Lecture 1: Introduction and the Boolean Model
Information Retrieval
Computer Science Tripos Part II
Ronan Cummins1
Natural Language and Information Processing (NLIP) Group
[email protected]
2016
1
Adapted from Simone Teufel’s original slides
1
Overview
1 Motivation
Definition of “Information Retrieval”
IR: beginnings to now
2 First Boolean Example
Term-Document Incidence matrix
The inverted index
Processing Boolean Queries
Practicalities of Boolean Search
What is Information Retrieval?
Manning et al, 2008:
Information retrieval (IR) is finding material . . . of an unstructured
nature . . . that satisfies an information need from within large
collections . . . .
2
What is Information Retrieval?
Manning et al, 2008:
Information retrieval (IR) is finding material . . . of an unstructured
nature . . . that satisfies an information need from within large
collections . . . .
3
Document Collections
4
Document Collections
IR in the 17th century: Samuel Pepys, the famous English diarist,
subject-indexed his treasured 1000+ books library with key words.
5
Document Collections
6
What we mean here by document collections
Manning et al, 2008:
Information retrieval (IR) is finding material (usually documents)
of an unstructured nature . . . that satisfies an information need
from within large collections (usually stored on computers).
Document Collection: text units we have built an IR system
over.
Usually documents
But could be
memos
book chapters
paragraphs
scenes of a movie
turns in a conversation...
Lots of them
7
IR Basics
Document
Collection
Query IR System
Set of relevant
documents
8
IR Basics
web
pages
Query IR System
Set of relevant
web pages
9
What is Information Retrieval?
Manning et al, 2008:
Information retrieval (IR) is finding material (usually documents)
of an unstructured nature . . . that satisfies an information need
from within large collections (usually stored on computers).
10
Structured vs Unstructured Data
Unstructured data means that a formal, semantically overt,
easy-for-computer structure is missing.
In contrast to the rigidly structured data used in DB style
searching (e.g. product inventories, personnel records)
SELECT *
FROM business catalogue
WHERE category = ’florist’
AND city zip = ’cb1’
This does not mean that there is no structure in the data
Document structure (headings, paragraphs, lists. . . )
Explicit markup formatting (e.g. in HTML, XML. . . )
Linguistic structure (latent, hidden)
11
Information Needs and Relevance
Manning et al, 2008:
Information retrieval (IR) is finding material (usually documents) of
an unstructured nature (usually text) that satisfies an information
need from within large collections (usually stored on computers).
An information need is the topic about which the user desires
to know more about.
A query is what the user conveys to the computer in an
attempt to communicate the information need.
A document is relevant if the user perceives that it contains
information of value with respect to their personal information
need.
12
Types of information needs
Manning et al, 2008:
Information retrieval (IR) is finding material . . . of an unstructured
nature . . . that satisfies an information need from within large
collections . . . .
Known-item search
Precise information seeking search
Open-ended search (“topical search”)
13
Information scarcity vs. information abundance
Information scarcity problem (or needle-in-haystack problem):
hard to find rare information
Lord Byron’s first words? 3 years old? Long sentence to the
nurse in perfect English?
. . . when a servant had spilled an urn of hot coffee over his legs, he replied to
the distressed inquiries of the lady of the house, ’Thank you, madam, the
agony is somewhat abated.’ [not Lord Byron, but Lord Macaulay]
Information abundance problem (for more clear-cut
information needs): redundancy of obvious information
What is toxoplasmosis?
14
Relevance
Manning et al, 2008:
Information retrieval (IR) is finding material (usually documents) of
an unstructured nature (usually text) that satisfies an information
need from within large collections (usually stored on computers).
Are the retrieved documents
about the target subject
up-to-date?
from a trusted source?
satisfying the user’s needs?
How should we rank documents in terms of these factors?
More on this in a lecture soon
15
How well has the system performed?
The effectiveness of an IR system (i.e., the quality of its search
results) is determined by two key statistics about the system’s
returned results for a query:
Precision: What fraction of the returned results are relevant to
the information need?
Recall: What fraction of the relevant documents in the
collection were returned by the system?
What is the best balance between the two?
Easy to get perfect recall: just retrieve everything
Easy to get good precision: retrieve only the most relevant
There is much more to say about this – lecture 6
16
IR today
Web search ( )
Search ground are billions of documents on millions of
computers
issues: spidering; efficient indexing and search; malicious
manipulation to boost search engine rankings
Link analysis covered in Lecture 8
Enterprise and institutional search ( )
e.g company’s documentation, patents, research articles
often domain-specific
Centralised storage; dedicated machines for search.
