This paper presents EDCoW (Event Detection with Clustering of Wavelet-based Signals), a method for detecting real-life events from Twitter. Twitter, as a social media platform, generates a large volume of tweets, many of which are irrelevant or trivial. Event detection in Twitter is challenging due to the high volume of tweets and the need to distinguish between significant events and trivial ones. EDCoW addresses these challenges by using wavelet analysis to build signals for individual words, filtering out trivial words based on their signal autocorrelations, and then clustering the remaining words to detect events using modularity-based graph partitioning. EDCoW first constructs signals for individual words by analyzing the frequency of word appearances over time. It then filters out trivial words by examining their signal autocorrelations. The remaining words are then clustered to form events using a modularity-based graph partitioning technique. This approach allows EDCoW to detect events without requiring a pre-set number of events, making it scalable for large volumes of tweets. The paper also discusses the significance of detected events, which is quantified based on the number of words and the cross correlation among the words. Experimental results show that EDCoW achieves promising performance in detecting events from Twitter data. The method is compared with other event detection approaches, such as topic modeling using Latent Dirichlet Allocation (LDA), and is found to be more effective in distinguishing between significant and trivial events. EDCoW has several advantages, including its ability to automatically determine the number of events and its efficiency in handling large datasets. However, there are areas for improvement, such as incorporating more factors for clustering words and studying the method's performance on larger datasets. The paper concludes that EDCoW is a promising approach for event detection in Twitter, with potential for further development and application.This paper presents EDCoW (Event Detection with Clustering of Wavelet-based Signals), a method for detecting real-life events from Twitter. Twitter, as a social media platform, generates a large volume of tweets, many of which are irrelevant or trivial. Event detection in Twitter is challenging due to the high volume of tweets and the need to distinguish between significant events and trivial ones. EDCoW addresses these challenges by using wavelet analysis to build signals for individual words, filtering out trivial words based on their signal autocorrelations, and then clustering the remaining words to detect events using modularity-based graph partitioning. EDCoW first constructs signals for individual words by analyzing the frequency of word appearances over time. It then filters out trivial words by examining their signal autocorrelations. The remaining words are then clustered to form events using a modularity-based graph partitioning technique. This approach allows EDCoW to detect events without requiring a pre-set number of events, making it scalable for large volumes of tweets. The paper also discusses the significance of detected events, which is quantified based on the number of words and the cross correlation among the words. Experimental results show that EDCoW achieves promising performance in detecting events from Twitter data. The method is compared with other event detection approaches, such as topic modeling using Latent Dirichlet Allocation (LDA), and is found to be more effective in distinguishing between significant and trivial events. EDCoW has several advantages, including its ability to automatically determine the number of events and its efficiency in handling large datasets. However, there are areas for improvement, such as incorporating more factors for clustering words and studying the method's performance on larger datasets. The paper concludes that EDCoW is a promising approach for event detection in Twitter, with potential for further development and application.