Information Retrieval Assignment

This project is part of an Information Retrieval (CS60092) assignment at IIT Kharagpur. (A PDF of the work can be found in the repository)

Programming Languages Used

• Python (Version - 3.8.10)

Libraries Used

• re
• sys
• string
• nltk
• pickle
• NumPy In case of any missing library, kindly install it using
  • pip3 install < library name > (for Python) (Some libraries mentioned above come as part of python3)

Assignment 1

We created an inverted index from the documents. Using the query, fetched the relevant documents from the inverted index.

Assignment 2

The results of the query are evaluated against gold standard relevance.

Other Details

• Length of the inverted index (Vocabulary) - 8813
• Number of queries that returned document IDs - 14 out of 225

Few evaluation metrics

• lnc.ltc
  1. Mean Average Precision (MAP) @ 10 = 0.4509512345679012
  2. Mean Average Precision (MAP) @ 20 = 0.40917921521906336
  3. Mean Normalized Discounted Cumulative Gain (NDCG) @ 10 = 0.35613102024509247
  4. Mean Normalized Discounted Cumulative Gain (NDCG) @ 20 = 0.387901807896173
• lnc.Ltc
  1. Mean Average Precision (MAP) @ 10 = 0.4509512345679012
  2. Mean Average Precision (MAP) @ 20 = 0.40917921521906336
  3. Mean Normalized Discounted Cumulative Gain (NDCG) @ 10 = 0.35613102024509247
  4. Mean Normalized Discounted Cumulative Gain (NDCG) @ 20 = 0.387901807896173
• anc.apc
  1. Mean Average Precision (MAP) @ 10 = 0.4321748383303936
  2. Mean Average Precision (MAP) @ 20 = 0.3955611461367082
  3. Mean Normalized Discounted Cumulative Gain (NDCG) @ 10 = 0.34492256327874254
  4. Mean Normalized Discounted Cumulative Gain (NDCG) @ 20 = 0.37787975030561444

Purpose

To understand how to create an inverted index, retrieve relevant documents as part of the query and various evaluation metrics.

References

1. http://web.stanford.edu/class/cs276/
2. https://slideplayer.com/slide/238597/1/images/41/tf-idf+weighting+has+many+variants.jpg