Event photo mining from Twitter using keyword bursts and image clustering

Kaneko, Takamu; Yanai, Keiji

doi:10.1016/j.neucom.2015.02.081

Cited by 42 publications

(16 citation statements)

References 25 publications

(27 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, Alqhtani et al [13] extract three types of features from images, including Histogram of Oriented Gradients descriptors, Grey-Level Co-occurrence Matrix and color histogram, which are then combined with features extracted from text to train a Support Vector Machine for event detection. In another example, Kaneko and Yanai [11] propose a method to select images from tweet streams for detected events. Specifically, the images are clustered based on densely sampled speeded-up robust features (SURF) and 64-dimensional RGB color histograms.…”

Section: Fusion Of Text and Image For Event Detectionmentioning

confidence: 99%

Image Analysis Enhanced Event Detection from Geo-Tagged Tweet Streams

Han

Karunasekera

Leckie

2020

Advances in Knowledge Discovery and Data Mining

View full text Add to dashboard Cite

Events detected from social media streams often include early signs of accidents, crimes or disasters. Therefore, they can be used by related parties for timely and efficient response. Although significant progress has been made on event detection from tweet streams, most existing methods have not considered the posted images in tweets, which provide richer information than the text, and potentially can be a reliable indicator of whether an event occurs or not. In this paper, we design an event detection algorithm that combines textual, statistical and image information, following an unsupervised machine learning approach. Specifically, the algorithm starts with semantic and statistical analyses to obtain a list of tweet clusters, each of which corresponds to an event candidate, and then performs image analysis to separate events from non-events-a convolutional autoencoder is trained for each cluster as an anomaly detector, where a part of the images are used as the training data and the remaining images are used as the test instances. Our experiments on multiple datasets verify that when an event occurs, the mean reconstruction errors of the training and test images are much closer, compared with the case where the candidate is a non-event cluster. Based on this finding, the algorithm rejects a candidate if the difference is larger than a threshold. Experimental results over millions of tweets demonstrate that this image analysis enhanced approach can significantly increase the precision with minimum impact on the recall.

show abstract

Section: Fusion Of Text and Image For Event Detectionmentioning

confidence: 99%

Image Analysis Enhanced Event Detection from Geo-Tagged Tweet Streams

Han

Karunasekera

Leckie

2020

Advances in Knowledge Discovery and Data Mining

View full text Add to dashboard Cite

show abstract

“…Besides, the similarity of the user's feature attribute is another dimension of relationship strength measurement, which is measured based on the posts of the user pair. Distance formulas are widely used in the similarity measurement, such as Euclidean distance, Manhattan distance [36,37], Chebyshev distance [38,39], Minkowski distance [40,41] and so on. Among those models, the squared Euclidean distance calculation is the most popular for practical application, therefore, it is adopted to measure the similarity of the user's feature attribute.…”

Section: Calculation Of Common Friend Rate and The Similarity Of Usermentioning

confidence: 99%

“…It considers both the importance of searching results and the relative location of searching results. NDCG is a metric that is widely used to evaluate the relationship strength measurement [34,41,44]. Hence, NDCG is chosen as another evaluation metric for performance comparison, which is defined as follows:…”

Section: Evaluation Metricmentioning

confidence: 99%

Research on Relationship Strength under Personalized Recommendation Service

Tao

2020

Sustainability

View full text Add to dashboard Cite

Relationship of users in an online social network can be applied to promote personalized recommendation services. The measurement of relationship strength between user pairs is crucial to analyze the user relationship, which has been developed by many methods. An issue that has not been fully addressed is that the interaction behavior of individuals subjected to the activity field preference and interactive habits will affect interactive behavior. In this paper, the three-way representation of the activity field is given firstly, the contribution weight of the activity filed preferences is measured based on the interactions in the positive and boundary regions. Then, the interaction strength is calculated, integrating the contribution weight of the activity field preference and interactive habit. Finally, user relationship strength is calculated by fusing the interaction strength, common friend rate and similarity of feature attribute. The experimental results show that the proposed method can effectively improve the accuracy of relationship strength calculation.

show abstract

“…The central location of keywords was estimated for geo‐locating the local event detected. Moreover, both visual and textual information from Twitter were analyzed by Kaneko and Yanai () through adding image clustering with keyword burst detection. The local events detected were shown with representative photos on the map.…”

Section: Related Workmentioning

confidence: 99%