Évaluation de la pertinence (question ↔ article juridique) en français. Pipeline complet (prépa → modèles → soumission) avec CamemBERT en bi-encodeur calibré (MSE/Spearman), + variantes cross-encoder.

한국어/영어 발음 유사도 기반 매핑 시스템. ASR 결과를 데이터베이스 용어와 정확히 매칭해주는 오픈소스 도구입니다.한국어-영어 간 발음 변환, Levenshtein 거리 기반 매칭, 사용자 정의 매핑을 지원하여 음성 인터페이스 정확도를 높입니다

code for paper https://arxiv.org/pdf/2510.00324 to be presented at sigir-ap 2025

AI-powered semantic search engine for products. FastAPI + MySQL + sentence-transformers.

Model that converts e-commerce product descriptions into natural user search queries.

NoMIRACL: A multilingual hallucination evaluation dataset to evaluate LLM robustness in RAG against first-stage retrieval errors on 18 languages.

REST service implemented in Python using FastAPI that allows for training and querying a model for object relevance.

Tezeta is a Python package designed to optimize memory in chatbots and Language Model (LLM) requests using relevance-based vector embeddings. This, in essence, provides support for using much longer conversations and text requests than supported by the context window.

a repo for describing how different selections of degrees of freedom affect the amount of information retained about the full system

Python project for generating search results based on a given text using TF Universal Sentence Encoder and ElasticSearch database.

Solr Relevance Ranking Analysis and Visualization Tool

Content-based Video Relevance Prediction on "cold-start" problems. Recommending new uploaded videos to users based on the extracted features.

Search engine vector voting

Evaluation Measures for Relevance and Credibility in Ranked Lists

Python implementation of BM25 function for document retrieval

A tool for retrieving relevant content from the Web using Google's Custom Search API

SIGIR 2017 Candidate Selection Tutorial (http://sigir.org/sigir2017/program/tutorials/#candidate)

A search query is processed using NLP after which the results are displayed by crawling the web for relevant web pages & displaying them after normalizing the scores

What is the nicest way to crowdsource IR relevance judgements?