Most prevalent IR evaluation scenario: US intelligence analyst’s
searches
Personal information retrieval (email, pers. documents; )
e.g., Mac OS X Spotlight; Windows’ Instant Search
Issues: different file types; maintenance-free, lightweight to run
in background
17
A short history of IR
1945 1950s 1960s 1970s 1990s 2000s
1980s
Term Cranfield Salton;
IR coined experiments TREC
by Calvin VSM
memex Moers
Boolean
IR
Literature
searching SMART
systems; Multimedia
evaluation Multilingual
pagerank
by P&R (CLEF)
(Alan Kent)
1
Recommendation
Systems
rec all
precision/
recall
prec ision
no items retrieved
18
IR for non-textual media
19
Similarity Searches
20
Areas of IR
“Ad hoc” retrieval and classification (lectures 1-5)
web retrieval (lecture 8)
Support for browsing and filtering document collections:
Evaluation lecture 6)
Clustering (lecture 7)
Further processing a set of retrieved documents, e.g., by using
natural language processing
Information extraction
Summarisation
Question answering
21
Overview
1 Motivation
Definition of “Information Retrieval”
IR: beginnings to now
2 First Boolean Example
Term-Document Incidence matrix
The inverted index
Processing Boolean Queries
Practicalities of Boolean Search
Boolean Retrieval
In the Boolean retrieval model we can pose any query in the
form of a Boolean expression of terms
i.e., one in which terms are combined with the operators and,
or, and not.
Shakespeare example
22
Brutus AND Caesar AND NOT Calpurnia
Which plays of Shakespeare contain the words Brutus and
Caesar, but not Calpurnia?
Naive solution: linear scan through all text – “grepping”
In this case, works OK (Shakespeare’s Collected works has less
than 1M words).
But in the general case, with much larger text colletions, we
need to index.
Indexing is an offline operation that collects data about which
words occur in a text, so that at search time you only have to
access the precompiled index.
23
The term-document incidence matrix
Main idea: record for each document whether it contains each
word out of all the different words Shakespeare used (about 32K).
Antony Julius The Hamlet Othello Macbeth
and Caesar Tempest
Cleopatra
Antony 1 1 0 0 0 1
Brutus 1 1 0 1 0 0
Caesar 1 1 0 1 1 1
Calpurnia 0 1 0 0 0 0
Cleopatra 1 0 0 0 0 0
mercy 1 0 1 1 1 1
worser 1 0 1 1 1 0
...
Matrix element (t, d) is 1 if the play in column d contains the
word in row t, 0 otherwise.
24
Query “Brutus AND Caesar AND NOT Calpunia”
We compute the results for our query as the bitwise AND between
vectors for Brutus, Caesar and complement (Calpurnia):
Antony Julius The Hamlet Othello Macbeth
and Caesar Tempest
Cleopatra
Antony 1 1 0 0 0 1
Brutus 1 1 0 1 0 0
Caesar 1 1 0 1 1 1
Calpurnia 0 1 0 0 0 0
Cleopatra 1 0 0 0 0 0
mercy 1 0 1 1 1 1
worser 1 0 1 1 1 0
...
This returns two documents, “Antony and Cleopatra” and
“Hamlet”.
25
Query “Brutus AND Caesar AND NOT Calpunia”
We compute the results for our query as the bitwise AND between
vectors for Brutus, Caesar and complement (Calpurnia):
Antony Julius The Hamlet Othello Macbeth
and Caesar Tempest
Cleopatra
Antony 1 1 0 0 0 1
Brutus 1 1 0 1 0 0
Caesar 1 1 0 1 1 1
¬Calpurnia 1 0 1 1 1 1
Cleopatra 1 0 0 0 0 0
mercy 1 0 1 1 1 1
worser 1 0 1 1 1 0
...
This returns two documents, “Antony and Cleopatra” and
“Hamlet”.
26
Query “Brutus AND Caesar AND NOT Calpunia”
We compute the results for our query as the bitwise AND between
vectors for Brutus, Caesar and complement (Calpurnia):
Antony Julius The Hamlet Othello Macbeth
and Caesar Tempest
Cleopatra
Antony 1 1 0 0 0 1
Brutus 1 1 0 1 0 0
Caesar 1 1 0 1 1 1
¬Calpurnia 1 0 1 1 1 1
Cleopatra 1 0 0 0 0 0
mercy 1 0 1 1 1 1
worser 1 0 1 1 1 0
AND 1 0 0 1 0 0
Bitwise AND returns two documents, “Antony and Cleopatra” and
“Hamlet”.
