SahaayaAI: An Inclusive AI-Driven Women Safety Platform with Complaint Analysis and Early Alert Mechanism

Abstract

Women’s safety is a pressing social issue as it still takes time for help to arrive at the scene of the incident, incidents continue to be under reported and existing mechanisms for safety are not widespread. Despite being presented with a plethora of probable digital solutions, the vast majority are reactionary content and provide relatively little preventative intelligence or analytical desperately required for success at this level. In this paper, we explore a comprehensive AI empowered system for women safety, which combines the processing of emergency alert with smart complaint analysis and spatial risk estimation. The proposed system, named SahaayaAI, further integrates a 3-click SOS-trigger to handle prompt real-time assistive response in consent-based manner and multimodal grievance reporting using texts, speech, and images for making it accessible to visually-impaired / hard of hearing / speech-impaired women. Automatic Speech Recognition is utilized to convert speech complaints into text, which are then processed by applying the Natural Language Processing methods. We use TF-IDF for feature extraction and Naive Bayes for classifier prediction of complaint severity levels, while K-means technique is used to categorize raw spatial complaint data into unsafe/complain prone zones. Experimental analysis conducted using simulated and anonymized datasets shows higher alert accuracy, less emergency response time, and efficient detection of potentially risky areas with respect to traditional women safety applications. These results demonstrate that interpretable AI models can be effectively integrated into emergency workflows to help optimize both real-time response efficiency and preventive safety awareness. Our method is scalable, privacy-preserving, and deployable in real-world city scenarios. Evaluation was performed on an anonymized and simulated set of complaints that were developed to represent real world safety situations.