently, feedback is received via star ratings and thumbs up/down buttons that users provide after interacting with the chatbot.
Another area the team is exploring is how to incorporate implicit feedback via the types of queries posed and their associated contexts.
Reach out to the product team with your queries or feedback through this form.
Last updated 26 April 2023
Thanks for letting us know that this page is useful for you!
If you've got a moment, please tell us what we did right so that we can do more of it.
Did this page help you? - No
Thanks for letting us know that this page still needs work to be done.
If you've got a moment, please tell us how we can make this page better.