Introduction to

Information Retrieval
Introducing ranked retrieval
 Ch. 6

Ranked retrieval
Thus far, our queries have all been Boolean.
Documents either match or don’t.
Good for expert users with precise understanding of their needs and the
collection.
Also good for applications: Applications can easily consume 1000s of results.
Not good for the majority of users.
Most users incapable of writing Boolean queries (or they are, but they think it’s
too much work).
Most users don’t want to wade through 1000s of results.
 This is particularly true of web search.
 Ch. 6

Problem with Boolean search:

feast or famine
Boolean queries often result in either too few (≈0) or too many (1000s)
results.
Query 1: “standard user dlink 650” → 200,000 hits
Query 2: “standard user dlink 650 no card found” → 0 hits
It takes a lot of skill to come up with a query that produces a manageable
number of hits.
AND gives too few; OR gives too many
 Ch. 6

Drawbacks of Boolean model

Ranked retrieval models

Rather than a set of documents satisfying a query expression, in ranked

retrieval models, the system returns an ordering over the (top) documents in
the collection with respect to a query
Free text queries: Rather than a query language of operators and expressions,
the user’s query is just one or more words in a human language
In principle, there are two separate choices here, but in practice, ranked
retrieval models have normally been associated with free text queries and
vice versa

5
 Ch. 6

Feast or famine: not a problem in ranked retrieval

When a system produces a ranked result set, large result sets are not an issue
Indeed, the size of the result set is not an issue
We just show the top k ( ≈ 10) results
We don’t overwhelm the user

Premise: the ranking algorithm works

 Ch. 6

Scoring as the basis of ranked retrieval

We wish to return in order the documents most likely to be useful to the
searcher
How can we rank-order the documents in the collection with respect to a
query?
Assign a score – say in [0, 1] – to each document
This score measures how well document and query “match”.
 Ch. 6

Query-document matching scores

We need a way of assigning a score to a query/document pair

Let’s start with a one-term query
If the query term does not occur in the document: score should be 0
The more frequent the query term in the document, the higher the score
(should be)
We will look at a number of alternatives for this
 Ch. 6

Take 1: Jaccard coefficient

A commonly used measure of overlap of two sets A and B is the Jaccard

coefficient
jaccard(A,B) = |A ∩ B| / |A ∪ B|
jaccard(A,A) = 1
jaccard(A,B) = 0 if A ∩ B = 0
A and B don’t have to be the same size.
Always assigns a number between 0 and 1.
 Ch. 6

Jaccard coefficient: Scoring example

What is the query-document match score that the Jaccard coefficient

computes for each of the two documents below?
Query: ides of march
Document 1: caesar died in march
Document 2: the long march
jaccard(q,d1) = | q ∩ d1 | / | q ∪ d1| = march/ ides of caesar died in march
= 1/6
jaccard(q,d2) = | q ∩ d2 | / | q ∪ d2| = march/ ides of march the long
= 1/5
 Ch. 6

Issues with Jaccard for scoring

It doesn’t consider term frequency (how many times a term occurs in a
document)
Rare terms in a collection are more informative than frequent terms
Jaccard doesn’t consider this information
We need a more sophisticated way of normalizing for length
we’ll use
. . . instead of |A ∩ B|/|A ∪ B| (Jaccard) for length normalization.
| A B | / | A B |
 Introduction to
Information Retrieval
Term frequency weighting
 Sec. 6.2

Recall: Binary term-document incidence matrix

Antony and Cleopatra Julius Caesar The Tempest Hamlet Othello Macbeth

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

The dimension of the vector is the size of the vocabulary.

