GCN-FP — Graph Convolution Network-Assisted Natural Language Feature Processing for Social Computing Systems

Social computing systems rely on environmental and behavioral inputs for providing fair processing in a wide range of application support. The input is fetched as text, audio, observation, etc., in which natural language processing is being applied in recent days. Voice-based actuation and processing in these systems result in uncertain events, increasing computational tardiness. This paper introduces a graph convolutional network-based feature processing (GCN-FP) model for addressing the above-mentioned issue. The network augments the voice/audio input features like a connected cyclic graph for succeeding computations. In this process, the training and recurrent graph mapping and disconnection are performed using the certainty factor. This certainty factor relies on the mapped and exhausted features observed in a single graph iteration. Based on the single mapping, decisions for social computing systems are provided, with controlled errors. The features that are not involved in the cyclic process or remain unmapped are identified as errors, and hence, they are included in the training part. Therefore, the proposed scheme achieves better recommendation accuracy, processing ratio, computation complexity, and uncertainty factor.