Tingzhi Wang¹ and Austin Amakye Ansah², ¹USA, ²California State Polytechnic University, USA

MindBalance is a mobile application developed to provide real-time support for individuals experiencing symptoms of schizophrenia. The platform integrates cognitive assessment, auditory hallucination validation, positive affirmations, and psychoeducation into a single accessible tool [6]. It features a semantic fluency test that utilizes a Siamese neural network trained on the AnSim dataset to detect signs of disorganized thinking, and an auditory validation program that compares user-described sounds to real-world audio recordings using a large language model [7]. Positive affirmations are generated through the Google Gemini model to reframe negative thought patterns in real time. Experiments were conducted to validate the system's core functions: a Siamese network assessed semantic coherence between animal name pairs, and a large language model evaluated discrepancies between perceived and recorded sounds. Results demonstrated that lightweight models could accurately detect semantic disorganization and auditory inconsistencies, supporting the feasibility of using AI-driven mobile tools for real-time schizophrenia symptom monitoring [8].

AI, Schizophrenia, Diagnostic, YaMNet, Siamese Neural Network