Automatic indexing

1
Automatic indexing

• Salton
• When the assignment of content identifiers is
  carried out with the aid of modern computing
  equipment the operation becomes automatic
  indexing

2
Pros and Cons of Automatic Indexing

• Pros
• (Lots of stuff out there to index)
• Consistency
• Cost reduction
• Time reduction
• Cons / limitations
• Human intellect
• Term relationships
• Misleading in retrieval
• Good algorithms, but generally domain-specific

3
Approaches and Methods

• Initial approach
• Create an inverted file
• Inverted file contains all the index terms
  automatically drawn from the document records
  according to the indexing technique used
• On-the-fly (facilitate retrieval)
• natural language processing (NLP)

4

• Recall
• No relevant doc. in DB retrieved
• Total no. of relevant docs in DB
• Retrieval of all the relevant documents from the
  whole database!!
• Precision
• No. of relevant doc. retrieved
• Total no. of documents retrieved
• Percentage of relevant documents retrieved.

5
Methods

• Stemming
• Used to conflate, or reduce morphological
  variants to a single term
• Should increase recall by grouping similar
  terms
• Achieve through stemming algorithms
• Porter algorithm, simple S stemmer
• Weakness, too severe
• morals and more stemmed to more
• Most effective general domain w/plurals, ing
  ed
• Utility in retrieval, truncation

6
Automatic Classification/Indexing Methods

• Statistical analysis
• Based on theory that terms frequency determines
  aboutness informetric properties
• High frequency / low frequency poor candidates
• Simple to perform
• Effective in creating index of single terms or
  broad class groupings
• Phrase indexing more complicated, and difficult
  to render
• E.g., running water highway traffic control
  in North Carolina
• Domain sensitive (better in some than others)
• Weighting options
• Within document tf (term frequency)
• Proportion of documents within the collection
  that have a term/s IDF (inverse document
  frequency)
• Combination of tfIDF quite successful
• Term discrimination

7
Methods

• Syntactical analysis
• Phrase (or grammatical) indexing
• Intended to address ambiguities with statistical
  methods
• Subject, object predicate
• Rules complex, NLP, domain specific
• Anaphoric statement problematic He ate apples
• Research shows only a slight improvement over
  statistical methods
• Probabilistic methods
• Term is a good indicator is it appears in a
  relevant document
• Basically tfIDF method, can include weighting
• Clustering, bringing like documents together
  based on the above methods, integrate thesauri,
  or other tools
• Classification/categorization, assignment of a
  top node (class notation or term)

8
Natural language vs. Controlled Vocabulary

• Natural language continuum
• lt-basic key word------------IR------------full
  NLP-gt
• Full discourse
• History
• Uniterm approach (Taube, 1953), Optical
  coincidence cards, edge hold punch cards
• Cranfield studies

9
Uniterm
10
Peek-a-boo card
11
Natural language vs. Controlled Vocab.

• Pros
• Cons

• Production cost
• Cost to the end-user
• Facilitate specificity in terms of access
• Exhaustivity indexing
• Handling of errors

12
What is Automatic Classification?

• Automatic manipulation of a documents contents
  to support logical grouping with other similar
  documents for organization and/or retrieval
  activities. Can include the assignment of, or
  manipulation of, classification notation.

13
Clustering techniques

• A technique by which relationships among data
  elements such as documents or document
  attributes are determined and closely related
  elements are grouped into clusters. (Korfhage,
  1997)
• Group vectors using various coefficients
• E.g., Dice Coefficient

14
Other methods

• Term weighting
• Various other IR approaches, automatic indexing

15
Why Automatic Classification?

• Classification is time consuming and expensive
• Knowledge structuring
• To much information
• Status of automatic classification
• Fairly experimental, although not completely
• Operational systems for e-mail

16
Why Automatic Classification?

• More expedient than a human
• Less costly /??
• More consistent than human
• Easier to fix errors

17
Why Automatic Classification?

• Your articles.
• How defined
• What was the purpose
• How was the automatic classification done, or
  discussed
• Outcome
• Experimental / Operational setting?