Start of main content

Inspec classification searching cascaded codes

When searching on Inspec it is sometimes desirable to restrict an answer set to a particular subject area within the database. This is especially true when a search is conducted using free language keywords. It is often the case that a word can have more than one meaning, and searching for that word can lead to the retrieval of many unwanted records which are not from the desired subject area. The Inspec classification can help minimise this problem.

The Inspec classification provides a hierarchical breakdown of the subjects covered in the database. It is updated annually and is incorporated many of the platforms that Inspec is available on. It can also be obtained separately from Inspec either in book form (2004 revision) or electronically.

Example search

Consider a search for articles discussing the use of iteration as a software technique. Simply searching the single keyword "iteration" is not appropriate as the word is not exclusively used to describe concepts within the subject area of computing. Many records, most of which will not be relevant, will be retrieved.

 Additional keywords such as “program…” can be added in order to make the search more focused, but care should be taken when doing this as the word “Program” like “iteration” has alternative meanings in a variety of subject fields (e.g. computing, television etc.). Although using these words will result in records being retrieved from the desired subject area of software there is a high chance that unwanted records from other technological areas such as television will also be included in the answer set.

Start by searching for the word “iteration” in the title field of records.

?S ITERATION/TIS1 2239 ITERATION/TI?T Sl/TI,CC/1-3

1/TI,CC/1

Title: Efficient corrector iteration for DAE time integration in multibody dynamics

Class Codes: E2200 (Mechanical components, systems and devices); E0210L (Numerical analysis)

1/TI,CC/2

Title: Acceleration and filtering in the generalized Landweber iteration using a variable shaping matrix

Class Codes: A0765G (IR spectroscopy and spectrometers); A7820D (Optical constants and parameters (condensed matter)); A0260 (Numerical approximation and analysis)

1/TI,CC/3

Title: Iteration procedure for the capacity of discrete memoryless channel

Class Codes: B0240 (Probability and statistics); B0290F (Interpolation and function approximation (numerical analysis)); B6110 (Information theory)

The records above are typical of the 2239 that were retrieved. They come from different subject areas. Looking at the information contained in the titles it is apparent that records 1 and 2 are irrelevant to the scope of my search. Record 3 however, is relevant.

The classification codes on this record suggest that that  C6110 would be an appropriate code to use in conjunction with the keyword “iteration” for restricting the search to the relevant subject area of the database and make the search more focused. Additional codes could be found by reviewing the codes on other records or by using the online classification list available on Dialog. This is accessed by using the EXPAND command:

?E CC=C6000 

Ref

Items

RT

Index-term

E1 193389   CC=C6
E2 420   CC=C60
E3 420 1 *CC=C6000 Computer software
E4 192991    
E5 2616 CC=C6100 Software techniques & systems
E6 27775 19 CC=C6110 Sustems analysis & programming
E7 12320 13 CC=C110B Software engineering techniques
E8 2129 6 CC=C6110J Object-oriented programming
E9 1451 3 CC=CC6110L Logic programming
E10 2294 1 CC=C6110P Parallel programming
E11 14315 13 CC=C6115 Programming support
E12 17730 11

CC=C6120 File Organisation

Expanding on C6000 in place of the specific code C6110 ensures that all of the classifications in the relevant hierarchy are displayed. Reviewing the codes displayed show that any record which has been classified with a code beginning C61… may be of interest.

These codes can be searched in a number of ways; search each code individually and combine into one answer set using the Boolean logic “OR” operator (i.e. S CC=(C6100 OR C6110 OR …..OR C6120)); use truncation (i.e. S CC=C61?); select the codes I want from the expand list (i.e. (SEL E1 – E12) then (S E1 – E2)); or use the cascaded code feature.

There are problems with the first 3 options. A very long search string would be required for option 1 and it is possible that an overflow might occur; the time required for the query to process can be quite long on option 2; and option 3 requires a 2 stage search strategy.

Option 4 is the best solution. The cascade feature on Dialog allows an entire hierarchy to be searched without the use truncation. It circumvents the problems associated with using truncation, takes much less time to complete, and requires only a simple strategy:

?S Sl AND CC=C61

1027 Si

192991 CC=C61

S2 70 Sl AND CC=C61

?T S2/6/1-4

 

2/6/1

04400066 Inspec Abstract Number: C9306-7310-017

Title: Graph algorithms=iteration+data structures? The structure of graph algorithms and a corresponding style of programming

 

2/6/2

04342491 Inspec Abstract Number: C9303-4240-052

Title: Improving the results of static analyses of programs by local decreasing iteration

 

2/6/3

04340143 Inspec Abstract Number: C9303-4240-045

Title: Bounded fixed-point iteration

 

2/6/4

04318678 Inspec Abstract Number: C9302-615OC-032

Title: Recursion vs. iteration in Prolog

There are various levels of the Classification where cascade searching can be applied. Truncation is only required at the lowest level. A table of levels is below:

Classification Section

Dialog

C (all Computers & Control) S CC=C
C6 (Computer software) S CC=C6
C61 (Software techniques...) S CC=C61
C6110 without subsections S CC=C6110
C6110 plus subsections S CC=C6110?
C6110L (single subsection) S CC=C6110L