Lecture II Database Structure, Organization and Selection

1
Lecture II - Database Structure, Organization and
Selection

• Review - Database, Producer, Vendor???
• Online Industry
• Effective Searcher Characteristics
• Online Timeline

2
Database Selection

• Single most important factor in performing a good
  search
  
• Requires minimal understanding of database
  structure
  
• Practical v. Technical level approach

3
Basic Database Groups

• Bibliographic - Consists of a citation to a
  published article, book, or report, with a
  summary (or abstract) and indexing terms
  describing the publication.
• Non-Bibliographic - Often serves as the final
  source, usually. Provides actual answers.
  Fulltext. Financial reports. Searching for actual
  piece of information.

4
Types of Databases

• Bibliographic
• Full-Text- Complete text plus basic citation
• Numeric - Contains statistical data, such as
  company financials, demographics, etc.
  
• Directory - Provides succinct information about a
  company or product. Ex. - company name, address,
  phone number, sales figures, names of officers,
  etc.

5
Basic Database Terminology

• Database - Collection of items or entities
• Unit Record - Discrete representation of
  individual items in the database. Card in library
  catalog is UR representing a book. Also referred
  to as an index record. Surrogate of an entity.

6
Unit Record

• Subdivided into different fields/paragraphs
• Each has a label (field) and may be searchable or
  only displayable
  
• Vendor takes tapes, then formats the UR to fit
  the systems protocol. (BIOSIS - sep. YR field
  may create unique tags)
  
• Sub-fields such as journal name, city of
  publication in PU field, or subheadings.

7
Unit Record (contd)

• Vendor may take 6 mo. to 1 year to mount a new
  database
  
• How UR is formatted explained in documentation
  produced by vendor
  
• MUST UNDERSTAND THE UR from both vendors
  producers view to search database effectively.
  Fields present and searchable. More than search
  engine!

8
Linear File

• Set of index records each record describes one
  item, arranged in an order based on the values of
  one or more attributes
  
• Entered by accession number, highest number the
  record input most recently
  
• Linear file of card catalog the librarys
  accession list
  
• Linear file of a book pages in numerical order

9
Inverted File

• Created from the linear file
• Already familiar - card catalog has two of them
  author/title index and subject index index in
  back of a book
  
• Make information accessible and useable
• Creation enables searching by subject, author,
  page number, or any entity deemed searchable

10
Inverted File (contd)

• When performing a search, system does not have to
  search each individual record in database for
  your terms
  
• Quickly searches the inverted file and find all
  documents that had a particular entity
  
• Usually referred to as the dictionary file or the
  basic index
  
• Varies from system to system

11
Dictionary/Inverted File or Basic Index

• Set of records created from a linear file.
  Typically, lists of accession numbers where each
  AN is associated with a different entity. Ex.-
  The index at back of book.
• Makes the information accessible/useable
• Variable by system

12
Standardization

• Reputable vendors attempt to standardize unit
  records (tho they didnt start that way) for the
  fields that can be
  
• PY or AU fields are susceptible to variations in
  the way they have been loaded
  
• Not always vendors fault but can be producers
  as well
  
• Impossible to become expert searcher in all
  systems or all databases

13
Why is there no standardization?

• Complex problem
• Vendors that are essentially competitors have no
  reason to cooperate
  
• Producers often vying to be the source of
  information in a subject area and arent about to
  get together with a rival
  
• Will continue to be a problem

14
Parsing

• Separating and sorting operations performed by a
  search service (vendor) on a given data field
  when inverted index is prepared from the linear
  file
• Vendor determines items that are searched as one
  unit
  
• Multi-word descriptor or an entire journal name
  may be considerd one word in the dictionary file

15
Parsing (contd)

• To flexibility and comprehensiveness, vendor
  may elect to enter field elements into the
  dictionary file more than one way
  
• Each searchable filed is indexed (or parsed) by
  words or phrases being extracted and entered into
  an alphabetical list so they can be searched
  separately
• Varies from database to database and from vendor
  to vendor. Vendor determines items to be searched
  as a single unit

16
Field may be parsed into

• Single words only
• Multiple-word phrases only
• Both single words and multiple-word phrases
• Word fragments (such as segments of chemical
  compounds)
  

17
Parsing Rules

• Double-Posting - Process of entering
  multiple-word search terms both as one unit and
  as individual words
  
• Binding - Hyphenating or otherwise parsing
  multi-unit search phrases so that they are
  searched as one word

18
Parsing (contd)

• Double-Posting - Consider the descriptor
  computer-assisted instruction could be entered
  as one searchable with hyphens between the words
  and as three separate terms. Ex.-computer-assiste
  d-instruction as one unit computer, assisted,
  instruction as three individual words.

