Pursuing the holy grail by interrelating user intentions and bag of visual words to perform retrieval adaptation

When searching for visual content in large image repositories users face a serious problem. Due to the large amount of resources to tap, they cannot find the content, which fits their information need best. The information need is typically based on the user's intention or the goal s/he wants to achieve and is expressed by the user's search behavior. For instance, if someone wants to buy a bicycle, the information need - find pictures of bicycles - is based on the intention to buy a bicycle. In this paper we consider users with a more specific and less specific, hence vague intent. We apply retrieval adaptation in terms of the degree of intentionality using hard and soft assignment techniques of the well known Bag of Visual Word(BoVW) retrieval method. Our evaluations on two common data sets reveal interesting results, which pose the potential of our approach.