Improved Image Caption Rating - Datasets, Game, and Model.

How well a caption fits an image can be difficult to assess due to the subjective nature of caption quality. What is a good caption? We investigate this problem by focusing on image-caption ratings and by generating high quality datasets from human feedback with gamification. We validate the datasets by showing a higher level of inter-rater agreement, and by using them to train custom machine learning models to predict new ratings. Our approach outperforms previous metrics - the resulting datasets are more easily learned and are of higher quality than other currently available datasets for image-caption rating.

You Described, We Archived: A Rich Audio Description Dataset.

The You Described, We Archived dataset (YuWA) is a collaboration between San Francisco State University and The Smith-Kettlewell Eye Research Institute. It includes audio description (AD) data collected worldwide 2013-2022 through YouDescribe, an accessibility tool for adding audio descriptions to YouTube videos. YouDescribe, a web-based audio description tool along with an iOS viewing app, has a community of 12,000+ average annual visitors, with approximately 3,000 volunteer describers, and has created over 5,500 audio described YouTube videos. Blind and visually impaired (BVI) viewers request videos, which then are saved to a wish list and volunteer audio describers select a video, write a script, record audio clips, and edit clip placement to create an audio description. The AD tracks are stored separately, posted for public view at https://youdescribe.org/ and played together with the YouTube video. The YuWA audio description data paired with the describer and viewer metadata, and collection timeline has a large number of research applications including artificial intelligence, machine learning, sociolinguistics, audio description, video understanding, video retrieval and video-language grounding tasks.

Alzheimer's disease-related overexpression of the cation-dependent mannose 6-phosphate receptor increases Abeta secretion: role for altered lysosomal hydrolase distribution in beta-amyloidogenesis.

Prominent endosomal and lysosomal changes are an invariant feature of neurons in sporadic Alzheimer's disease (AD). These changes include increased levels of lysosomal hydrolases in early endosomes and increased expression of the cation-dependent mannose 6-phosphate receptor (CD-MPR), which is partially localized to early endosomes. To determine whether AD-associated redistribution of lysosomal hydrolases resulting from changes in CD-MPR expression affects amyloid precursor protein (APP) processing, we stably transfected APP-overexpressing murine L cells with human CD-MPR. As controls for these cells, we also expressed CD-MPR trafficking mutants that either localize to the plasma membrane (CD-MPRpm) or to early endosomes (CD-MPRendo). Expression of CD-MPR resulted in a partial redistribution of a representative lysosomal hydrolase, cathepsin D, to early endosomal compartments. Turnover of APP and secretion of sAPPalpha and sAPPbeta were not altered by overexpression of any of the CD-MPR constructs. However, secretion of both human Abeta40 and Abeta42 into the growth media nearly tripled in CD-MPR- and CD-MPRendo-expressing cells when compared with parental or CD-MPRpm-expressing cells. Comparable increases were confirmed for endogenous mouse Abeta40 in L cells expressing these CD-MPR constructs but not overexpressing human APP. These data suggest that redistribution of lysosomal hydrolases to early endocytic compartments mediated by increased expression of the CD-MPR may represent a potentially pathogenic mechanism for accelerating Abeta generation in sporadic AD, where the mechanism of amyloidogenesis is unknown.