Due to globalization technology growth from day to day rapidly and a number of textual information is available in different platforms. An event is a situation that described in different arguments and event modeling is a process of automatic information identification, extraction, and representing from unstructured text automatically. Event modeling is used to identify event instances from a text and to discover event-related information in terms of time, locations, agents that participate in events, methods, and precedence-follows causal connection. However, a lot of information was available and the need for information also increased by individuals, governmental and non-governmental agencies but due to the existence of a large volume of documents certain systems or machines need to automatically detect, extract, and represent a piece of information to reduce the complexity of text understanding by users/machines. The previously conducted Amharic event modeling was proposed event extraction and representation model using machine learning algorithms to recognize events and hand-crafted rules for extracting events with arguments but in both approaches feature engineering is challenging and they are not consider all event descriptive event elements. Therefore, in this study, we have proposed an event modeling to recognize events and to extract event element in terms of 5W1H dimensional arguments using deep learning from a manually collected Amharic Bible book dataset. In this study we have compared the CNN, BiLSTM, and the combined CNN with BiLSTM neural network approaches due to their features extraction capability and appropriate for sequential inputs. To provide the learned representation for texts and to assess the effects of word embedding techniques on deep neural networks by comparing between Word2Vec (CBOW) and the default Keras supported embedding methods. Confusion matrix, precision, recall, and f1 metrics are used for evaluating the proposed model. In this study BiLSTM based event identification and event argument extraction has better performance with the accuracy of 94.6% for Word2Vec based and 98.4% for Keras supported embedding based event identification and 86.8% for event argument extraction. Key Words: Event modeling, Event element, Event identification, Event extraction, 5W1H event argument