Quest for relevant tags using local interaction networks and visual content

Abstract

Typical tag recommendation systems for photos shared on social networks such as Flickr, use visual content analysis, collaborative filtering or personalization strategies to produce annotations. However, the dependence on manual intervention and the knowledge of sufficient personal preferences coupled with the folksonomic issues limit the scope of these strategies. In this paper, we present a fully automatic and folksonomically scalable tag recommendation model that can recommend tags for a user's photos without an explicit knowledge of the user's personal tagging preferences. The model is learned using the collective tagging behavior of other users in the user's local interaction network, which we believe approximates the user's preferences, at least partially. The tag recommendation model generates content-based annotations and then uses a Naïve Bayes formulation to translate these annotations to a set of folksonomic tags selected from the tags used by the users in the local interaction network. Quantitative and qualitative comparisons with 890 Flickr networks show that this approach is highly useful for tag recommendation in the presence of insufficient information of a user's own preferences.

Tag suggestion on youtube by personalizing content-based auto-annotation

“…To do so, we first evaluate common tag recommendations strategies on a large-scale dataset of YouTube clips where the original user-generated tags are used as ground truth. Here, our results confirm earlier findings [14] that a history-based suggestion of tags offers a simple and strong strategy. Second, as our focus moves towards crowd-sourcing, we introduce a novel approach that extends history-based tag suggestion with a visual content analysis crowd-sourced from YouTube: In our system, two similarity matchings are conducted for the questioned video, once with a large-scale reference dataset of user-tagged content and once with content in the user's personal history.…”

Personalized image annotation via class-specific cross-domain learning

“…For generic image annotation, we compare the proposed cs-WNN model with four most relevant methods, including the visual model [7] that is used for personalization with cross-entropy based learning algorithm (CEL) [1], the weighted nearest neighbor model as described in Eq.…”

“…Then, Sawant et al [7] leveraged users' labeling preferences with the observation that simply annotating a user's image with the tags most frequently used by the same user yielded a much higher accuracy than many generic methods. Moreover, considering that folksonomy or social tagging systems allow users to label their favourite images with free tags or phrases, the work in [8] develop a personalized photo tagging approach to recommend users preferred tags for their newly uploaded photos based on the history information in their social communities.…”

“…So how we can utilize users' image datasets and their labeling histories to learn their annotation profiles is the key point we investigate. Previous work [6][7][8] on this line mostly focuses on learning users' annotation models on their own datasets. However, the problems of cold start for users (i.e.…”

Modeling tagged photos for automatic image annotation

“…In this sense, the approach does not cater to the preferred vocabularies of users. To address this issue, we developed a personalized tagging extension [1]. We proposed a transfer learning model to translate the set of machine annotations to a user's vocabulary using a Naïve Bayes formulation.…”