27
The results: two documents
Antony and Cleopatra, Act III, Scene ii
Agrippa [Aside to Dominitus Enobarbus]: Why, Enobarbus,
When Antony found Julius Caesar dead,
He cried almost to roaring, and he wept
When at Philippi he found Brutus slain.
Hamlet, Act III, Scene ii
Lord Polonius: I did enact Julius Caesar: I was killed i’ the
Capitol; Brutus killed me.
28
Bigger collections
Consider N=106 documents, each with about 1000 tokens
109 tokens at avg 6 Bytes per token ⇒ 6GB
Assume there are M=500,000 distinct terms in the collection
Size of incidence matrix is then 500,000 ×106
Half a trillion 0s and 1s
29
Can’t build the Term-Document incidence matrix
Observation: the term-document matrix is very sparse
Contains no more than one billion 1s.
Better representation: only represent the things that do occur
Term-document matrix has other disadvantages, such as lack
of support for more complex query operators (e.g., proximity
search)
We will move towards richer representations, beginning with
the inverted index.
30
The inverted index
The inverted index consists of
a dictionary of terms (also: lexicon, vocabulary)
and a postings list for each term, i.e., a list that records which
documents the term occurs in.
Brutus 1 2 4 11 31 45 173 174
Caesar 1 2 4 5 6 16 57 132 179
Calpurnia 2 31 54 101
31
Processing Boolean Queries: conjunctive queries
Our Boolean Query
Brutus AND Calpurnia
Locate the postings lists of both query terms and intersect them.
Brutus 1 2 4 11 31 45 173 174
Calpurnia 2 31 54 101
Intersection 2 31
Note: this only works if postings lists are sorted
32
Algorithm for intersection of two postings
INTERSECT (p1, p2)
1 answer ← <>
2 while p1 6= NIL and p2 6= NIL
3 do if docID(p1) = docID(p2)
4 then ADD (answer, docID(p1))
5 p1 ← next(p1)
6 p2 ← next(p2)
7 if docID(p1) < docID(p2)
8 then p1← next(p1)
9 else p2← next(p2)
10 return answer
Brutus 1 2 4 11 31 45 173 174
Calpurnia 2 31 54 101
Intersection 2 31
33
Complexity of the Intersection Algorithm
Bounded by worst-case length of postings lists
Thus “officially” O(N), with N the number of documents in
the document collection
But in practice much, much better than linear scanning,
which is asymptotically also O(N)
34
Query Optimisation: conjunctive terms
Organise order in which the postings lists are accessed so that least
work needs to be done
Brutus AND Caesar AND Calpurnia
Process terms in increasing document frequency: execute as
(Calpurnia AND Brutus) AND Caesar
Brutus 8 1 2 4 11 31 45 173 174
Caesar 9 1 2 4 5 6 16 57 132 179
Calpurnia 4 2 31 54 101
35
Query Optimisation: disjunctive terms
(maddening OR crowd) AND (ignoble OR strife) AND (killed OR slain)
Process the query in increasing order of the size of each
disjunctive term
Estimate this in turn (conservatively) by the sum of
frequencies of its disjuncts
36
Practical Boolean Search
Provided by large commercial information providers
1960s-1990s
Complex query language; complex and long queries
Extended Boolean retrieval models with additional operators –
proximity operators
Proximity operator: two terms must occur close together in a
document (in terms of certain number of words, or within
sentence or paragraph)
Unordered results...
37
Examples
Westlaw : Largest commercial legal search service – 500K
subscribers
Medical search
Patent search
Useful when expert queries are carefully defined and
incrementally developed
38
Does Google use the Boolean Model?
On Google, the default interpretation of a query [w1 w2 ... wn ] is
w1 AND w2 AND ... AND wn
Cases where you get hits which don’t contain one of the w−i :
Page contains variant of wi (morphology, misspelling,
synonym)
long query (n is large)
Boolean expression generates very few hits
wi was in the anchor text
Google also ranks the result set
Simple Boolean Retrieval returns matching documents in no
particular order.
Google (and most well-designed Boolean engines) rank hits
according to some estimator of relevance
39
Reading
Manning, Raghavan, Schütze: Introduction to Information
Retrieval (MRS), chapter 1
40
