<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
<head>
   <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
   <meta name="Author" content="IBM">
   <meta name="GENERATOR" content="Mozilla/4.51 [en] (WinNT; I) [Netscape]">
   <title>Package-level Javadoc</title>
</head>
<body>
Provides the core functionality for spell-checking documents
<h2>
Package Specification</h2>
This package provides the interfaces for the notions of dictionary, edit distance, phonetic hash, 
spell event and spell-check iterator. For most of these interfaces a default implementation 
for english languages is provided. These implementations can be reused in custom dictionaries or 
spell-check iterators, or replaced by more specialized algorithms for a particular group of languages.
<h3>
Spell Check Engine</h3>
The central point to access the spell-checker functionality is the interface <tt>ISpellCheckEngine</tt>. 
Implementations of this interface provide support for life-cycle management, registering and unregistering
dictionaries, changing the locale of the engine and creating a spell-checker for a specific language.
<p>
The following steps are needed to obtain a spell-checker for a specific language:
<ul>
<li>Create an instance of <tt>ISpellCheckEngine</tt>. In this package, no default implementation is provided, 
since the management of the dictionary registering and loading is application dependent. Usually, instances 
of <tt>ISpellCheckEngine</tt> are implemented as singletons.</li>
<li>Create the appropriate dictionaries that should be used during the spell-check process. All dictionaries that 
can be registered with <tt>ISpellCheckEngine</tt> must implement the interface <tt>ISpellCheckDictionary</tt>. 
For this interface, an abstract implementation is provided in the class <tt>AbstractSpellDictionary</tt>.
Depending on the language of the words contained in this dictionary, custom algorithms for the phonetic hash
 (<tt>IPhoneticHashProvider</tt>) and the edit distance (<tt>IPhoneticDistanceAlgorithm</tt>) should be implemented 
 and registered with the dictionary.</li>
 <li>Instances of spell-checkers can now be created by calling <tt>createSpellChecker(Locale)</tt>, where the locale 
 denotes the language that the spell-checker should use while executing.</li>
</ul>
When requesting a new spell-checker with a different locale via <tt>createSpellChecker(Locale)</tt>, the spell-checker is 
reconfigured with the new dictionaries. More concretely, the old dictionary is unregistered and a new one registered for the 
desired locale is associated with the spell-checker. If no such dictionary is available, no spell-checker is returned and 
the locale of the engine is reset to its default locale.
<h3>
Dictionaries</h3>
Dictionaries are the data structures to hold word lists for a particular language. All implementations of dictionaries must 
implement the interface <tt>ISpellDictionary</tt>. It provides support for life-cycle management as well as the facility to query 
words from the list, add words to the list and get correction proposals for incorrectly spelt words.
<p>
This package provides a default implementation of a dictionary (<tt>AbstractSpellDictionary</tt>) that uses algorithms 
convenient for english languages. <br>
Every dictionary needs two kinds of algorithms to be plugged in:
<ul>
<li>An edit distance algorithm: Edit distance algorithms implement the interface <tt>IPhoneticDistanceAlgorithm</tt>. The algorithm 
is used to determine the similarity between two words. This package provides a default implementation for languages using the latin alphabet (<tt>DefaultPhoneticDistanceAlgorithm</tt>). 
The default algorithm uses the Levenshtein text edit distance.</li>
<li>A hash algorithm: Phonetic hash providers implement the interface <tt>IPhoneticHashProvider</tt>. The purpose of 
phonetic hashes is to have a representation of words which allows comparing it to other, similar words. This package provides a default 
implementation which is convenient for slavic and english languages. It uses the double metaphone algorithm by published 
Lawrence Philips.</li>
</ul>
By plugging in custom implementations of one or both of these algorithms the abstract implementation <tt>AbstractSpellDictionary</tt> can 
be customized to specified languages and alphabets.
<h3>
Spell Check Iterators</h3>
Instances of <tt>ISpellChecker</tt> are usually language-, locale- and medium independent implementations and therefore need an input provider. The 
interface <tt>ISpellCheckIterator</tt> serves this purpose by abstracting the tokenizing of text media to a simple iteration. The actual spell-check process 
is launched by calling <tt>ISpellChecker#execute(ISpellCheckIterator)</tt>. This method uses the indicated spell-check iterator to determine the 
words that are to be spell-checked. This package provides no default implementation of a spell-check iterator.
<h3>
Event Handling</h3>
To communicate the results of a spell-check pass, spell-checkers fire spell events that inform listeners about the status 
of a particular word being spell-checked. Instances that are interested in receiving spell events must implement 
the interface <tt>ISpellEventListener</tt> and register with the spell-checker before the spell-check process starts.<p>
A spell event contains the following information:
<ul>
<li>The word being spell-checked</li>
<li>The begin index of the current word in the text medium</li>
<li>The end index in the text medium</li>
<li>A flag whether this word was found in one of the registered dictionaries</li>
<li>A flag that indicates whether this word starts a new sentence</li>
<li>The set of proposals if the word was not correctly spelt. This information is lazily computed.</li>
</ul>
Spell event listeners are free to handle the events in any way. However, listeners are not allowed to block during 
the event handling unless the spell-checking process happens in another thread.
</body>
</html>