This paper addresses the problem of identifying rumors in microblogs, focusing on three types of features: content-based, network-based, and microblog-specific memes. The authors propose a retrieval model that achieves over 0.95 in Mean Average Precision (MAP) on a manually annotated dataset of over 10,000 tweets. The dataset is the first large-scale dataset for rumor detection and can help analyze online misinformation and other aspects of microblog conversations. The study shows that the proposed features are effective in identifying both rumors and disinformers. The authors also explore the effectiveness of different features in retrieving rumors and identifying users' beliefs about them. The results show that content-based features are particularly effective in achieving high precision and recall, while network-based features and Twitter-specific memes (such as hashtags and URLs) can also be useful. The study concludes that the proposed framework is effective in identifying rumors and disinformers in microblogs. The authors also highlight the importance of using domain training data to improve the retrieval of specific rumors. The paper presents a general framework that uses statistical models and maximizes a linear function of log-likelihood ratios to retrieve rumorous tweets that match a general query. The authors also show that the proposed features are effective in capturing tweets that show user endorsement, which helps identify disinformers or users that spread false information in online social media. The study contributes to the field of information retrieval and social media analysis by providing a new dataset and framework for detecting rumors in microblogs.This paper addresses the problem of identifying rumors in microblogs, focusing on three types of features: content-based, network-based, and microblog-specific memes. The authors propose a retrieval model that achieves over 0.95 in Mean Average Precision (MAP) on a manually annotated dataset of over 10,000 tweets. The dataset is the first large-scale dataset for rumor detection and can help analyze online misinformation and other aspects of microblog conversations. The study shows that the proposed features are effective in identifying both rumors and disinformers. The authors also explore the effectiveness of different features in retrieving rumors and identifying users' beliefs about them. The results show that content-based features are particularly effective in achieving high precision and recall, while network-based features and Twitter-specific memes (such as hashtags and URLs) can also be useful. The study concludes that the proposed framework is effective in identifying rumors and disinformers in microblogs. The authors also highlight the importance of using domain training data to improve the retrieval of specific rumors. The paper presents a general framework that uses statistical models and maximizes a linear function of log-likelihood ratios to retrieve rumorous tweets that match a general query. The authors also show that the proposed features are effective in capturing tweets that show user endorsement, which helps identify disinformers or users that spread false information in online social media. The study contributes to the field of information retrieval and social media analysis by providing a new dataset and framework for detecting rumors in microblogs.