19
Binding or Bound

• Author names may be bound in a database and this
  means that the last name plus first name or
  initials can be searched quickly as one word
  
• A bound descriptor such as computer-assisted
  instruction can be searched as one word
  
• And, double-posted means it can be searched using
  any of the three terms computer, assisted, or
  instruction

20
Stopwords

• Common words used so often and are so heavily
  posted in the database quite literally those
  words that the search system ignores
  
• Otherwise, inordinate amount of processing time
  expended when cluttered with millions of
  occurrences of words
  
• Such terms in dictionary file would also require
  huge amounts of storage space
  
• Each system has determines its own list

21
Stopwords as Concepts

• Type A Behavior or Cyclosporin A used to be
  extremely difficult to search before bound
  descriptors evolved
  
• OK when they occur within a bound descriptor
  phrase
  
• If Type A Behavior is double-posted, however,
  only the words type and behavior are entered
  into the dictionary file
  
• Bypass surgery another case in point

22
Punctuation

• Many systems elect to drop as it occurs within
  the text or descriptors
  
• Apostrophes, hyphens, asterisks, parentheses,
  etc. will be disregarded and not appear in the
  dictionary file
  
• But, they do still print out in retrieval text
• Check system documentation for rules
• Childrens literature 3 words!

23
System Limits

• Each system will have limits on how many
  characters make a search statement
  
• Also, a limit on the number of search queries or
  statements the system will hold

24
Remember.......

• Rules for parsing, text-editing, and query memory
  may seem purely mechanical and arbitrary, but
  because they vary widely and can affect your
  searching, you must be aware of them for the
  system youre using before you begin a search.

25
Search Logic

• Basic and common to most online systems
• Order of Operators
• George Boole - Mathematical Logic
• John Venn - Pictorial Representations

26
Logical Operators

• Also called Boolean Operators
• Logically represent relationships between
  concepts
  
• AND, OR, and NOT

27
AND

• Simply means two or more concepts are contained
  within one set or within one document and is
  represented by two intersecting circles, where
  the ANDed area is shaded in .
• AND Intersection
• Decreases retrieval

28
OR

• Creates the larger set by looking for any
  occurrence of the terms you have listed in your
  search
  
• Merely an addition process or Union
• OR is represented by the entire area of all the
  circles
  
• Increases retrieval
• Eliminates duplication within the set

29
NOT

• Excludes a term or terms
• NOT Exclusivity or Difference
• Must be used Judiciously!
• Effective use is to employ NOT in strategies to
  remove sets already viewed or printed
  
• Cost Savings Effect

30
Order of Processing

• Processing Order varies by system for
  logical/boolean operators
  
• DIA processes NOT first, followed by AND and then
  OR
  
• Most systems process the ANDs first
• Check system documentation

31
NESTING

• Allows us to group sets and effectively use
  boolean operators
  
• Use of Parentheses directs the system to process
  certain operations first
  
• Again, important to now the operator order
• Works because most systems follow the rule that
  what is enclosed in ( ) will be processed first
  and a good idea to nest concepts

32
Positional Operators

• Also called Proximity Operators
• Express relationships between terms contained in
  the same document
  
• Ordinarily, a proximity operator stipulates that
  one term take a certain order
  
• Specifies word order of search terms
• ADJ, WITH (or NEAR), SAME(or FIELD), LINK, etc.

33
Positional Operators (contd)

• ADJ adjacency very specific word order term
  must be directly next to another term in that
  specified order most specific of proximity
  operators
• WITH Terms can occur in any order leaving room
  for unanticipated adjectives between two terms
  preference is WITH over ADJ bi-directional
  
• Both are ususally defined to a sentence which may
  vary in definition
  

34
Positional Operators (contd)

• SAME(or FIELD) Words in same field but no word
  order specified
  
• LINK Words in same descriptor unit. Used to
  link heading and subheadings
  
• n Up to n intervening words

35
Hierarchy of Positional Operators

• ADJ is first and most specific in the hierarchy
• WITH is broader in the hierarchy
• SAME is broadest in the hierarchy
• All combine concepts within the same document as
  opposed to logical operators
  
• Order of processing, generally is ADJ, followed
  by WITH followed by SAME

36
Other System Commands

• Logon and Logoff
• Cost
• Time
• Passwords
• Save Search
• Execute Search

• Type/Print
• Delete Sets
• Limits
• Sort
• Change Databases
• Display Sets