A state government agency wanted to take their administration closer to the citizens by offering multiple citizen-centric services in an integrated, convenient, citizen-friendly manner, using IT tools.
Background:
- The state agency wanted its constituents to enroll themselves for various government programs.
- Disparate programs had different government websites with a plethora of documents to be uploaded for each service.
- There were too many ‘Frequently Asked Questions’ (FAQ’s) listed on the websites for every government program and the answers did not meet all language dialect requirements.
- Constituents found it difficult to find the correct website and FAQ section to get their queries addressed.
Solution:
- Client needs were evaluated and we implemented a chatbot to address queries across multiple websites.
- The open source chatbot was implemented on-premise to address queries in English and local languages.
- The solution utilized Text Analytics and Natural Language Processing (NLP) techniques to analyze customer queries and provided automated responses.
- The Natural Language Processing (NLP) engine broke down sentences into components, which were mapped against keywords of FAQ’s, to identify the appropriate response to the user question.
- The engine also performed a syntax check of the user question and mapped the components against a lexicon repository, to understand the meaning and context.
- Based on the context, the chatbot also displayed documents to be uploaded by the users for different programs for enrollment.
- For registered users, the chatbot saved chat history and referenced previous history to address current needs, if they were related.
Benefits:
- Chatbot responses were available 24 x 7 x 365 versus during regular working hours only.
- The chatbot provided responses in multiple languages.
- Response time was reduced dramatically.
- The chatbot redirected the customer to appropriate websites for more detailed information.
Technologies Utilized:
- Text Analytics
- National Language Processing