Each document is represented by a binary vector ∈ {0,1}|V|
 Sec. 6.2

Term-document count matrices

Consider the number of occurrences of a term in a document:
Each document is a count vector in ℕ|V|: a column below

Antony and Cleopatra Julius Caesar The Tempest Hamlet Othello Macbeth

Antony 157 73 0 0 0 0
Brutus 4 157 0 1 0 0
Caesar 232 227 0 2 1 1
Calpurnia 0 10 0 0 0 0
Cleopatra 57 0 0 0 0 0
mercy 2 0 3 5 5 1
worser 2 0 1 1 1 0
The bag of words representation

I love this movie! It's sweet, but

with satirical humor. The dialogue is
great and the adventure scenes are
fun… It manages to be whimsical and
romantic while laughing at the
conventions of the fairy tale genre.
I would recommend it to just about
anyone. I've seen it several times,
and I'm always happy to see it again
whenever I have a friend who hasn't
seen it yet.
The bag of words representation: using a subset of words
x love xxxxxxxxxxxxxxxx sweet
xxxxxxx satirical xxxxxxxxxx
xxxxxxxxxxx great xxxxxxx
xxxxxxxxxxxxxxxxxxx fun xxxx
xxxxxxxxxxxxx whimsical xxxx
romantic xxxx laughing
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxx recommend xxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xx several xxxxxxxxxxxxxxxxx
xxxxx happy xxxxxxxxx again
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxx
The bag of words representation
great 2
love 2

recommend 1
laugh 1
happy 1
... ...
Bag of words model
Vector representation doesn’t consider the ordering of words in a document
John is quicker than Mary and Mary is quicker than John have the same
vectors

This is called the bag of words model.

Term frequency tf
The term frequency tft,d of term t in document d is defined as the number of
times that t occurs in d.
We want to use tf when computing query-document match scores. But how?
Raw term frequency is not what we want:
A document with 10 occurrences of the term is more relevant than a document
with 1 occurrence of the term.
But not 10 times more relevant.
Relevance does not increase proportionally with term frequency.

NB: frequency = count in IR

 Sec. 6.2

Log-frequency weighting
The log frequency weight of term t in d is

𝑤𝑡 , 𝑑=
0, {
1 + lo g 10 t f 𝑡 , 𝑑 , if t f 𝑡 , 𝑑 > 0
otherwise
0 → 0, 1 → 1, 2 → 1.3, 10 → 2, 1000 → 4, etc.

Score for a document-query pair: sum over terms t in both q and d:

score ¿ ∑ (1 + log t f 𝑡 , 𝑑)
𝑡 ∈𝑞∩ 𝑑
The score is 0 if none of the query terms is present in the document.
 Introduction to
Information Retrieval
(Inverse) Document frequency weighting
 Sec. 6.2.1

Document frequency
Rare terms are more informative than frequent terms
Recall stop words
Consider a term in the query that is rare in the collection (e.g., arachnocentric)
A document containing this term is very likely to be relevant to the query
arachnocentric
→ We want a high weight for rare terms like arachnocentric.
 Sec. 6.2.1

Document frequency, continued

Frequent terms are less informative than rare terms

Consider a query term that is frequent in the collection (e.g., high, increase,
line)
A document containing such a term is more likely to be relevant than a
document that doesn’t
But it’s not a sure indicator of relevance.
→ For frequent terms, we want positive weights for words like high, increase,
and line
But lower weights than for rare terms.
We will use document frequency (df) to capture this.
 Sec. 6.2.1

idf weight

dft is the document frequency of t: the number of documents that contain t

dft is an inverse measure of the informativeness of t
dft  N
We define the idf (inverse document frequency) of t by

We use log (N/dft) instead of N/dft to “dampen” the effect of idf.

idf t log10 ( N/df t )

Will turn out the base of the log is immaterial.

 Sec. 6.2.1

idf example, suppose N = 1 million

term dft idft
calpurnia 1
animal 100
sunday 1,000
fly 10,000
under 100,000
the 1,000,000

idf t log10 ( N/df t )

There is one idf value for each term t in a collection.
Effect of idf on ranking
Question: Does idf have an effect on ranking for one-term queries, like
iPhone

26
Effect of idf on ranking

Question: Does idf have an effect on ranking for one-term queries, like
iPhone
idf has no effect on ranking one term queries
idf affects the ranking of documents for queries with at least two terms
For the query capricious person, idf weighting makes occurrences of capricious
count for much more in the final document ranking than occurrences of person.

