An associative knowledge network model for interpretable semantic representation of noun context

Abstract Uninterpretability has become the biggest obstacle to wider application of deep neural network, especially in most human–machine interaction scenes. Inspired by powerful associative computing ability human brain system, a novel interpretable semantic representation model noun context, knowledge network model, is proposed. The proposed structure composed only pure relationships without relation label and dynamically generated analysing neighbour between words text, which incremental updating reduction reconstruction strategies can be naturally introduced. Furthermore, method designed for practical problem checking coherence context. In method, learned from text corpus first regarded as background then multilevel contextual coupling degree features given detection document are computed. Finally, location those inconsistent realized using an classification such decision tree. Our sufficient experimental results show that above obtain excellent performance completely reach or even partially exceed obtained latest methods F1 score metric. addition, natural interpretability learning our should extremely valuable than methods. So, this study provides very enlightening idea developing machine methods, tasks writing error detection.