Scribd
Upload
52 views

Solutions To Exercises: Chapter 1 - Information Retrieval Models

Uploaded by

Sai Pawan
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
52 views

Solutions To Exercises: Chapter 1 - Information Retrieval Models

Uploaded by

Sai Pawan
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Solutions to Exercises

Chapter 1 – Information Retrieval Models


Djoerd Hiemstra

1.1(c) The Venn diagrams of Figure 1.2 show exactly 8 disjoint subsets of documents, including
the area around the diagram. Whatever the final result of a Boolean query, each subset is
either selected or not, so in total 28 = 256 subsets can be defined.
1.2(b) Vector spaces are metric spaces, i.e., a set of objects equipped with a distance, where the
distance from q to d is the same as from d to q, so if we change q for d and d for q,
the distance should remain the same. In practice however, many practical term weighting
algorithms do not use the same weights for queries and documents. In such a case, the
similarity might not be equal to 0.08. One might argue that such a model is not a vector
space model, though.
1.3(c) If we add a single document, the document frequencies of terms that occur in the added
document will increase by 1. Furthermore, the number of documents N changes, which
affects the idfs of all other terms.
1.4(b) Assuming that the idf is calculated using the number of documents N as shown in Equation
(1.5), all weights need to change, as is the case in Exercise 3.
1.5(a) For each term, the probabilistic model considers either presence of the term in the document,
or absence of the term in the document. So, per term, there are two cases, hence 23 = 8
different scores in the case of three query terms.
1.6(c) As said above, the model only considers presence or absence, but in the case of presence
it does not consider the number of occurrences of the term in the document. If the term is
present in both D and E, then they will be assigned the exact same score. If the system
needs to provide a total ranking, then the implementation has to determine which document
is ranked first.
1.7(b) Without smoothing, the probability is a simple fraction of the number of occurrences, divided
by the total number of terms in the document. Of course, language models do not need an
additional term weighting algorithm.
1.8(c) If λ = 0, the model does not use smoothing. Without smoothing, terms that do not occur
in the document are assigned zero probability. The score of a document is determined by
multiplying the probabilities of the single terms. If one of them is zero, the final score of
the document is zero.

Information Retrieval: Searching in the 21st Century Edited by A. Göker & J. Davies
© 2009 John Wiley & Sons, Ltd. ISBN: 978-0-470-02762-2

You might also like