Polylingual Multimodal Learning
Published: 2015 September
Herausgeber: Hollmén, Jaakko and Papapetrou, Panagiotis
Buchtitel: Proceedings of the ECMLPKDD 2015 Doctoral Consortium
Seiten: pp. 155-164
Verlag: Aalto University publication series SCIENCE + TECHNOLOGY, 12/2015
The growth of multimedia content on the web raise diverse challenges. Over the last decades various approaches are designed to support search, recommendations, analytics and advertising majorly based on the textual content. Now, due to the overwhelming availability of media (images and videos) require advancements in the existing technologies to leverage multimedia information. Recent progress made in machine learning to foster continuous representations of the text and effectual object detection in videos and images provide new opportunities. In this aspect, my research work aims to leverage data generated from multimedia to support various applications by finding cross-modal semantic similarity. In particular, it aims to compare semantically similar content generated across different channels by jointly modeling two different modalities. Modeling one or more modalities together can be helpful to generate missing modalities and retrieve cross-modal content. My research also extends textual information in multiple languages to support the growth of polylingual content on the web.
Weitere Informationen unter: Link