1 of 22

1
CSA1013Historical Perspectives of
Information Search and Retrieval

• Dr. Christopher Staff
• Department of Computer Science AI
• University of Malta

2
Aims and Objectives

• What is Information Search and Retrieval?
• Whats the state-of-the-art?
• How did we get here?
• Where are we likely to go next?

3
Whats Information Search and Retrieval?

• Whats information?
• Structured vs. unstructured
• Where is it?
• Question answering vs. Information lack or
  information need

4
Whats the state-of-the-art?

• Information Retrieval in the real world
• Web-based search engines
• Google, AllTheWeb, AltaVista, etc.
• Web directories
• Yahoo, Excite, etc.

5
Whats the state-of-the-art?

• Google, and Google-like search engines
• Index gt 8 billion web pages (pdf, doc, html, )
• User expresses Query
• terms, natural language query, etc
• System compares query to indexed documents
• Returns list of relevant documents

6
Whats the state-of-the-art?

• Recent study by Jansen Spink Jansen shows
• Query 2.14 terms Spink
• Queries with 1 term 53!
• 54 of users are satisfied with first page of
  results (list of 10 documents)
• 80 of users view not more than 10 - 20 results
• 27.6 read only one document!
• 66 read lt 5 documents

7
Has life always been this good?

• It would seem that were living in information
  heaven
• Any info we seek is just a couple of query terms
  away
• In reality, although majority of queries appear
  to be trivial, the reality is quite different

8
Has life always been this good?

• What if we want to find all relevant information?
  (The Invisible Web)
• What if we want to find something that is
  difficult to describe?
• What if we dont know what were looking for?
• What tools do we use to find info in
  encyclopaedias, dictionaries, newspapers,
  reference manuals, novels and other books?

9
Here beginneth the history lesson

• People have devised tools to find information
  again ever since we learnt to write things down
• Think of information stored on your personal
  computers how do you find something that you
  wrote last month, last year?

10
Prehistory!

• Well, nearly!
• Early writings
• Papyrus scrolls
• No paragraph, page numbers, etc
• Couldnt scroll to the end to read an index
• Instead, Greek/Roman libraries used
  sillybus/index of title

11
Greeks/Romans

• 3BC, Greeks probably use alphabetization in
  Library of Alexandria
• Around 2BC (Rome), evidence of hierarchies of
  information/classification systems
• Greeks probably earlier
• Also, Tables of Contents date from around 2BC
  (Pliny the Elder reports before 79AD)

12
Printing Press

• Not much else was to happen until 1455, with the
  advent of the printing press
• Previously, still difficult to refer to
  information within a book, because copies were
  inaccurate
• Info on one page in one book could be on a
  different page in other copies

13
Indices and the Printing Press

• Still, alphabetization was on initial letter,
  then on first four letters
• Not until 18th Century did full alphabetization
  occur!

14
The Second World War and beyond

• In 1945, Vannevar Bush publishes As We May
  Think in the Atlantic Monthly
• In 1949, Warren Weaver writes that if Chinese is
  English codification, then Machine Translation
  should be possible
• These give rise to intelligent and
  statistical (or surface-based) approaches to
  Information Search and Retrieval respectively
  (amongst other things -))

15
Intelligent vs. Surface-based

• Concepts
• 1950s
• Lay in waiting for years, because
  hardware/software not around

• Words
• 1950s
• First approaches were Key Words in Context
  (KWIC)

16
Intelligent vs. Surface-based

• 1960s
• Generality in AI (John McCarthy)

• 1960s
• Boolean Search
• Measures of performance effectiveness
• Thesaural Lookup
• Vector Space Model

17
Intelligent vs. Surface-based

• 1970s
• Expert Systems
• Still about understanding information and
  reasoning with and about it

• 1970s
• Explosion in availability of electronic text
  collections
• Library Retrieval Systems
• Full-text indexing
• Probabilistic IR
• Relevance Feedback

18
Intelligent vs. Surface-based

• 1980s
• Conceptual IR
• Knowledge Rep Langs
• Lenats CYC
• Contextual Reasoning
• 5th Generation Computing, Japan
• LSI feeds Statistical IR

• 1980s
• OPACs
• IR used by non-specialists
• Extended Boolean IR
• Word Sense Disambiguation
• Statistical IR (LSI, etc)
• Internet

19
Intelligent vs. Surface-based

• 1990s
• Better language processing
• information extraction
• entity name recognition
• Advances in contextual reasoning, ontologies

• 1990s
• WWW (1995 c. 10M pages, 2003 c. 3B!)
• Multimedia Indexing Retrieval
• Web-based search engines

20
Intelligent vs. Surface-based

• 2000s
• Semantic Web

• 2000s
• Faster processors
• More memory
• Cheaper storage space
• More superficial comparisons

21
Intelligent vs. Surface-based

• The future
• Computers that can find precisely the information
  you seek
• Even if the answer is non-obvious
• Or the answer needs to be the result of reasoning
• MyLifeBits

• The future
• Computers that can approximate the information
  you seek
• At much less cost
• At the expense of correctness
• MyLifeBits

22
Main References

• Paijmans, J.J., last updated 2004, The Retrieval
  of Information from historical perspective,
  http//pi0959.kub.nl/Paai/Onderw/V-I/Content/histo
  ry.html
• American Society of Indexers, last updated 2005,
  How Information Retrieval Started,
  http//www.asindexing.org/site/history.shtml
• Jansen Jansen, B.J., and Spink, A., 2003, An
  Analysis of Web Documents Retrieved and Viewed,
  in Proceedings of the 4th International
  Conference on Internet Computing, Las Vegas,
  Nevada, 23-26 June 2003. http//ist.psu.edu/facult
  y_pages/jjansen/academic/pubs/pages_viewed.pdf
• Spink Spink, A., et. al., 2001, Searching the
  Web The Public and their Queries, in JASIST
  2001. http//jimjansen.tripod.com/academic/pubs/ja
  sist2001/jasist2001.html