Flickr dataset:
We provide a weighted co-occurrence network between Flickr tags together with the corresponding tag frequencies. This network is resulting from the tags (free words) co-appearing on photos in rather long list of queries from Flickr. In the preparation of the network we kept only English nouns and took into account co-occurrences only if they were present on photos from at least 10 different users.
Reference
G. Tibély et al:
Extracting tag hierarchies
PLoS ONE 8(12): e84133 (2013).
Files:
| File name | Description | Format | Size |
|---|---|---|---|
| Flickr_co-occurrence_net.zip | Co-occurrence network of tags on Flickr photos | compressed plain text file 1st. and 2nd. columns: co-appearing tag ids 3d. column: number of co-occurrences. |
5.5Mb |
| Flickr_tag_frequencies.zip | Frequency of Flickr tags | compressed plain text file 1st. column: tag id, 2nd column: number of photos. |
0.1Mb |
| Flickr_tag_names.zip | Tag names | compressed plain text file 1st. column: tag id, 2nd column: name. |
1.3Mb |
| Flickr_dataset.zip | All files in the Flickr dataset as a zip archive | 6.9Mb |
|---|
Note:
Each file header contains instructions for processing the data with the
Hierarchy Extracting Algorithms
Note2:
We provide a smaller version of the dataset, where you may get results faster