Jawaharlal Nehru Technological University Hyderabad MCA V INFORMATION RETRIEVAL SYSTEMS Previous Question Paper (December - 2019 - R17)

Information Retrieval (IR) models are categorized based on how they represent documents and queries, and how they calculate relevance. The taxonomy is primarily divided into three classes:

• Boolean Model: Based on set theory and Boolean algebra, where documents are retrieved if they satisfy exact logical conditions (AND, OR, NOT).
• Vector Space Model: Represents documents and queries as vectors in a multi-dimensional space, calculating relevance via similarity measures like Cosine Similarity.
• Probabilistic Model: Estimates the probability that a document is relevant to a query based on the probability distribution of terms, often using the Binary Independence Retrieval (BIR) framework.

Document clustering is the process of grouping similar documents into clusters without predefined labels, making it an unsupervised learning task. The primary goal is to organize large collections to improve retrieval efficiency and browsing experience. Key approaches include:

• Hierarchical Clustering: Builds a tree-like structure of clusters (dendrogram) using either agglomerative (bottom-up) or divisive (top-down) strategies.
• Partitional Clustering: Divides documents into a fixed number of non-overlapping groups, such as the K-means algorithm, which iteratively refines cluster centroids to minimize intra-cluster variance.