How to use your library

(and the WWW)

effectively

Sally Jo Cunningham

(sallyjo@cs.waikato.ac.nz)

email me if you have trouble doing your literature survey

Remember, six months in the lab can save you a day in the library

-old reference librarian proverb

Most trustworthy sources are refereed

• journals
• conference proceedings

But working papers (technical reports) can be useful

• often longer, with more detail
  • can be more cutting-edge

Related item searches: how do I find another good article like this one?

Start with one good article

finding similar articles published BEFORE it:

• look in the article’s references

finding similar articles published AFTER it:

• use the citation indexes
• CMCI CompuMath Citation Index
• Science (SciCite),
• Social Sciences citation index, etc.

finding similar articles published at the SAME TIME:

• check the article’s subject headings,
• significant terms, etc.
• run a new search for that same term,
• heading, etc. in the ACM guide,
• CompuMath, etc.

Formal Reference Sources in Computer Science

ACM Guide to Computing Literature (book format)

CMCI CompuMath Citation Index (book)

CDROMs: remember, they’re not CS- specific!!!

ABI, SciCite, Eric, Legal Track, …

Online databases

• UnCover http://uncweb.carl.org/
• FirstSearch (in library, $1.70 PER SEARCH)
  (note: FirstSearch no longer includes INSPEC)
• Current Contents http://gateway.ovid.com/

Internet technical report servers

• http://www.nzdl.org/

other (smaller) commercial databases exist. Check the library links page off http://www2.waikato.ac.nz/library/

(particularly the TRIAL DATABASES)

known item search

(title, title keyword, author)

The idea: you know that you want an article with title XXX, or that author YYY has done something interesting

The technique: look in a "fielded" source such as UnCover, ACM Guide, etc.

Remember to use the correct syntax

• can you type in any words in the title, or do they have to be the first words in the title in exact order?
• Do you type in Smith, J or J Smith?

The dreaded subject search

also uses the ACM Guide, CompuMath, etc.

Most search engines/digital libraries don’t have subject headings. How do I figure out what terms to search with?

keyword searches: will match any word in the document surrogate (title, title+abstract, etc.)

• what is the document surrogate for this document set?

Remember:

• the smaller the document surrogate, the more difficult it is to find the "right" term to match a document.
• overly large document surrogates (for example, the entire document) make it harder to eliminate irrelevant documents from your search results

Tips for finding the right words: remember to try

phrase searching, if available ("finite state machine") ;

synonyms: the word you think of may not be the one the author used

truncation: computers/computing/computed/compute/computer

woman/women

abbreviations: AI/artificial intelligence

alternative spellings: visualization/visualisation

more general terms: machine learning/AI

more specific terms: AI/machine learning/neural nets/logic programming/search spaces/...

Then try:

Getting a friend to look

Waiting a week or so, and looking again yourself

• Asking colleages about accepted terminology in your field

Ex: "diffusion of innovation" vs "dissemination of invention"

Using subject headings

subject headings come from a thesaurus, or list of

"legal" headings (ex: Library of Congress Subject

Headings)

headings assigned manually by a cataloger

relatively few headings assigned (1-3 for a book)

good for capturing:

• what the document is most "about" (instead of picking up any term that’s present, like keyword search);
• subject that the document is about, where the term doesn’t appear in the document (ex: when the term is invented after the document is published, like "AIDS"; when the document is about "AI", but talks about machine learning)

must locate the catalog headings that most closely match your topic

must use EXACTLY that heading

problem: items indexed under the most general term are not usually also under the specific topic, and vice-versa

computer aided design

computer-aided design, Mathematics

computer-aided design, periodicals

Ranked vs Boolean searching

Boolean: uses AND, OR, NOT operators

• "machine learning" and text
• machine and learning and text
• returns all documents exactly meeting the search criteria

Ranked:

• returns documents matching at least one of the search terms
  • machine learning text

sorts ("ranks") returned documents by how well the documents match the query (how many of search terms matched; how many times each term appears in the document; …)

Boolean searching can be very exact, particularly using AND and NOT.

Ranked searching is better for getting large numbers of potentially relevant documents

General things to worry about:

• use a general topic database (UnCover) or a very specific one (LibraryLit)?
• what’s the time coverage of my source?
  (many online databases only go back a few years)
• what’s the search syntax on this system?
  (can I do phrase searching? Boolean searching?)
• what exactly is indexed, anyway?
  (title and author? Abstract? Full text?)

Other useful techniques: