7.2 Results
7.2.13 Tag editing
The tagging feature included an administration tool to edit or delete tags. This was used mainly to delete tags. Usually, tags were deleted because they did not correspond to the content of the article. All use of this tool was logged into TagAdministrationLog. It shows the tag, whether it was deleted, excluded or edited, at what time. If the editor edited the tag, the transaction log includes the new tag. An edited or excluded tag could be treated in the administration tool again. A deleted tag was gone.
If an editor deleted an article with tags applied to it, these deleted tags would not be in the log. Thus, the log only contains cases where a tag was treated independently, and the editor wanted to edit or delet the tag itself. Tags that were deleted with articles were most likley deleted for other reasons than the content or other porperties of the tag itself. On the other hand, some tags that were deleted together with the article they were applied to, would have been deleted or edited if the article were not to be deleted.
Because of this, some tags are ‘missing’ in the TagAdministrationLog.
The first months, the tag editing tool was mainly used by the chief editor Øyan. Then the responsibility was handed over to editor E3 (see chapter 9). She was a student who worked part-time for the Danish Cancer Society when she took over the responsibility.
Øyan gave instructions based on his initial experience with editing tags. The other editors E1 and E2 had access to use it the editing tool, but only editor E2 reported that she had edited and/or deleted tags, and only rarely.
Tags that did not relate to the aboutness of the article were most challenging to the editors. From the beginning, these tags were seen as promblematic and irrelevant to the tag collection. First, the editors were hesitant about deleting them also because they knew we were studying the tags and I had stated that I did not want them to delete too many tags. In addition to this, the tag-editing tool that not not give enough information to locate tags to be edited in between the relevant ones. Only tag and URL were visible in the editing tool, not the document text itself. But in cases where it was obvious that the tag was far away from the article content, it was deleted if it was discovered (E-mail from Tor Øyan fr 09.12.2011 13:53). And the editors eventually established experience and routine when editing tags.
8,574 tags were treated in the administration tool. Most of them, 8,384, or 97.8% were deleted. 189, or 2.2%, were edited. In Table 20, the Number of transactions refer to the total number of treated tags, per month. Number of deleted tags show how many of the treated tags were deleted each month. Distinct tags show the number of distinct tags treated each month.
The option to exclude tags, and thus ban this word from being saved as a tag, was used only once. The excluded tag was lort (crap). Tags that were applied later, like fuck lort and fucking lort [crappier] was not excluded, only deleted. Thus, this seems like this option was not really used by the editors. When looking at distinct tags, 7,990 distinct tags were treated. 7,814 were deleted, 186 were edited and one excluded.
When compareing the number of tags treated in the administration tool with the total numbers of tags, 34% of the tags were treated in the administration tool. 33% of all applied tags were deleted.
Month Applied
tags
Number of transactions
Number of deleted tags
Distinct tags November
(from Nov. 8) 992
50 48 43
December
2011 2502
131 128 105
Jaunary 2012 2740
301 297 289
February 2319
2410 2408 2396
March 3287
1234 1230 1220
April 2719
203 168 198
May 3354
1369 1345 1343
June 2412
181 181 172
July 1655
210 205 203
August 2025
671 604 665
September 721
1249 1212 1237
October 38
276 269 270
November 134
35 35 32
December 200
220 220 212
January 2013 94
22 22 21
February 2013 59
12 12 12
Average
535,88 526,13
All year
8574 8384
Table 20 Transactions in the tag editing tool, per month.
Figure 29 visualizes the numbers of deleted tags each month. The numbers vary and it is not clear why. But for the months after the feature change, there were less new tags to delete or otherwise treat. In February, September, October and December 2012 more tags are deleted than applied, thus older tags were treated as well as new ones.
Figure 29 Applied tags and deleted tags per month
7.2.13.1 Categorization of edited and deleted tags
Unfortunately, the TagAdministrationLog did not include the id or the URL of the tag that was deleted or edited. Thus, it is complicated to say which occurrence of a tag was deleted, in cases where a similar tag was applied to many URLs, or many times to the same URL. However, a comparison of the list of tags and the list of deleted and edited ones gives numbers that indicate how often a tag formulation was deleted.
When categorixed, the numbers in each category indicate how editors treated tags in various categories.
98% of all distinct tags from external users were deleted once or more. 70% of all distinct tags from internal users were deleted once or more. Thus, tags from external users were more likley to be deleted than tags from internal users. This corresponds to my expectations that users from internal ip-addresses were able to apply better tags than external users. In addition to the central editorial board of Cancer.dk, approximately 20 other decentralized editors applied tags. And all Cancer.dk employees had access to tag. But the editors who deleted and edited tags, were two of the internal taggers, and this may have influenced their judgement when doing this.
Still, there is a distance between 70% and 98%.
0
Deleted or edited tags
Total number
of tags %
Lay 1 2838 3284 86.42
Neutral 2 8738 11684 74.79
Professional 3 1374 1917 71.67
Hard to categorize 4 1707 1923 88.77
No categorization
needed 5 6183 6272 98.58
Table 21 Distinct tags that were deleted, distributed on lay/professional categories
Professional and neutral vocabulary is more likley not to be deleted than lay vocabulary. 75% of neutral distinct tags and 72% of professional distinct tags were deleted once or more, compared to 86% of lay distinct tags. See Table 21 where tags that were deleted or edited are distributed on the lay/professional categories. The numbers do not explain whether editors were more liberal to lay vocabulary, or maybe tags with lay vocabulary were better. If so, it corresponds to a view that lay users perform better with lay vocabulary.