The paper addresses the challenge of localizing specific moments in videos using natural language descriptions. To achieve this, the authors propose the Moment Context Network (MCN), which integrates both local and global video features with temporal context to accurately identify the start and end points of a moment within a video. The key innovation is the inclusion of temporal endpoint features, which help determine when a moment occurs in the video. To train and evaluate the MCN model, the authors collect the Distinct Describable Moments (DiDeMo) dataset, which consists of over 40,000 pairs of localized video moments and corresponding natural language descriptions. The DiDeMo dataset is unique in that it includes unedited personal videos and diverse visual settings, making it suitable for open-world video moment localization tasks. The authors demonstrate that the MCN model outperforms several baseline methods, highlighting the importance of integrating local and global video features, as well as temporal context, for effective moment localization. The paper also discusses the challenges and future directions in this field, emphasizing the need for better handling of complex sentence structures and generalization to unseen vocabulary.The paper addresses the challenge of localizing specific moments in videos using natural language descriptions. To achieve this, the authors propose the Moment Context Network (MCN), which integrates both local and global video features with temporal context to accurately identify the start and end points of a moment within a video. The key innovation is the inclusion of temporal endpoint features, which help determine when a moment occurs in the video. To train and evaluate the MCN model, the authors collect the Distinct Describable Moments (DiDeMo) dataset, which consists of over 40,000 pairs of localized video moments and corresponding natural language descriptions. The DiDeMo dataset is unique in that it includes unedited personal videos and diverse visual settings, making it suitable for open-world video moment localization tasks. The authors demonstrate that the MCN model outperforms several baseline methods, highlighting the importance of integrating local and global video features, as well as temporal context, for effective moment localization. The paper also discusses the challenges and future directions in this field, emphasizing the need for better handling of complex sentence structures and generalization to unseen vocabulary.