Query Expansion and Synonym Recognition

CS 410 Text Information Systems, 2024 Fall

• Lisi Lei (lisilei2@illinois.edu)
• Ivy He (qiuhanh2@illinois.edu)
• Hilda Liu (hildal2@illinois.edu)
• Puyu Yang (pyang33@illinois.edu)
• Meilin Liu (meilinl2@illinois.edu)

Abstract

This project provides a web-based interface for exploring query expansion techniques through synonym recognition. By leveraging various NLP models and frameworks, users can input a query term, choose an algorithm, and retrieve expanded queries with semantically similar terms. This tool is designed to showcase the practical applications of synonym recognition and query expansion in text-based information retrieval.

Introduction

Effective information retrieval often hinges on understanding and expanding user queries to include related terms. This project demonstrates how synonym recognition can be integrated with query expansion using state-of-the-art NLP models and frameworks. Users can select between multiple backend algorithms to observe the impact of different approaches, such as statistical and neural-based methods, on synonym retrieval.

The application integrates:

• NLTK for traditional synonym matching.
• ConceptNet combined with Sentence BERT for semantic similarity.
• Llama-Vision-Free (via the Together API) for large language model-based synonym generation.

The system is designed to provide a user-friendly interface while demonstrating advanced back-end processing capabilities.

Backend

Tech Stack

• Python: Version 3.12.2
• Flask: Web framework for serving the application
• NLTK: For lexical-based synonym retrieval
• ConceptNet API: Knowledge graph for semantic relationships
• Sentence BERT: Pretrained model for similarity scoring
• Together API: Integration for the Llama-Vision-Free LLM

Core Functionality

1. Query Processing: Accepts a user-inputted query and extracts the main term.
2. Synonym Retrieval:
   • NLTK: Provides synonyms from lexical resources.
   • ConceptNet + Sentence BERT: Retrieves related terms and ranks them based on semantic similarity.
   • Llama-Vision-Free: Uses an LLM for dynamic synonym generation.
3. Query Expansion: Outputs the original query expanded with the retrieved synonyms.

Frontend

Features

• A responsive and intuitive web page for user interaction.
• Input fields for the query term and maximum number of synonyms.
• Algorithm selection via radio buttons:
  • BERT-based retrieval
  • LLM-based retrieval
  • NLTK-based retrieval
• Dynamically generated results displayed as a list with options to copy individual or all results.

Installation Requirements

Prerequisites

To get started with the Query Expansion Tool, ensure you have the following installed on your machine:

• Python: Version 3.12.2 or later
• pip: Python package manager

Backend Installation

1. Clone the repository:

   git clone https://github.com/0lisil0/query-expansion-and-synonym-recognition

2. Run the following command in the project directory to install dependencies:

pip install -r requirements.txt

Environment Setup for LLM Integration

This project supports Llama-Vision-Free through the Together API for LLM-based synonym retrieval.

1. Create an account (or log in if you already have one) on Together AI.
2. To temporarily set the environment variable, use the following command in PowerShell:

   $env:TOGETHER_API_KEY="your_actual_api_key_here"
   

   Alternatively, you can save the key in a .env file and load it from there.

Example usage of LLM synonyms finder

synonym_finder = LlamaQuerySynonymFinder(model_name="meta-llama/Llama-Vision-Free")
query = "happy"
synonyms = synonym_finder.get_synonyms(query, num_synonyms=5)
print(f"synonyms: {synonyms}")

Usage Instructions

Start the Application

1. Navigate to the project directory.

2. Start the Flask backend server:

   python app.py
   

3. Open your browser and access the application:

• URL: http://127.0.0.1:5000/

Tutorial

• A video tutorial can be found here: https://www.youtube.com/watch?v=Jo5L4WYMV0w
• Slide used for the video is here: https://docs.google.com/presentation/d/1xhyK7P2VIbOAA2pDB1NlCfhPpJOgJRvff1tjUUtrjR8/edit?usp=sharing

Contribution of Team Members

• Backend Implementation (llm_query_expansion.py, nltk_query_expand.py, sentence_bert_query_expand.py): Lisi Lei, Meilin Liu
• Frontend Implementation (index.html): Hilda Liu, Ivy He
• Flask Implementation (index.html, requirements.txt, app.py): Puyu Yang
• Testing: Meilin Liu, Lisi Lei
• Tech Review: Lisi Lei , Meilin Liu, Hilda Liu
• Short Video Presentation: Hilda Liu