JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD MCA 2022 INFORMATION RETRIEVAL SYSTEMS Previous Question Paper (March/April - 2022 - R20)

The Vector Space Model (VSM) is an algebraic model used in information retrieval to represent documents and queries as vectors in a multi-dimensional space. Each dimension corresponds to a unique term in the collection vocabulary.

The core components and process of the Vector Space Model are as follows:

• Term Representation: A document is represented as a vector D = (w₁, w₂, ..., w_n), where w_i represents the weight of term i in the document. Weights are typically calculated using TF-IDF (Term Frequency-Inverse Document Frequency) to reflect the importance of a term within a document relative to the entire collection.
• Weighting Scheme: TF-IDF is defined as:
  w_i,j = tf_i,j × log(N / df_i)
  where tf_i,j is the frequency of term i in document j, N is the total number of documents, and df_i is the number of documents containing term i.
• Similarity Measurement: To rank documents, the model measures the similarity between the query vector Q and document vector D. The most common metric is Cosine Similarity, which calculates the cosine of the angle between the two vectors:
  Similarity(Q, D) = (Q · D) / (|Q| × |D|)
  This measure is independent of document length, as it focuses on the orientation of the vectors rather than their magnitude.
• Ranking: Documents are ranked in descending order of their similarity scores to the query.

The advantages of VSM include its ability to provide partial matching, ranked retrieval results, and the flexibility to adjust term weights. However, it assumes term independence, meaning it fails to capture semantic relationships between terms (synonymy) or the specific order of words in a document.

The Non-Binary Independence Model is an extension of the Binary Independence Retrieval (BIR) model that accounts for term frequency (TF) information within documents. While the basic BIR model treats documents as binary vectors (presence or absence of a term), the non-binary model incorporates the frequency of terms to better estimate the probability of relevance.

In this model, the probability of a document being relevant is conditioned not just on the presence of terms, but on their specific counts. The model assumes that the distribution of term frequencies in relevant and non-relevant documents follows specific statistical distributions, typically the Poisson or Negative Binomial distribution.

The core components of the model include:

• Term Frequency Modeling: Instead of a binary indicator variable, the model uses a random variable representing the frequency of term t in document d.
• Independence Assumption: It retains the core assumption of the independence model, which posits that terms occur independently of one another within the classes of relevant and non-relevant documents.
• Weighting Function: The retrieval status value (RSV) is calculated by summing the weights of query terms. The weight for a term t is derived from the ratio of the probability of observing a specific frequency f in a relevant document to the probability of observing that same frequency in a non-relevant document.

The weight of a term t with frequency f can be expressed as:

Where P(f | R) is the probability of term t appearing f times in a relevant document, and P(f | NR) is the probability in a non-relevant document. By using these distributions, the model effectively captures the intuition that a higher frequency of a query term in a document increases the likelihood of that document being relevant to the user's information need. This approach provides a more nuanced ranking mechanism than the binary model, as it distinguishes between documents that merely contain a term and those where the term is central to the content.