1. INFORMATION RETRIEVAL TECHNIQUES BY DR. ADNAN ABID
Lecture # 17
Processing a phrase query
Proximity queries

2. ACKNOWLEDGEMENTS
The presentation of this lecture has been taken from the underline sources
“Introduction to information retrieval” by Prabhakar Raghavan, Christopher D. Manning, and Hinrich Schütze
“Managing gigabytes” by Ian H. Witten, ‎Alistair Moffat, ‎Timothy C. Bell
“Modern information retrieval” by Baeza-Yates Ricardo, ‎ 
“Web Information Retrieval” by Stefano Ceri, ‎Alessandro Bozzon, ‎Marco Brambilla

3. Outline Processing a phrase query Proximity queries
Combination schemes

4. Processing a phrase query
Extract inverted index entries for each distinct term: to, be, or, not.
Merge their doc:position lists to enumerate all positions with “to be or not to be”.
to:
2:1,17,74,222,551; 4:8,16,190,429,433; 7:13,23,191; ...
be:
1:17,19; 4:17,191,291,430,434; 5:14,19,101; ...
Same general method for proximity searches
00:06:20  00:07:45 (merge their)
00:07:50  00:16:05 (to: & be:)(NLE: layering Required)
00:18:30  00:18:42

5. Proximity queries
LIMIT! /3 STATUTE /3 FEDERAL /2 TORT Here, /k means “within k words of”.
Clearly, positional indexes can be used for such queries; biword indexes cannot.
Exercise: Adapt the linear merge of postings to handle proximity queries. Can you make it work for any value of k?
00:19:30  00:19:50
00:21:12  00:21:50(limit)
00:22:50  00:23:40 (clearly)
00:24:20  00:24:40 (exercise)

6. Positional index size
Can compress position values/offsets as we did with docs in the last lecture
Nevertheless, this expands postings storage substantially
00:29:20  00:29:40(can)
00:30:00  00:30:30(nevertheless)

7. Positional index size
Need an entry for each occurrence, not just once per document
Index size depends on average document size
Average web page has <1000 terms
SEC filings, books, even some epic poems … easily 100,000 terms
Consider a term with frequency 0.1%
Why?
00:32:10  00:33:10(Index size)
00:33:14  00:33:25(Consider)
00:33:30  00:34:30(Table)
100 1
100,000
1000
Positional postings
Postings
Document size

8. Rules of thumb
Positional index size factor of 2-4 over non-positional index
Positional index size 35-50% of volume of original text
Caveat: all of this holds for “English-like” languages
00:34:50  00:35:00(Positional 2-4)
00:35:16  00:36:00(Positional 35-50)

9. Combination schemes
A positional index expands postings storage substantially (Why?)
Biword indexes and positional indexes approaches can be profitably combined
For particular phrases (“Michael Jackson”, “Britney Spears”) it is inefficient to keep on merging positional postings lists
Even more so for phrases like “The Who”
00:39:50  00:40:10
00:42:00  00:42:20

10. Wild Card Queries Example Stan*  Standard, Stanford S*T  Start
*ion  Option
Pa*an  Pakistan
Pa*t*an etc…
00:48:55  00:50:15

11. Resources for today’s lecture
MG 3.6, 4.3; MIR 7.2
Porter’s stemmer: http//
H.E. Williams, J. Zobel, and D. Bahle, “Fast Phrase Querying with Combined Indexes”, ACM Transactions on Information Systems.