מושגי ליבה
Developing a Farsi chatbot to guide users through the Self-Attachment Technique (SAT) and evaluating its effectiveness in enhancing mental health.
תקציר
In this work, a Farsi chatbot was developed to guide users through the Self-Attachment Technique (SAT), a novel psychological technique based on attachment theory. The chatbot uses rule-based and classification-based modules to recommend appropriate SAT exercises. A 12-class emotion recognition module with high accuracy was developed, keeping conversations engaging by retrieving responses from a large dataset. The platform was evaluated in a non-clinical study with positive results, showing engagement and satisfaction among users. The study aimed to assess the feasibility of using the chatbot in human trials and its efficacy in interacting empathetically with users while reducing negative emotions and increasing positive ones.
The research highlighted the importance of digital technologies in supplementing traditional therapy sessions, especially in middle-income countries where access to mental health professionals may be limited. Previous studies have shown that chatbots can effectively ease mental disorders and achieve high levels of user satisfaction. The development of a Farsi-speaking chatbot for mental health interventions is significant, considering the rise in anxiety and depression globally post-COVID-19 pandemic.
The study focused on developing datasets in Farsi to address language resource challenges, sharing three datasets with the research community. The evaluation involved quantitative feedback on empathy, engagement, emotion interpretation, effectiveness at improving emotions, overall performance, SAT impact on well-being, and SAT Teacher satisfaction. Qualitative feedback highlighted themes such as helpfulness, user-friendliness, beneficial exercises, lack of open-endedness, difficulty of exercises, rigid answers, insufficient guidance, accessibility issues, delayed responsiveness, lack of images for engagement, and lack of empathy during conversations.
סטטיסטיקה
A 12-class emotion recognition module with accuracy above 92%.
Results from an evaluation study with N=51 volunteers.
Positive user feedback: 76% found the platform engaging; 73% felt better after interactions; 74% were satisfied with SAT Teacher's performance.
ציטוטים
"Our main contributions are: developing a Farsi chatbot for mental health enhancement."
"We collected monolingual Farsi datasets to move towards making Farsi a high-resource language."
"The study showed promise in applying chatbot technologies for psychotherapeutic interventions."