27
 Sec. 6.2.1

Collection vs. Document frequency

The collection frequency of t is the number of occurrences of t in the
collection, counting multiple occurrences.
Example:

Word Collection frequency Document frequency

insurance 10440 3997

try 10422 8760

Which word is a better search term (and should get a higher weight)?
 Introduction to
Information Retrieval
tf-idf weighting
 Sec. 6.2.2

tf-idf weighting

The tf-idf weight of a term is the product of its tf weight and its idf weight.

w t ,d (1  log tf t ,d ) log10 ( N / df t )

Best known weighting scheme in information retrieval
Note: the “-” in tf-idf is a hyphen, not a minus sign!
Alternative names: tf.idf, tf x idf

Increases with the number of occurrences within a document

Increases with the rarity of the term in the collection
 Sec. 6.2.2

Final ranking of documents for a query



31
 Sec. 6.3

Binary → count → weight matrix

Antony and Cleopatra Julius Caesar The Tempest Hamlet Othello Macbeth

Antony 5.25 3.18 0 0 0 0.35

Brutus 1.21 6.1 0 1 0 0
Caesar 8.59 2.54 0 1.51 0.25 0
Calpurnia 0 1.54 0 0 0 0
Cleopatra 2.85 0 0 0 0 0
mercy 1.51 0 1.9 0.12 5.25 0.88
worser 1.37 0 0.11 4.15 0.25 1.95

Each document is now represented by a real-valued vector of tf-idf

weights ∈ R|V|
 Introduction to
Information Retrieval
The Vector Space Model (VSM)
 Sec. 6.3

Documents as vectors

Now we have a |V|-dimensional vector space

Terms are axes of the space
Documents are points or vectors in this space
Very high-dimensional: tens of millions of dimensions when you apply this to
a web search engine
These are very sparse vectors – most entries are zero
 Sec. 6.3

Queries as vectors

Key idea 1: Do the same for queries: represent them as vectors in the space
Key idea 2: Rank documents according to their proximity to the query in this
space
proximity = similarity of vectors
proximity ≈ inverse of distance
Recall: We do this because we want to get away from the you’re-either-in-or-
out Boolean model
Instead: rank more relevant documents higher than less relevant documents
 Sec. 6.3

Formalizing vector space proximity

First cut: distance between two points

( = distance between the end points of the two vectors)
Euclidean distance?
Euclidean distance is a bad idea . . .
. . . because Euclidean distance is large for vectors of different lengths.
 Sec. 6.3

Why distance is a bad idea

The Euclidean distance

between q and d2 is large even
though the distribution of
terms in the query q and the
distribution of terms in the
document d2 are very similar.
 Sec. 6.3

Use angle instead of distance

Thought experiment: take a document d and append it to itself. Call this

document d′.
“Semantically” d and d′ have the same content
The Euclidean distance between the two documents can be quite large
The angle between the two documents is 0, corresponding to maximal
similarity.

Key idea: Rank documents according to angle with query.

 Sec. 6.3

From angles to cosines

The following two notions are equivalent.

Rank documents in decreasing order of the angle between query and document
Rank documents in increasing order of
cosine(query,document)

Cosine is a monotonically decreasing function for the interval [0 o, 180o]

 Sec. 6.3

From angles to cosines

But how – and why – should we be computing cosines?

