##Overview This project implements the following basic algorithms for document-query similarity computation and ranking:

• Overlap: uses the overlap between non-capitalized words to rank the documents relative to a given query
• Tf-idf: uses tf-idf as the weighting function for words and the weighted sum as the similarity function between the document and the query s(Q, D)
• Rocchio: As in 2, this algorithm uses td-idf and the weighted sum formula for an initial ranking, but then optimizes the original query using Rocchio’s algorithm

report.pdf contains further description of the implementation and key results.

##How to run the code All code is found in the ranking/ folder. The output of each algorithm is written to output_ranking/ folder into a file with name 'algo_name'.top.

The performance of each of the algorithms can be evaluated using the trec_eval program, also included under ranking/. The script compares the results printed out by the search algorithm into a .top file to the relevant documents included in truth.rel. To perform the evaluation, from the command line run:

• ranking/trec_eval -o -c -M1000 data/truth.rel data/overlap.top

The program prints a number of standard effectiveness measures for each algorithm.