In the last four years, the ever-growing usage of social media in Ethiopia has fueled the country’s problem with the peaceful coexistence of its people. Illegitimate social media usage has played a big role in widening the gap between people. So that it is difficult to manually identify the emotions of a million users and aggregate them towards a rapid and efficient decision, it is quite a challenging task due to the rapid growth of Amharic language usage in social media. As such, there is a necessity to develop an intelligent system that automatically detects such negative and non-negative contents by filtering them into socially, religiously, and politically relevant categories and filtering Toxic online contents. The researchers utilized comment exporter software to add 29962 comments collected from social media to the dataset. Common evaluation metrics such as accuracy, recall, F1 score, and precision were used to measure our proposed model's performance. Finally, this study with four categories of classification (CNN, GRU, LSTM, and Bi-LSTM) based on the experiments of Amharic text negative emotion detection and filtering classification models has an accuracy of 83%, 50%, 84%, and 86%, respectively. In the experiments conducted, Bi-LSTM achieved the highest accuracy of 86%. Therefore, different deep learning methods are used by using social media comments of users to evaluate them and perform better for Amharic text comment emotion detection and filtering negative emotion-bearing content. The target of this paper is to use the concept of automatic detection and filtering of negative emotion-bearing contents from social media in Amharic text using sentiment analysis and a study of deep learning algorithms. Keywords –: Emotion, Negative Emotion, Emotion Detection, social media filtering sentiment analysis, deep learning