Database Source Coverage: Myth and Reality
See the article in  Online Information Review  24(6), 2000. 

 

 Journal lists may not tell you when a title was dropped unceremoniously from the sources covered by a database, like Online from LISA after 1998. 
The search reflects the last update that took place on November 16, 2000 as shown by the ZD=index. It seems that LISA stopped covering Online for good.

Click here to see the whole screenshot!

 

 Its equally popular and respected sister publication, EContent (known until August 1999 as Database) seems to head for the same fate. After the title change EContent was still covered in 1999, but not a single record for articles published in 2000 made it to the LISA database. Given the clout of these journals, and the emphasis of LISA on the topic of information retrieval, dropping these two journals is quite a baffling decision. Then again, there are databases that never covered journals that are essential for the professed database subject coverage as shown later.

Click here to see the whole screenshot!

 

 In November 2000, IBM donated a supercomputer to our Information and Computer Science Department. To find some current papers about supercomputers and supercomputing I turned to the DIALOG Bluesheets database to find which databases mention prominently this subject on their bluesheet. There was a single database of the nearly 500 in DIALOG's stable.

 

 It was the ISA database. It was an extra pleasure to see that not only supercomputers but computer science in general is listed among the prominently mentioned subjects purportedly covered by ISA in its bluesheet that was updated twice in 2000, presumably to keep it current. It seemed also to be appropriate to show to information and computer science students in my DIALOG course.

Click here to see the whole screenshot!

 

 As with most science topics, current information is typically the most interesting for users. There is no denying that I was disappointed that there was not a single record in the ISA database for articles, or conference papers published in 2000 that would have included the word supercomputers in the title, or the descriptor or the identifier field.
Neither in 1999 or 1998.

Remember, this is the database which claims to cover super computing and computer science.

Click here to see the whole screenshot!

 

Truncating the search term finally regaled me with one record about an 1998 article. I wondered if supercomputers are not written about or the new editors of the database ditched the topic and the journals covering it the best.

Click here to see the whole screenshot!

 

 To corroborate that supercomputers are hot topics, and there is an abundance of current literature, I made a search in some databases solely based on my own experience.
They clearly showed that indeed there has been significant literature about supercomputers in the past three years.

Click here to see the whole screenshot!

 

 Luckily, there are various software tools including free and inexpensive ones which help in exploring the real subject and source coverage of databases. 
Through the free DIALOG InfoPro Portal you may search -with some limitations- the excellent Journal Name Finder Database to see the databases that have the best coverage for specific journals, such as Applied Artificial Intelligence.

 

The list clearly ranks the databases by the depth of coverage of the journal. In this example there are only 12 databases that cover this journal, so all of them are listed. 
When a database is split into two or more segments by time period, like INSPEC, MEDLINE, EMBASE, BIOSIS are, they show up twice or more on the list even if the journal coverage did not span the time frame of one of the segments. This makes the list redundant and inconvenient.

 

 

 It is inconvenient because only the top 25 databases are listed in the free journal Name Finder version.
As there is no way to search for exact journal title, or using proximity and positional operators, there can be irrelevant hits. 
If a title is not distinct in itself, as is the case with Artificial Intelligence, other journals that include the same words in their title, also show up. 
Even with these restrictions, the list clearly shows, for example, that Information Science Abstracts does not make the list  for any journals with artificial intelligence in the title even though it is one of its touted subject specialties in more ways than one.

 

A search for a more specific journal title that finds 7 databases shows the  lack of coverage by the ISA database of another important journal in artificial intelligence.

 

 

And  so does the search for AI Communications, another respected journal in artificial intelligence.

Click here to see the whole screenshot!

 

 In some cases the limit of 25 databases is not a problem because after the first few databases the coverage in the rest of the databases is not significant, as is the case with Computers and Artificial Intelligence.

Click here to see the whole screenshot!

 

The same applies to the journal of Artificial Intelligence in Medicine that is relevant both for the artificial intelligence and the medical informatics coverage in a database that touts both topics.

 

 And there are cases when the excellent coverage of a key journal in a free database like MEDLINE drives you away from any of the fee-based databases, and ...

 

 ... accentuates the absurdity of the subject coverage claim of the producer of an expensive database.

 

 For fairness, when searching the Journal Name Finder database and you don't find a database listed for a journal, you should try it in a different format ...

Click here to see the whole screenshot!

 

... even if it is a wrong title for the same journal, and the result tunrs out not to worth the efforts.
At least you learn it for free instead of mucking around in an expensive  database trying to find inaccurate and inconsistent journal name variants.

Click here to see the whole screenshot!

 

 When the limitations of the free options cannot be accepted, you should try the inexpensive version of the Journal Name Finder database with a DialUnit price of $1.25.
This is the case for a title like Information Sciences that the free version cannot handle appropriately. 
But isn't that obvious that a database with the name Information Science Abstracts would cover it very well? No, it is not, and you better check out the variations with subtitles in the browsable index of Journal Name Finder, and pick all the relevant entries to make the search.

Click here to see the whole screenshot!

 

 Even the most discriminating information scientist can't argue that this journal from Elsevier is not enough "information scientish" to match the high standard of admission set by the new editors of ISA. 
But as you will find out, its very short time span of acceptable coverage ended after 1998.

Click here to see the whole screenshot!

 

To get  to the bottom of the matter you may wish to see how this journal was covered across the years. 
By ranking the set retrieved by the journal name variations you find out that it was covered substantially only from 1995 and its coverage stopped in 1998 with a really shallow coverage in that year when the database was taken over.

 

 If you wonder how this coverage compares to that of other, less expensive databases, follow the same procedure to produce the timeline below, illustrating this journal's coverage by the MathSci and  PASCAL databases. 

 

 Another way to quickly determine what key  journals were grossly neglected or not at all covered in a given year is to create an alphabetic index for that year's journal from the database.
A look at the 'I' section of the index shows that in addition to Information Sciences ISA has not added at all records in 2000 (as of December 7) for such journals as IEEE Transactions on Information Theory, Information Systems Management, Information Systems Journal, Information Systems Research, or International Journal on Information and Management - to name a few journals that clearly should be covered in an information science database. 

Click here to see the whole screenshot!

 

 Dropping that many relevant information science journals it is no wonder that the update volume of the database reached an all-time low. By  December, it is reasonable to expect databases to have added 80-85% of the current year's record, especially when it is specifically claimed that the database has no backlog. As of December 7, ISA added 2,027 records for the current year's publications. 
  Such withering away is a warning sign for the potential extinction of a database, a topic that  I discuss  in the December issue of Information World Review (Endangered Database Species) and illustrate it in another  EXTRA piece as a Christmas bonus. This figure is an updated version, reflecting the last update of the database.

Click here to see a larger graph !

Back to the eXTRA menu