Abstract:

The current article presents the chatbot service within the CareProfSys system, which is used to find out information about professions in the European classification of professions: it serves as an experienced advisor for people seeking guidance in their career search, being useful for two distinct types of users, each with specific requirements. The first type targets aspiring learners, e.g. high school students or students who want to practice a job related to their field of study. The chatbot provides details about universities found in different cities across the country and admission requirements, helping users make informed choices about their educational path. The second type of users is those who want to make a career change. Our conversational agent is an important source for these individuals, providing useful information on topics such as salaries and retraining requirements. In addition, it identifies companies that could hire people in the desired field and provides guidance for securing internships. Through relevant and useful guidance, the chatbot within CareProfSys supports these mature users in their professional development, as well as in the desired career change, to reach the professions recommended by the system. The chatbot has multilingual interaction capability, suitable for discussions in both Romanian and English.

The conversational agent built by us is based on the Pandorabots platform: the user enters the chatbot platform, chooses the conversation language (English or Romanian) and engages in a conversation; the replies received from the bot conform to the rules previously created using the tag-based Artificial Intelligence Markup Language (AIML). A dialogue takes place according to the following flow: after entering some necessary information, the user asks questions, and the answers received from the chatbot are based on templates developed in AIML files based on which the Pandorabots platform works. If the questions are not found in the templates, an exception is thrown and a friendly message is displayed asking the user to enter another question.

AIML relies on pattern matching as its main mechanism and harnesses the power of recursion. AIML files define templates to be found in the questions asked and the corresponding answers. These patterns act as triggers for the chatbot to identify user inputs and appropriate responses are generated accordingly. AIML reproduces the natural writing style of humans. Recursion allows AIML to handle complicated conversational flows and generates responses dynamically. It empowers the chatbot to refer to its previous responses within a new response, leading to a continuous loop of pattern matching and template generation. This recursive approach improves AIML's ability to address various conversational scenarios while maintaining context throughout the interaction.

For creation, testing and optimization, we used Pandorabots, but for the integration of the presented chatbot into the CareProfSys system, we used the React JS and Node JS technology stack. Thus, we created a web interface to our chatbot that can be accessed by the main CareProfSys web platform via a RESTful API.

The conversation agent is presented in the context of various chatbot frameworks and applications. Preliminary testing results are also available, after an experiment with 27 highschool students who played with CareProfSys during a summer school.

Keywords:

Chatbot, conversational agent, career guidance, student support.