Tag refinement in an image folksonomy using visual similarity and tag co-occurrence statistics

Lee, Sihyoung; Neve, Wesley De; Ro, Yong Man

doi:10.1016/j.image.2010.10.002

Cited by 24 publications

(16 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We evaluated the effectiveness of image tag refinement by adopting the noise level (N L) metric [9] [15], which represents the proportion of noisy tags in the set of user-supplied tags of an image folksonomy. When N L is close to one, the number of noisy tags in a folksonomy is high.…”

Section: Methodsmentioning

confidence: 99%

“…The authors of [15] formulate the problem of tag relevance estimation as a maximum a posteriori (MAP) problem. Given a seed image, the proposed approach computes a posteriori probability for each tag associated with a seed image, taking advantage of the observation that the Euclidean distance between folksonomy images that have been annotated with the same tag follows a Gaussian distribution in feature space.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Visually weighted neighbor voting for image tag relevance learning

Lee

Neve

2013

Multimed Tools Appl

Self Cite

View full text Add to dashboard Cite

The presence of non-relevant tags in image folksonomies hampers the effective organization and retrieval of user-contributed images. In this paper, we propose to learn the relevance of user-supplied tags by means of visually weighted neighbor voting, a variant of the popular baseline neighbor voting algorithm proposed by Xirong Li et al. in 2009. To gain insight into the effectiveness of baseline and visually weighted neighbor voting, we qualitatively analyze the difference in tag relevance when using a different number of neighbors, for both tags relevant and tags not relevant to the content of a seed image. Our qualitative analysis shows that tag relevance values computed by means of visually weighted neighbor voting are more stable and representative than tag relevance values computed by means of baseline neighbor voting. This is quantitatively confirmed through extensive experimentation with MIRFLICKR-25000, studying the variation of tag relevance values as a function of the number of neighbors used (for both tags relevant and tags not relevant with respect to the content of a seed image), as well as the influence of tag relevance learning on the effectiveness of image tag refinement, tag-based image retrieval, and image tag recommendation.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Visually weighted neighbor voting for image tag relevance learning

Lee

Neve

2013

Multimed Tools Appl

Self Cite

View full text Add to dashboard Cite

show abstract

“…Lee at al. propose a tag refinement technique that aims at differentiating noisy tag assignments from correct tag assignments (Lee et al, 2010). Each tag is assigned a probability of being noisy based on the visual similarity of the images and tag co-occurrence statistics.…”

Section: Related Workmentioning

confidence: 99%

Topical Video Search: Analysing Video Concept Annotation through Crowdsourcing Games

Gligorov

Hildebrand

Ossenbruggen

et al. 2017

HCj

View full text Add to dashboard Cite

Games with a purpose (GWAPs) are increasingly used in audio-visual collections as a mechanism for annotating videos through tagging. One such GWAP is Waisda?, a video labeling game where players tag streaming video and win points by reaching consensus on tags with other players. The open-ended and unconstrained manner of tagging in the fast-paced setting of the game has fundamental impact on the resulting tags. Consequently, the Waisda? tags predominately describe visual objects and rarely refer to the topics of the videos. In a previous study [ECIR 2013, 50-61], Gligorov at al. showed that the Waisda? tags are effective in finding video segments that depict a specific entity (person, object, etc.) of interest. This study evaluates the performance of the game tags for retrieval of videos that are about a given topic. To this end, we setup a Cranfield-style experiment for which we use an evaluation dataset that consists of: (i) a collection of videos tagged in Waisda?, (ii) a set of queries derived from real-life query logs, and (iii) relevance judgements. While we reuse the collection of videos and the set of queries compiled by Gligorov at al., we designed the set of relevance judgments specifically for this study. The novelty aspects of this paper are as follows. Our results demonstrate that the raw, unprocessed game tags are not well suited for retrieving videos based on topic. We perform a qualitative analysis of the search results which reveals that this is mainly caused by the presence of general tags which refer to visual objects unrelated to the topics of the video. Thus, we characterize the quality of the game tags as topical annotations in order to detect and filter out the non-topical ones. We explore several features of the game tags which could serve as an indication of their quality as topical descriptors. Our results show that after filtering, the game tags perform equally well compared to the manually crafted metadata when it comes to accessing the videos based on topic. An important consequence of this finding is that tagging games can provide a cost-effective alternative in situations when manual annotation by professionals is too costly.

show abstract

“…Normalization in Asymmetric takes place by using frequency of one of the tags [65,100] as shown in Eq 8.…”

Section: Co-occurring Tagsmentioning

confidence: 99%

Semantics discovery in social tagging systems: A review

Jabeen

Khusro

Majid

et al. 2014

Multimed Tools Appl

View full text Add to dashboard Cite

Web 2.0 has brought many collaborative and novel applications which transformed the web as a medium and resulted in its exponential growth. Tagging systems are one of these killer applications. Tags are in free-form but represent the link between objective information and users' cognitive information. However, tags have ambiguity problem reducing precision. Hence search and retrieval pose a challenge on folksonomy systems which have flat, unstructured, non-hierarchical organization with unsupervised vocabulary. We present a brief survey of different approaches for adding semantics in folksonomies thus bringing structure and precision in search and navigation. We did comparative analysis to estimate the significance of each source of semantics. Then, we have categorized the approaches in a systematic way and summarized the feature set support. Based on the survey we end up with recommendations. Our survey and conclusion will prove to be relevant and beneficial for engineers and designers aiming to design and maintain well structured folksonomy with precise search and navigation results.

show abstract

Tag refinement in an image folksonomy using visual similarity and tag co-occurrence statistics

Cited by 24 publications

References 36 publications

Visually weighted neighbor voting for image tag relevance learning

Visually weighted neighbor voting for image tag relevance learning

Topical Video Search: Analysing Video Concept Annotation through Crowdsourcing Games

Semantics discovery in social tagging systems: A review

Contact Info

Product

Resources

About