 Sec. 6.3

Length normalization

A vector can be (length-) normalized by dividing each of its components by its

length – for this we use the L2 norm:

x 2
 i i
x 2

Dividing a vector by its L2 norm makes it a unit (length) vector (on surface of
unit hypersphere)
Effect on the two documents d and d′ (d appended to itself) from earlier slide:
they have identical vectors after length-normalization.
Long and short documents now have comparable weights
 Sec. 6.3

cosine(query,document)
Dot product Unit vectors

|𝑉 |

⃗•⃗
𝑞 𝑑 ⃗
𝑞 ⃗
𝑑
∑ 𝑞𝑖 𝑑𝑖
⃗
⃗ , 𝑑)=
cos (¿ 𝑞 = • =
𝑖=1
¿
𝑞||𝑑| |𝑑|
√ √
|⃗ ⃗ |⃗
𝑞 | ⃗ |𝑉 | |𝑉|
∑ 𝑞2𝑖 ∑ 𝑑𝑖2
𝑖=1 𝑖=1

qi is the tf-idf weight of term i in the query

di is the tf-idf weight of term i in the document

cos(q,d) is the cosine similarity of q and d … or,

equivalently, the cosine of the angle between q and d.
Cosine for length-normalized vectors

For length-normalized vectors, cosine similarity is simply the dot product (or
scalar product):

for q, d length-normalized.

43
Cosine similarity illustrated

44
 Sec. 6.3

Cosine similarity amongst 3 documents

How similar are
the novels term SaS PaP WH

SaS: Sense and affection 115 58 20

Sensibility jealous 10 7 11

PaP: Pride and gossip 2 0 6

Prejudice, and wuthering 0 0 38

WH: Wuthering
Term frequencies (counts)
Heights?

Note: To simplify this example, we don’t do idf weighting.

 Sec. 6.3

3 documents example contd.

Log frequency weighting After length normalization

term SaS PaP WH term SaS PaP WH
affection 3.06 2.76 2.30 affection 0.789 0.832 0.524
jealous 2.00 1.85 2.04 jealous 0.515 0.555 0.465
gossip 1.30 0 1.78 gossip 0.335 0 0.405
wuthering 0 0 2.58 wuthering 0 0 0.588

cos(SaS,PaP) ≈
0.789 × 0.832 + 0.515 × 0.555 + 0.335 × 0.0 + 0.0 × 0.0 ≈ 0.94
cos(SaS,WH) ≈ 0.79
cos(PaP,WH) ≈ 0.69

Why do we have cos(SaS,PaP) > cos(SAS,WH)?

 Introduction to
Information Retrieval
Calculating tf-idf cosine scores
in an IR system
 Sec. 6.4

tf-idf weighting has many variants

 Sec. 6.4

Weighting may differ in queries vs documents

Many search engines allow for different weightings for queries vs. documents
SMART Notation: denotes the combination in use in an engine, with the
notation ddd.qqq, using the acronyms from the previous table
A very standard weighting scheme is: lnc.ltc
Document: logarithmic tf (l as first character), no idf and cosine normalization
Query: logarithmic tf (l in leftmost column), idf (t in second column), cosine
normalization …
A bad idea?
 Sec. 6.4

tf-idf example: lnc.ltc

Document: car insurance auto insurance
Query: best car insurance
Term Query Document Prod
tf- tf-wt df idf wt n’lize tf-raw tf-wt wt n’lize
raw
auto 0 0 5000 2.3 0 0 1 1 1 0.52 0
best 1 1 50000 1.3 1.3 0.34 0 0 0 0 0
car 1 1 10000 2.0 2.0 0.52 1 1 1 0.52 0.27
insurance 1 1 1000 3.0 3.0 0.78 2 1.3 1.3 0.68 0.53

Exercise: what is N, the number of docs?

Doc length =
Score = 0+0+0.27+0.53 = 0.8
 Sec.
6.3

Computing cosine scores

Summary – vector space ranking

Represent the query as a weighted tf-idf vector

Represent each document as a weighted tf-idf vector
Compute the cosine similarity score for the query vector and each document
vector
Rank documents with respect to the query by score
Return the top K (e.g., K = 10) to the user