I finished comparing the MeSH terms with the citations in PubMed/Connotea & CiteULIke

My results:

the number of connotea tags that = MeSH headings  55/240 (22.9166667%)

the number of connotea citations that were also found in PubMed  118/240 (49.1666667%)

the number of citeUlike tags that = MeSH headings  69/240 (28.75%)
the number of citeUlike citations that were also found in PubMed  156/240 (65%)

Next step is to start writing the paper.  I will start with the background information on CiteULike – hard to find. 

7 more Connotea/CiteULike tags to check for connections to PubMed & MeSH. 

Here is the current breakdown:

# of tags in Connotea that match a MeSH term for the citation 39/161 or 24%                                

# of tags in CiteULike that match a MeSH term for the citation 44/170 or 25+%

# of  citations in Connotea that are also found PubMed 81/161 or 50+%

#of citations in CiteULike that are also found in PubMed 105/170 or 61+%

It is looking more & more that tagging in Connotea/CiteULike does not have any relationship to MeSH terms; the taggers are using PubMed to find their information.  (As an aside, many of these citations are for articles that  freely available on PubMed.  That is not part of this research but would make for another paper.  Are social taggers using more free materials?  What are the origins of the taggers?)

Last night, my computer caught a virus. This the first one in 2 years; the depth of the virus made up for it. It took a few false starts, but all is up & cleanly running.

Back to the business at hand. Almost half-way through the duplicate tags –

for the Connotea citations –

36 tags = MeSH terms for 121 citations* = 29.7%.

66 tags had direct PubMed connections for 121 citations* = 54.5%

for the CiteULike citations-

35 tags = MeSH terms for 130 citations = 26.9%

88 tags had direct PubMed connections for 130 citations = 67.6%

*the tag “drosophilia” has only one usable bookmark on Connotea (2 tags total by 2 different taggers).

Emerging pattern – the Connotea/CiteULike tags only match MeSH 1 out of 4 times ; how can it be efficiently used?

After figuring out that not everyone would use the Pub2Conn/Pub2CiteULike tool, I checked the bookmarkings in both sites.  Some of the Connotea citations listed the PMID; some did not.  Some of the remaining Connotea citations did connect directly to PubMed; some did not.  Same for CiteULike.

I have reviewed tags in both sites; the work is going slowly because I am doing it manually.  The number of Connotea tags that equal the citations’ PubMed MeSH terms = 17/71 (23.943662%).  The number of tags with both Connotea/PubMed connection = 39/71 (54.9295775%).  The number of CiteULike tags that equal the citations’ PubMed MeSH terms = 21/80 (26.25%).  The number of tags with both CiteULike/PubMed connections = 56/80 (70%).

Can’t wait to get more information; is this a pattern?

After experimenting with different ways to gather the final bit of information, I finally settled on a process.  Since I could not find any citations that are tagged in PubMed with Connotea/CiteULike, I took another approach; I had always thought the taggers started their searches in PubMed.  If they did not, those citations would not be tagged. 

Plan C:  record the first 10 citations from Connotea & CiteULike (that equals 20) for each tag.  Connotea lists each tag in order from oldest date of tagging to newest date of tagging.  CiteULike does not follow that rule, so it’s first page of tags are not from last month.  I listed the first 10 tags that are indexed in PubMed, no limits put on the PubMed information.  I am trying to discover if social tagging is really a useful social phenomenom. 

Tag/mesh heading “attention” = 8/10 of Connotea citations are indexed with the same term in PubMed,  In addition, this term is the major topic of  all 8.  9/10 of CiteULike citations are indexed with the same term in PubMed.  In addition this term is the major topic of all 9.

Tag/mesh heading “biology” = None of the 20 selected citations are indexed in PubMed with this term. 

I SO wish I had the time to interview the taggers to find out why the particular term was used. 

Up to the point, I have

1.recorded the tags for one week on Connotea & CiteULike

2. determined the duplicate tags (those that appear in both websites on the same day)

3. found the MeSH category for each tag (not just the duplicates)  using the MeSH Browser, 2008 edition

4. decided to use tags that were categorized as MeSH Headings for the rest of the project (=25 tags)

5. used PubMed Limits to find number of citations attached to each tag for 5 different time periods (60/90/180 days, 1 year, any date), that were published in English and are freely full-text available on PubMed (Google doc  entitled “# of articles-duplicate tags-MeSH Headings)

6. downloaded GreaseMonkey to find connotea/citeUlike tagged items on PubMed

Next step – to see how many PubMed citations were tagged by Connotea/CiteULike users…

Step 2 – Done.  I investigated the MeSH Browser for each tag.  Some interesting discoveries: the word “ehealth” is not yet included in MeSH.  Hm, wonder why? 

Step 3 – I started today with 2 tags.  First – I rechecked the tag in PubMed mesh.  (The annotations/scope notes were impossible to find there.)  I used the “send to” box to send the item to PubMed (Search box with AND).  I then added limits – English language, free full text & published in the last 30/60 days, depending on when the items were added to PubMed.

Because I spent time playing around with Step 3, I spent lots of time working on the project.  I found that items in CiteULike are found on Google by searching the article title.  Not so the Connotea tagged items; I still have to search that website.

After spending MANY hours recording data on number of times tags were used this week, I have finally finished step 1.  Final count – 492 tags accumulated; 55 duplicates. 

This week will be spent examining MeSH content for tags.  I will be ready for the meat of this project by next Sunday.

During the past few days, I have hinted that I am a fan of citeUlike.  I have found it’s achilles heel this morning.  After recording the most popular tags for today, the next step is to record the ending numbers for all citeUlike tags.  One problem – the directions are not easily accessible.  I figured it out, after a little sleuthing, but written instructions would be appreciated. 

454  tags; 51 duplicate tags – not as many as I originally thought, but enough to do the work.  I have not read all of the research; what I have stated that although this system is uncontrolled, it ultimately controls itself.  I have found that at CiteULike; those tags are pretty much the same every day.  Connotea, on the other hand, has many new tags everyday; it does not fit the mold.

One more day of gathering data – thank goodness.