Chatbot for Social Networks Surveys
Understanding the needs and expectations of consumers is vital for any business, because the level of customer satisfaction has a direct impact on profits. One of our clients needed a tool to get fast, complete and reliable feedback from the users of his e-commerce site. The solution also couldn’t be time-consuming for his employees. As users are usually more responsive on social networks, it was the preferred communication channel. However, a questionnaire wasn’t a good option, because customers don’t like filling out forms, they want live communication.
It was decided to create an automatic data acquisition system to collect customer feedback through communication on social networks. The system maintains a free-form dialogue on behalf of an employee, recording customer's answers to a general database. The data is collected anonymously to ensure privacy and objectivity of further analysis.
The project comes as a stand-alone application with social networks API and a user-friendly interface to meet the client’s needs. The modular structure of the object architecture is supported. The user database, client application, and artificial intelligence module are separate services which allow flexibility in installation and use.
The project is quite versatile, it can be configured for various social networks and topics relative to customer's business.
The survey system for social networks uses AI which knows how to start a conversation with a user of a social network, explain the purpose of communication and make a survey on a selected topic. It supports a notification system displaying the survey results, collects quantitative data and analyses the results automatically.
The client set the operational hours for the system, which subsequently reduced the workload of the staff by 75%.
Since the introduction of the automatic survey system for social networks, the effectiveness of the data obtained has increased by 80% compared to the method of customer questionnaires.
The project supports English and Russian. After extra training on additional data, the AI module can be expanded for any language. Also, including new dialogs in the module makes it possible to enhance the variability of responses the client wishes to receive from customers. This will only take the replacement of the AI module, instead of updating the entire system.
The choice of Python for the project allowed relatively quick and simple software development, as well as access to a wide range of libraries. Artificial intelligence is powered by RASA, an efficient open source machine learning platform for intent recognition and automation of text and voice conversations. The project was created as a separate application for PC and can be easily ported to other platforms with a web interface and implementation in the cloud service. The application interface was developed by our team after the analysis of the client's workflows. Python, PyQt, RASA.