{"id":338,"date":"2025-12-08T00:14:04","date_gmt":"2025-12-08T00:14:04","guid":{"rendered":"http:\/\/ijeesoo.com\/?page_id=338"},"modified":"2025-12-08T01:37:16","modified_gmt":"2025-12-08T01:37:16","slug":"probabilistic-information-retrieval","status":"publish","type":"page","link":"http:\/\/ijeesoo.com\/?page_id=338","title":{"rendered":"Probabilistic Information Retrieval"},"content":{"rendered":"\n
\n
\n



 Foundations and Language Models<\/h1>\n<\/div>\n\n\n\n
\"\"<\/figure>\n<\/div>\n\n\n\n
\n
\n

Probabilistic Information Retrieval<\/strong><\/h2>\n\n\n\n
<\/div>\n\n\n\n

Foundations<\/p>\n<\/div>\n\n\n\n

\n
\n

Probabilistic Information Retrieval<\/strong><\/h3>\n\n\n\n
    \n
  • Motivation: Why Probability in IR?<\/li>\n\n\n\n
  • Contrast with Boolean & Vector Space Models<\/li>\n\n\n\n
  • Key Assumptions in Probabilistic IR<\/li>\n\n\n\n
  • Probability Basics<\/li>\n\n\n\n
  • Example: Probability of Relevance for a document<\/li>\n<\/ul>\n<\/div>\n\n\n\n
    \n

    Probability Ranking Principle (PRP)<\/h3>\n\n\n\n
      \n
    • Definition of PRP<\/li>\n\n\n\n
    • Intuition: Ranking by Probability of Relevance<\/li>\n\n\n\n
    • Case 1: 1\/0 Loss Scenario<\/li>\n\n\n\n
    • Case 2: Incorporating Retrieval Costs<\/li>\n\n\n\n
    • Worked Example: PRP Applied to a Query<\/li>\n\n\n\n
    • Limitations of PRP<\/li>\n<\/ul>\n<\/div>\n\n\n\n
      \n

      Binary Independence Model (BIM)<\/h3>\n\n\n\n
        \n
      • Assumptions of BIM (Term Independence)<\/li>\n\n\n\n
      • Deriving Ranking Function \u2013Step 1<\/li>\n\n\n\n
      • Deriving Ranking Function \u2013Step 2<\/li>\n\n\n\n
      • Formula for Ranking Score (Log-Odds Weight)<\/li>\n\n\n\n
      • Probability Estimates in Theory<\/li>\n\n\n\n
      • Probability Estimates in Practice<\/li>\n\n\n\n
      • Example Calculation<\/li>\n\n\n\n
      • Relevance Feedback in Probabilistic Models<\/li>\n\n\n\n
      • Rocchio vs Probabilistic Feedback<\/li>\n\n\n\n
      • Strengths and Weaknesses of BIM<\/li>\n<\/ul>\n\n\n\n


        <\/p>\n<\/div>\n<\/div>\n\n\n\n

        \n
        \n

        Extensions & Modern Models<\/h3>\n\n\n\n
          \n
        • Appraisal of Probabilistic Models<\/li>\n\n\n\n
        • Tree-Structured Dependencies<\/li>\n\n\n\n
        • Okapi BM25: Motivation and Formula<\/li>\n\n\n\n
        • BM25 Example with Query\/Document Scoring<\/li>\n\n\n\n
        • Bayesian Network (BN) Approaches to IR<\/li>\n\n\n\n
        • Summary of Probabilistic IR<\/li>\n<\/ul>\n<\/div>\n\n\n\n
          \n

          Language Models for IR<\/strong><\/h3>\n\n\n\n
            \n
          • Definition of Language Models in IR<\/li>\n\n\n\n
          • Finite Automata and Language Models<\/li>\n\n\n\n
          • Types of Language Models<\/li>\n\n\n\n
          • Multinomial Distributions over Words<\/li>\n\n\n\n
          • Example: Unigram Language Model<\/li>\n<\/ul>\n<\/div>\n\n\n\n
            \n

            Query Likelihood Model<\/h3>\n\n\n\n
              \n
            • Query Likelihood Principle<\/li>\n\n\n\n
            • Using Query Likelihood in IR<\/li>\n\n\n\n
            • Estimating Query Generation Probability<\/li>\n\n\n\n
            • Maximum Likelihood Estimation (MLE)<\/li>\n\n\n\n
            • Smoothing Techniques<\/li>\n\n\n\n
            • Ponte & Croft\u2019s Experiments \u2013Setup<\/li>\n\n\n\n
            • Ponte & Croft\u2019s Experiments \u2013Results<\/li>\n\n\n\n
            • Lessons Learned from Query Likelihood<\/li>\n<\/ul>\n<\/div>\n<\/div>\n\n\n\n
              \n
              \n

              Comparisons & Extensions<\/h3>\n\n\n\n
                \n
              • Language Modelling vs Probabilistic IR<\/li>\n\n\n\n
              • Language Modelling vs VSM<\/li>\n\n\n\n
              • Strengths of LM Approaches<\/li>\n\n\n\n
              • Weaknesses of LM Approaches<\/li>\n\n\n\n
              • Extended LM Approaches: Mixture Models<\/li>\n\n\n\n
              • Topic Models (e.g., LDA) in IR<\/li>\n\n\n\n
              • Neural Language Models in Modern IR (Word Embeddings, Transformers)<\/li>\n\n\n\n
              • Example: BERT in Document Ranking<\/li>\n<\/ul>\n<\/div>\n\n\n\n
                \n

                Wrap-Up<\/strong><\/h3>\n\n\n\n