Deep Neural Networks achieve state-of-the-art performances in several semantic NLP tasks but lack of explanation capabilities as for the limited interpretability of the underlying acquired models. In other words, tracing back causal connections between the linguistic properties of an input instance and the produced classification is not possible. In this paper, we propose to apply Layerwise Relevance Propagation over linguistically motivated neural architectures, namely Kernel-based Deep Architectures (KDA), to guide argumentations and explanation inferences. In this way, decisions provided by a KDA can be linked to the semantics of input examples, used to linguistically motivate the network output.
Croce, D., Rossini, D., Basili, R. (2018). On the readability of deep learning models: The role of Kernel-based deep architectures. In CEUR Workshop Proceedings. CEUR-WS.
On the readability of deep learning models: The role of Kernel-based deep architectures
Croce, Danilo;Basili, Roberto
2018-12-12
Abstract
Deep Neural Networks achieve state-of-the-art performances in several semantic NLP tasks but lack of explanation capabilities as for the limited interpretability of the underlying acquired models. In other words, tracing back causal connections between the linguistic properties of an input instance and the produced classification is not possible. In this paper, we propose to apply Layerwise Relevance Propagation over linguistically motivated neural architectures, namely Kernel-based Deep Architectures (KDA), to guide argumentations and explanation inferences. In this way, decisions provided by a KDA can be linked to the semantics of input examples, used to linguistically motivate the network output.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.