Week 4

Question 1

Boolean queries

Answer 1

• process a query in the form of boolean expression
• AND , OR , NOT to join terms
• Simplest IR model

Question 2

Boolean Query optimization

Answer 2

• for each n terms, get postings then AND them together

- Execute smaller frequency first ( Calpurnia AND Brutus) AND Caesar

Question 3

Boolean general optimization

Answer 3

• OR AND
• get document frequency for all terms
• estimate size of each OR by summating document frequencies
  Process in increasing order of OR sizes

Question 4

Biword Indices

Answer 4

• Index every consecutive pair of terms in text as a phrase
• Friends, Romans, Countrymen = Friends Romans + Romans Countrymen
• each biword is a dictionary term

Question 5

Longer phrase queries, what are problems

Answer 5

• Break them down
• Wilfred laurier university waterloo -> Wilfred Laurier AND laurier university AND university waterloo
  Problems
• issue is false positive
• Index blow up due to big dictionary
• not standard solution for TR

Question 6

Positional indices

Answer 6

• In postings index table store Document frequency but also position in which token appears

Question 7

Parse phrase query in positional indices

Answer 7

• Extract inverted index entries for each distinct terms; to, be, or, not
• Merge document position list to enumerate all positions leading to “to be or not to be”

Question 8

Positional index size

Answer 8

• positional index expands postings storage substantially
• positional index used as standard for TR due to use value of phrase and proximity queries
• size depends on avg size of document

Question 9

What can positional size index be used for

Answer 9

• can be used explicitly for implicitly ranking retrieval system
• needs entry for all occurrence of the term, not just on per document

Question 10

Combination of biword and positional index

Answer 10

• solution to large index size