School of Science and Technology 科技學院
Computing Programmes 電腦學系
| Programme | Bachelor of Computing with Honours in Internet Technology Bachelor of Science with Honours in Computer Science |
| Supervisor | Dr. Keith Lee |
| Areas | Intelligent Applications |
| Year of Completion | 2023 |
Hidden youth, who avoid social situations and gatherings, make up about 2% of the population and are becoming more common in Hong Kong. The COVID-19 pandemic has made the phenomenon more serious, as people are forced to stay at home. 6PM, an online support team from St. James’ Settlement in Hong Kong, helps hidden youth solve emotional problems and provides online resources such as counseling and gatherings. They have provided a Discord platform for events and to connect with hidden youth.
Hidden youth who stay at home for prolonged periods experience a deterioration of their social and communication skills, leading to disconnection and nervousness. To address this, we partnered with SJS to develop a Discord bot and web application using natural language processing (NLP). The bot can communicate more humanely, helping hidden youth train their social skills and become more confident. The web application reduces analysis and management time, replacing the use of Excel. Our goal is to help hidden youth become more outgoing and sociable, improving their mental health in the long term.
We focus on the following objectives, derived from our functions, to help meet the aim of our goal:
The chatbot’s objectives include bringing informative news and chatting with hidden youth to train their social skills and detect emotions. It can intervene in disputes and reduce social workers’ workload by substituting some of their tasks. A web application is available for social workers to analyze organized events and group chat logs for emotion analysis. Promoting 6PM can be done by reposting information on Discord. The evaluation process involves granting access to a predetermined number of users to evaluate the bot’s functionality and a comprehensive Google form for feedback. Social workers assess the chatbot and web application’s user-friendliness, implementations, and features.
Demonstration Video
Presentation Video
Figure 1: Architecture of Discord Chatbot and web application
To tackle the problems mentioned previously, a Discord chatbot and a web application were designed. The Discord chatbot included basic conversational functions, the ability to detect user emotions, and the ability to join events. The web application included event creation, posting events on Discord, and chatlog analysis. Pycord with the Python programming language was mainly used to develop the Discord chatbot, which could be classified into server-side and client-side. An additional web application was developed using the PHP programming language.
Hosting Discord chatbot
Figure 2: Overview of the Discord Bot’s functionality
Figure 3: Dialog Management Flowchart with NLP
Figure 4: Flowchart of Emotion Detection
Hosting web application:
Figure 5: Overview of the Web Application functionality
Figure 6: Overview of the Web Application functionality
Hosting API:
Figure 7: Sending application confirmation to user
As we were in the prototype stage, Docker containers were used. The advantages of using Docker included:
Functionality Evaluation (6PM feedback)
Meetings were held with 6PM Organization to assess the concept and features of the web application and the Discord chatbot’s overall performance. They were satisfied with the features overall and impressed with the ability to create events, analyze chat, and detect user emotions when social workers were not around. Some additional suggestions were made, such as adding confirmation features to notify users of their event application.
Functionality Evaluation
The objectives of the Discord chatbot and web application were achieved. The chatbot was successful in providing informative news and promoting 6PM, but only partially satisfied the objective of chatting with hidden youth. The web application reduced social workers’ workload, especially in event organization. The improvements suggested in the evaluation were implemented in both the chatbot and web application, which improved the quality of service.
The chatbot’s interactive functions helped hidden youth become more engaged and interactive, which could alleviate the hidden youth phenomenon. The emotion detection function could detect negative messages and send emotional requests to social workers for follow-up or intervention. Users could also join activities posted by social workers from the web application. However, the NLP was limited to available datasets and needed to be updated manually, which could result in a loss of interest in interacting with the chatbot in the long run. Emotion detection was only capable of detecting emotions in simple sentences and could not detect complicated emotions.
To solve these problems, a better emotional model and NLP model should be used. Subscription paid services from Komprehend AI or BytesView could provide better performance of the emotion detection functionality. Subscription paid API services from OpenAI, such as GPT-3, could make extensive conversations and improve engagement and interaction with the chatbot.
The search event function of the web application could be improved by adding a description search to make it more efficient and convenient for social workers. Overall, the improvements suggested in the evaluation enhanced the quality of service of the chatbot and web application, benefiting hidden youth and social workers
In the future, the following work could be done to improve our project:
