Analysis of Ranking Models for Information Retrieval in Hindi

Suneet Kumar Gupta1, Vikas Kumar2, Gopal Gupta3 and Gaurav Kansal4

1NRI Institute of Information Science and Technology, Bhopal, India.

2Dhirubhai Ambani Institute of Information and Communication Technology (DAIICT), Gandhinagar, Gujarat, India.

3Bansal Institute of Science & Technology, Bhopal, India.

4ABES Engineering College, Ghaziabad.


As we know that there are many models which are used to rank the documents while we retrieve the data from a given corpus. Ranking of document is very necessary in information retrieval because ranking shows that which document of the corpus is how much relevance with our given query. Five models have been used for the experiments. For performing the experiment 122586 documents of UTF (Universal Text Format), with 50 queries has been used. And also there is need of another file known as Qrels (Query relevance), which shows the relevance between documents and queries. With the help of these three files different model making the ranking of documents in terrier open source. Experimentally it has found that model PL2c has given least Mean Average Precision, which is .0010 and the best Mean Average Precision has given by TF_IDF which is .1627. All the models have been compared for the ranking for Hindi Corpus, Hindi Topics and Hindi qrels.

     Back to TOC