Scribd
Upload
211 views

Tutorial Suffix Tree

The document introduces two data structures for text indexing: suffix trees and suffix arrays. A suffix tree is a compact tree that represents all suffixes of a text, with each leaf mapping to a suffix and paths spelling out suffixes. It allows searching for patterns in optimal O(m+occurrences) time by traversing the tree with the pattern. A suffix array simply stores all suffixes in lexicographic order in an array, allowing searches in O(m log n) time via binary search. Both use linear O(n) space but suffix arrays use less space in practice.

Uploaded by

Vikash Kumar
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
211 views

Tutorial Suffix Tree

The document introduces two data structures for text indexing: suffix trees and suffix arrays. A suffix tree is a compact tree that represents all suffixes of a text, with each leaf mapping to a suffix and paths spelling out suffixes. It allows searching for patterns in optimal O(m+occurrences) time by traversing the tree with the pattern. A suffix array simply stores all suffixes in lexicographic order in an array, allowing searches in O(m log n) time via binary search. Both use linear O(n) space but suffix arrays use less space in practice.

Uploaded by

Vikash Kumar
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

CS4311 Design and Analysis of Algorithms

Suffix Tree and Suffix Array

About this tutorial


Introduce two data structures for text indexing problem: Suffix Tree and Suffix Array

Text Indexing
String Matching problem: Given a text T and a pattern P, how to locate all occurrences of P in T ? KMP algorithm can solve this in O(|T|+|P|) time optimal In some applications, T is very long, and given in advance, and we will search different patterns against it later E.g., T= Human DNA, P = gene
3

Text Indexing
Text Indexing problem: Suppose a text T is known. Can we build a data structure for T, such that for any pattern P given later, we can find all occurrences of P in T quickly ? The data structure is called an index of T Target: search better than O(|T|+|P|) ??
4

Text Indexing
Two main kinds of text indexes: Word-Based: (for texts formed by words) Used by most text search engine E.g., Inverted Files Full-Text: (for texts with no word boundaries) Used in indexing DNA E.g., Suffix Tree, Suffix Array
5

Suffix Tree
Let T[1..n] be a text with n characters we assume T[n] is a unique character For any j, T[j..n] is called a suffix of T T has exactly n suffixes Weiner (1973) and McCreight (1976) independently invented the suffix tree a tree formed by putting all suffixes of T together
6

# 8 c a # 6 c # 3 a a c a

c # 7 a c # 4 c a a c # 1
7

a ca a

# 5

c # 2

Suffix Tree of acacaac#

Definition of a Suffix Tree


Suffix tree is an edge-labeled compact tree (no degree-1 nodes) with n leaves each leaf suffix leaf label starting pos of suffix If we traverse from root to leaf k : edge labels along path suffix T[k..n] edge-label to each child starts with different character
8

Searching in a Suffix Tree


Theorem: If a pattern P occurs at position j in T, P is a prefix of T[j..n] This suggests the searching algorithm below: Start from root of the suffix tree Traverse the suffix tree using P What we are doing is to match P with all suffixes of T at the same time
9

Searching in a Suffix Tree


Theorem: Pattern P occurs in T if and only if all chars of P are matched in the traversal of the searching algorithm Questions: 1. How to locate the occurrences? 2. What is the searching time? O(|P|+r) time, where r = #occurrences
10

Space Usage
There are O(n) nodes and O(n) edges in the suffix tree O(n) space ? Each edge needs to store its label, which can contain O(n) chars In the worst-case, total O(n2) chars Can we reduce space usage?
11

Space Usage
Observation: Each edge label must be equal to some substring of T Clever Idea: 1. Store T, and 2. Replace each edge label by 2 integers, telling which substring it is equal to Total space: O(n)
12

[8,8] 8 [6,8] 5 [8,8] 6 [6,8] 3 [1,1] [2,2] [3,3]

[2,2] [8,8] 7 [6,8] 4 [4,8] 2 [3,3]

[4,8] 1
13

Suffix Tree of acacaac#

Suffix Array
Although suffix tree takes O(n) space, the hidden constant is quite large around 40n to 60n bytes Manber and Myers (1990) simplified the suffix tree, and invented the suffix array An array storing the suffixes of T in the dictionary order
14

Suffix Array
Suffix Array of acacaac# 1 2 3 4 5 6 7 8 # aac# ac# acaac# acacaac# c# caac# cacaac#

The suffix array SA for T has n entries For any j, SA[j] stores the jth smallest suffix, based on alphabetical order Theorem: If P occurs in T, its occurrences correspond to consecutive region in SA
15

Suffix Array
Suffix Array of acacaac# 1 2 3 4 5 6 7 8 # aac# ac# acaac# acacaac# c# caac# cacaac#

Searching P takes O(|P| log n) time using binary search Space: We can represent each suffix by its starting position O(n) space In practice, around 14n bytes
16

You might also like