Skip to content

Latest commit

 

History

History
44 lines (28 loc) · 1.65 KB

README.md

File metadata and controls

44 lines (28 loc) · 1.65 KB
Raw

Information Retrieval Engine

Implementation of an information retrieval engine as AUT IR course project.

Instructor: Dr. A. Nickabadi

Semester: Spring 2022

Usage

See the help command by entering python main.py --help.

usage: MyIRSystem [-h] [-hl] [-zl] [-l index.pkl | -s index.pkl] [-r | -b] collection.json

positional arguments:
  collection.json       collection .json file path

optional arguments:
  -h, --help            show this help message and exit
  -hl, --heaps-law      demonstrate heaps law
  -zl, --zipf-law       demonstrate zipf law
  -l index.pkl, --load index.pkl
                        load previously saved index
  -s index.pkl, --save index.pkl
                        save built index
  -r, --ranked          ranked retrieval using tf-idf
  -b, --boolean         boolean retrieval using positional index

If the inverted index is saved beforehand, it can be loaded using --load switch. Otherwise, it will build a new index, which could be saved by --save switch.

There are two operation modes; boolean and ranked. In ranked mode, documents are scored based on a tf-idf weighting scheme. In boolean mode, the number of terms' occurrences is used to rank the results. Supported operators for boolean mode are:

  • And: The default operator by using space between terms in a query; A B
  • Not: By using ! before a term; A ! B
  • Phrase: By using " around the phrase; "A B"

A sample input collection is available here.

Preprocessing and tokenization are done by the hazm library (for the Persian language).