Conversational Agent to Serve as a Personal Companion

Abstract

Conversation agents can act as a personal companion to individuals in a variety of social applications and as proof of concept the same is being done for the domain of Career Counselling. Quarter of a million students in Pakistan are deprived of access to Career Counselling services owing to which they make uninformed decisions about their future at an early stage without any proper understanding or guidance. This decision has a major impact on their future directly. Existing modes of career counselling are restricted to that provided by career counsellor’s face to face and traditional online forms available that do not incorporate all aspects of a student’s life. Due to the inaccessibility of the former across the country and lack of flexibility shown by the latter, a chatbot application is proposed. The goal of the project is to develop a chatbot entitled ‘Sir Isaacbot’ that offers career counselling to individuals in an adaptive manner, considering both internal characteristics such as grades, educational background, abilities as well as external factors like geographical location, job market etc in order to guide them about what career stream is most suitable for them. The application is self-contained target oriented product aimed at advising college bound students about what career stream is most well suited to them. The application mimics human conversation that l converses with the user to gather data about him using a nonlinear approach to maintain the flow of conversation. The chatbot guides the user a series of questions that the user must respond to as accurately as possible. Once enough data is available in the profile, the chatbot will dispense advice. The developers of the project identified a number of internal and external factors that influence decision making towards a career stream using a variety of data gathering techniques including advice from field experts and ensured the questions asked and advice provided was tailored based on those. Data gathered through these techniques was used to train a model for multiple use cases to develop the minimal viable product. The product was able to satisfy all requirements specified in the current scope as verified by testing and was able to justify the initial idea of chatbots being able to aid individuals in their lives. As the nature of the application demands, subsequent training will be carried out continuously so the chatbot can be fine-tuned through an evolutionary process. Though currently text based, the chatbot can be integrated with speech as well as facial feature recognition. Its scope can incorporate additional age groups and factors to support the idea of career counselling being an ongoing process. The architecture used in the current project can be reused to model chatbots that can serve a meaningful purpose in multiple social applications.