Document Expansion by Query Prediction

[KeyurD1010]

Title

Document Expansion by Query Prediction

Team Name

The Indexers

Email

keyurdhanani456@gmail.com

Team Member 1 Name

Keyur Dhanani

Team Member 1 Id

202311045

Team Member 2 Name

Rameez Raja

Team Member 2 Id

202311068

Team Member 3 Name

Deepak Khatri

Team Member 3 Id

202311042

Team Member 4 Name

Nirmal Shah

Team Member 4 Id

202311043

Category

Evaluation Track Problem

Problem Statement

The problem is improving search engine retrieval effectiveness by expanding documents with predicted queries. This can be done using a sequence-to-sequence model trained on query-document pairs, which helps enhance document relevance and retrieval speed without relying on computationally expensive re-ranking methods.

Evaluation Strategy

MAP and MRR@10.

Dataset

MS MARCO #:- https://huggingface.co/datasets/microsoft/ms_marco TREC-CAR : - https://ir-datasets.com/car.html

Resources

Paper Title: - Document Expansion by Query Prediction

[parth126]

• This can be split into two independent problems
  1. Optimize indexing (or any other part of IR pipeline)
  2. Build a better matching algorithm (mostly LLM based) to solve VMware specific problem

The linked paper is relatively weak and not recommended for the project.

[parth126]

The project is changed to "Document expansion using query prediction" Approval is pending until the proposal is updated.

[KeyurD1010]

Respected sir,

As instructed via email, we have updated the github issue.

Project Title : - Document Expansion by Query Prediction Dataset:- MS MARCO and TREC-CAR Evaluation matrix :- MAP and MRR@10

[parth126]

Looks good now.