word sense systems:

Researchers have explored various system architectures to address the sense disambiguation problem.
We can classify these systems into four main categories:
1. Rule based or Knowledge based
2. Supervised
3. Unsupervised
4. Semi-supervied

1. Rule-based:

-Rule-based systems for word sense disambiguation are among the earliest methods developed to
tackle the problem of determining the correct meaning of a word based on its context.
-These systems rely heavily on dictionaries, thesauri, and handcrafted rules.
ALgorithms and Techniques:
(i) Lesk Algorithm:

One of the simplest and oldest dictionary-based algorithms.

The algorithm assigns the sense of a word that has the most overlap in terms of words with the words
in its context.
Example: If the word "bank" appears in a context with words like "money" and "deposit," the financial
sense of "bank" is chosen.

(ii) Enhanced Lesk Algorithm:

Banerjee and Pedersen extended the Lesk algorithm to include synonyms, hypernyms (more general
terms), hyponyms (more specific terms), and meronyms (part-whole relationships).
This increases the accuracy of overlap measurement and improves disambiguation performance.

(iii) Structural Semantic Interconnections (SSI):

Proposed by Navigli and Velardi.

Constructs semantic graphs using resources like WordNet, domain labels, and annotated corpora.
Uses an iterative algorithm to match the semantic graphs of context words with the target word until
the best matching sense is identified.

Working of Rule-based:

Context Collection
Dictionary/Thesaurus Matching
Weight Computation
Sense Selection

Advantages:

Simple and intuitive approach.

Can be very effective when precise dictionary definitions or thesaurus categories are available.

Limitations:

Heavily reliant on the availability and quality of lexical resources.

Handcrafted rules can be labor-intensive and may not cover all possible contexts.

2. Supervised:
 -Supervised systems for word sense disambiguation use machine learning to train classifiers on
manually annotated datasets.
-These systems typically perform better than unsupervised methods when tested on annotated data but
require significant manual effort for annotation and a predetermined sense inventory.
-Use various machine learning models, like Support Vector Machines (SVMs) and Maximum Entropy
(MaxEnt) classifiers.

Features Used in Supervised Systems:

*Lexical Context:
This feature comprises the words and lemmas of words occuring in the entire paragraph.
*Parts of Speech:
POS tags for words in the context window.
*Local Collocations:
Sequences of nearby words that provide semantic context.
Example: For the word "nail" in "He bought a box of nails," collocation features might include "box_of"
and "of_nails."
*Topic Feature:
The broad topic or domain of the text can indicate the likely sense of a word.

*Additional Rich Features:

Voice of the Sentence: Active, passive, or semi-passive.
Presence of Subject/Object: Whether the word has a subject or object.
Sentential Complement: Whether the word has a complement.
Prepositional Phrase Adjunct: Whether the word has a prepositional phrase.

Advantages:

Typically achieves high accuracy due to the rich feature sets and annotated data.

Limitations:

Requires a large amount of manually annotated data, which is time-consuming and expensive.

3. Unsupervised:

-Unsupervised systems for word sense disambiguation tackle the problem without relying heavily on
manually annotated training data.
-These systems are essential due to the scarcity of labeled data for every sense of each word in a given
language.
-The key strategies include clustering, distance metrics, and leveraging cross-linguistic evidence.

Key approaches:

(i) Group similar instances of a word into clusters, where each cluster represents a different sense of
the word.

(ii) Use a measure of semantic distance to determine the sense of a word by finding how close it is to
known senses in a semantic network like WordNet.

(iii) Start with a few examples of each sense (seeds) and grow these examples into larger clusters.

Advantages:

No need for extensive manual annotation, making it scalable and adaptable to various languages.
 Can discover new senses not present in predefined sense inventories.

Limitations:

May require sophisticated algorithms to achieve high accuracy.

Performance can be lower compared to supervised systems when tested on well-annotated data.

4. Semisupervised:

-Semisupervised systems for word sense disambiguation combine limited labeled data with a larger
pool of unlabeled data to iteratively improve classification performance.
-These methods aim to leverage the strengths of both supervised and unsupervised approaches.

Key Principles:

One Sense per Collocation:

Words that occur in specific syntactic relationships or nearby certain types of words often share the
same sense.
One Sense per Discourse:
Within a given discourse, instances of the same word tend to share the same sense.

Yarowsky Algorithm:

Introduced by Yarowsky, this algorithm is foundational for semisupervised WSD.

It uses initial seed examples to iteratively classify and expand the training set.

Initial Seed Selection

Training
Iteration
Termination
Application

Extensions and Variations

SALAAM Algorithm:

Groups words that translate into the same word in another language, identifies senses using
WordNet proximity, and propagates sense tags across the parallel text.

Unsupervised to Supervised Combination:

Use unsupervised methods to create labeled data, then train supervised models on this data. This
hybrid approach aims to combine the scalability of unsupervised methods with the precision of supervised
learning.

Advantages:

Reduces the need for extensive manual annotation.

Utilizes both labeled and unlabeled data, making it more scalable than purely supervised
approaches.

Limitations:

Performance depends on the quality and representativeness of initial seed examples.

Potential noise from incorrect automatic labeling, though mitigated by constraints like one sense per
discourse.

Performance
Studies have shown semisupervised methods to perform well, often achieving accuracy in the
mid-80% range when tested on standard datasets.

SOFTWARE:

Several software tools developed by the research community are available for performing word sense
disambiguation (WSD).

1. IMS (It Makes Sense)

Description: A complete word sense disambiguation system.

Functionality: Provides end-to-end WSD capabilities using state-of-the-art methods.

2. WordNet-Similarity-2.05

Description: A set of Perl modules for computing various word similarity measures.
Functionality: Facilitates quick computation of word similarity measures based on WordNet.

3. WikiRelate!

Description: A word similarity measure based on Wikipedia categories.

Functionality: Uses the categorization in Wikipedia to measure word similarity.

Key Benefits:

Comprehensive Solutions: Tools like IMS provide complete systems for performing WSD.
Modular Approaches: Tools like WordNet-Similarity offer modular functions for specific tasks such as
measuring word similarity.
Innovative Measures: WikiRelate! leverages the vast categorization in Wikipedia to derive word
similarities, providing a unique approach to WSD.

--------------------------------------------------------------------------------------------------------------------------------------------
------------