Figure 6 - uploaded by Jiyi Li
Content may be subject to copyright.
Contexts in source publication
Context 1
... show three detailed examples in Figure 6. In the example of Figure 6.(a), the affective information of the image includes "anticipation" which is annotated by our system. ...
Context 2
... show three detailed examples in Figure 6. In the example of Figure 6.(a), the affective information of the image includes "anticipation" which is annotated by our system. On the contrary, the emotion words such as "sadness" and "amazement" that are annotated by MVSO are not prominent in the image. ...
Context 3
... the contrary, the emotion words such as "sadness" and "amazement" that are annotated by MVSO are not prominent in the image. In the example of Figure 6.(b), the emotional labels collected by our system such as "apprehension" can better describe the affective information in this image. On the contrary, the emotion labels annotated by MVSO, Figure 5: The average accuracy and inaccuracy of the emotion words in the top-í µí± (í µí± ≤ 6) emotional labels by our systems and sub_MVSO. ...
Context 4
... 3 such as "ecstasy" and "joy" are not reflected in the image. In the example of Figure 6.(c), the affective information of the image includes "pensiveness" which is annotated by our system. On the contrary, the emotion words such as "acceptance" and "joy" that are annotated by MVSO are not prominent in the image. ...
Context 5
... the contrary, the emotion words such as "acceptance" and "joy" that are annotated by MVSO are not prominent in the image. Furthermore, for the examples in Figure 6, MVSO always uses the normal emotion words with high frequency in the corpus, such as "joy" and "sadness"; in contrast, our system can utilize diverse emotion words to annotate the image. It verifies that our system can collect diverse and reliable emotional labels. ...
Citations
When collecting answers from crowds, if there are many instances, each worker can only provide the answers to a small subset of the instances, and the instance-worker answer matrix is thus sparse. The solutions for improving the quality of crowd answers such as answer aggregation are usually proposed in an unsupervised fashion. In this paper, for enhancing the quality of crowd answers used for inferring true answers, we propose a solution with a self-supervised fashion to effectively learn the potential information in the sparse crowd answers. We propose a method named CrowdLR which first learns rich instance and worker representations from the crowd answers based on two types of self-supervised signals. We create a multi-task model with a Siamese structure to learn two classification tasks for two self-supervised signals in one framework. We then utilize the learned representations to complete the answers to fill the missing answers, and can utilize the answer aggregation methods to the complete answers. The experimental results based on real datasets show that our approach can effectively learn the representations from crowd answers and improve the performance of answer aggregation especially when the crowd answers are sparse.
