JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD B.Tech III Year INFORMATION RETRIEVAL SYSTEMS Previous Question Paper (September - 2021 - R18)

A Data Warehouse is a centralized repository designed to store, manage, and analyze large volumes of historical data from multiple heterogeneous sources to support business intelligence and decision-making processes. Unlike operational databases, data warehouses are optimized for Online Analytical Processing (OLAP) rather than transactional processing.

• Subject-Oriented: Data is organized around major subjects of the enterprise (e.g., customers, products, sales) rather than specific application functions.
• Integrated: Data is collected from disparate sources, cleaned, transformed, and standardized to ensure consistency in naming conventions, measurement units, and attributes.
• Time-Variant: Data is stored with a historical perspective, often spanning years, allowing analysts to identify trends and patterns over time.
• Non-Volatile: Once data enters the warehouse, it is generally not updated or deleted. It is read-only, ensuring a stable environment for analysis.

Common architectures include:

• Enterprise Data Warehouse (EDW): A centralized system providing a unified view of the entire organization's data.
• Data Mart: A subset of the data warehouse focused on a specific business line or department, such as marketing or finance.
• Virtual Data Warehouse: A logical view of data that remains in its source systems, accessed via middleware without physical consolidation.

The relationship between Information Retrieval (IR) Systems and Database Management Systems (DBMS) is defined by their distinct objectives, data structures, and query mechanisms. While both manage information, they serve different user needs.

• Data Structure: DBMS typically handles structured data organized in tables with rigid schemas. IR systems are designed for unstructured or semi-structured data, such as full-text documents, web pages, and multimedia files.
• Query Nature: DBMS queries are precise and deterministic (e.g., SQL 'SELECT' statements). IR queries are often vague, keyword-based, and probabilistic, relying on relevance ranking rather than exact matching.
• Retrieval Criteria: In a DBMS, a record either satisfies the query condition or it does not (Boolean logic). In IR, the system returns a ranked list of documents based on a similarity score, acknowledging that relevance is subjective.
• Performance Metrics: DBMS performance is measured by transaction throughput and data integrity (ACID properties). IR performance is measured by Precision (the proportion of retrieved documents that are relevant) and Recall (the proportion of relevant documents that were successfully retrieved).

Modern systems often exhibit a convergence, where DBMS incorporate full-text search capabilities (e.g., indexing blobs) and IR systems utilize database-like metadata management to improve retrieval efficiency.