1. Explain briefly about hierarchy of clusters.

A hierarchy of clusters is a way of grouping documents or data items into clusters that are
arranged in levels, like a tree. There are two main types:

• Agglomerative (bottom-up): Start with each item as its own cluster; merge the
closest clusters step by step.

• Divisive (top-down): Start with all items in one cluster; split into smaller clusters step by
step.

Advantages:

• No need to decide the number of clusters in advance.

• Gives a clear tree structure (dendrogram) which helps in understanding the data at
different levels.

• Flexible: can use different distance measures and linkage methods.

Disadvantages:

• Slow and not suitable for very large datasets (computationally intensive).

• Sensitive to noise and outliers, which can affect the cluster quality.

• Once clusters are merged or split, the decision cannot be changed.

• The results can change depending on the linkage method or distance measure used.

2. Illustrate information visualization techniques.

Information visualization techniques help users see and understand data better. Common techniques
include:

• Dendrograms: Tree diagrams that show how clusters are formed or split at each step in
hierarchical clustering.

• Scatter Plots: Show documents as points in 2D or 3D space based on their features, helping
to spot clusters or outliers.

• Heatmaps: Use colors to show the strength of relationships or similarities between items.

• Network Graphs: Show documents as nodes and similarities as links, making it easy to see
connections.

Advantages:

• Makes complex data easier to understand.

• Helps users spot patterns, trends, and relationships quickly.

Disadvantages:

• Can become cluttered or hard to read with very large datasets.

• May require some training or experience to interpret correctly.

Search statements are the queries or expressions that users create to tell the information retrieval
system what they are looking for. These statements can be simple keywords, phrases, or more
complex queries using Boolean operators like AND, OR, and NOT.

Bandings refer to the way search results are grouped or organized based on certain criteria, such as
relevance, similarity, or importance. For example, after a search, the system may present results in
bands like "highly relevant," "moderately relevant," and "less relevant." This helps users focus on the
most useful results first.

The process of binding is also important. Binding means connecting the user's search statement to
the system's vocabulary and database. This involves:

• Interpreting the user's words and mapping them to the terms used in the database.

• Assigning weights or importance to certain terms if the system allows.

• Translating the user's query into the system's internal language for processing.

4. Discuss about similarity measures

Similarity measures are mathematical methods used to determine how closely two documents,
queries, or items are related in an information retrieval system. They help the system rank and
retrieve the most relevant documents for a user's search.

Some common similarity measures include:

• Cosine Similarity: Calculates the cosine of the angle between two document vectors. If the
angle is small (cosine value close to 1), the documents are very similar.

• Jaccard Similarity: Compares the number of common terms between two documents to the
total number of unique terms in both documents.

• Euclidean Distance: Measures the straight-line distance between two points (documents) in
a multi-dimensional space; smaller distance means more similarity.

5. Explain the concept of cognition & perception with example

Cognition refers to the mental processes involved in understanding, thinking, learning, and
remembering information.

Perception is about how we receive and interpret information through our senses, such as seeing,
hearing, or touching.

In the context of information retrieval systems, both cognition and perception play important roles in
how users search for and interact with information.

Example:
Suppose a user is looking for a specific book in a library.

• Perception helps the user see the book titles and covers on the shelves, recognize the colors,
and read the labels.

• Cognition helps the user remember the author's name, understand the classification system,
and decide which book matches their need.
6. Discuss about Selective Dissemination

Selective Dissemination of Information (SDI) is a personalized information service provided by

How SDI Works:

• User Profile Creation: Each user creates a profile specifying their interests, keywords,
subjects, or topics.

• Continuous Monitoring: The system continuously monitors new documents, articles, or data
added to the database.

• Matching: Whenever new information matches a user’s profile, the system automatically
selects it.

• Delivery: The relevant information is sent directly to the user, often through email, alerts, or
a customized dashboard.

Example:

A medical researcher interested in “diabetes treatment” registers their interest with a digital library.
Whenever new research papers or articles about diabetes treatment are added, the system
automatically sends notifications or emails to the researcher.

Uses:

• Academic research updates

• Corporate intelligence gathering

• News alerts for journalists or analysts

7. Explain software text search algorithms

Software text search algorithms are methods used by computers to find specific words, phrases, or
patterns in large volumes of text quickly and efficiently. These algorithms are the backbone of search
engines, text editors, and database search functions.

Common Text Search Algorithms:

1. Brute-Force Search

• How it works: Checks every possible position in the text for the search pattern.

• Use case: Simple and works for small texts, but slow for large data.

2. Knuth-Morris-Pratt (KMP) Algorithm

• How it works: Uses information from previous matches to skip unnecessary comparisons,
making the search faster.

• Advantage: Efficient for long texts and repeated patterns.

3. Boyer-Moore Algorithm
 • How it works: Starts matching the pattern from the end, and skips sections of the text when
mismatches occur.

• Advantage: Very fast in practice, especially for large texts.

4. Rabin-Karp Algorithm

• How it works: Uses hash functions to compare the pattern with substrings in the text.

• Advantage: Good for searching multiple patterns at once.

Applications:

• Search engines: To quickly find web pages containing user queries.

• Text editors: For “Find” and “Replace” functions.

• Database systems: For searching within large tables or documents.

8. Explain the concept of image and video retrieval in IRS

Image and video retrieval in Information Retrieval Systems (IRS) refers to the process of searching for
and finding relevant images or videos from a large database based on a user's query.

The system uses various features to match the user's query with stored media:

• For images: It analyzes visual features like color, shape, texture, and patterns. For example, if
a user uploads a picture of a flower, the system searches for similar images using these
features.

• For videos: It looks at features like motion, scene changes, objects, and sometimes audio.
Users can search for videos by entering keywords, uploading a sample image, or even
providing a short video clip.

Modern systems may also use machine learning and deep learning techniques to improve the
accuracy of matching and retrieval. Image and video retrieval is widely used in digital libraries,
security systems, medical imaging, and multimedia search engines.