The fine-tuned_movie_retriever-bge-base-en-v1.5 open-source model - Customized for movie recommendation systems to optimize vector retrieval.

Fine Tuned Movie Retriever Bge Base En V1.5

Developed by JJTsao

A fine-tuned sentence transformer model customized for movie and TV show recommendation systems, optimizing high-quality vector retrieval in the RAG pipeline.

Text Embedding Open Source License:Apache-2.0 #Film and TV semantic retrieval #Atmosphere and style matching #RAG optimization

Downloads 540

Release Time : 5/5/2025

Model Overview

This model is optimized for the retrieval-augmented generation (RAG) pipeline in film and TV recommendations, supporting rich atmosphere and style and natural language queries based on metadata, and adopting an intelligent negative sampling strategy.

Model Features

Support for rich atmosphere and style queries

Capable of understanding and processing natural language queries describing the atmosphere and style of movies, such as 'Emotionally intense space exploration movies with themes of love and sacrifice'.

Optimized metadata query

Optimized natural language queries based on metadata such as directors, eras, and genres, such as 'Any crime movies about robberies directed by Quentin Tarantino in the 1990s?'.

Intelligent negative sampling

Adopting more intelligent negative sampling strategies such as type contrast, theme mismatch, and star-theme confusion to improve retrieval quality.

Large-scale training data

Trained using approximately 32,000 synthetic natural language queries across metadata and atmosphere style prompts.

Model Capabilities

Film semantic retrieval

TV show semantic retrieval

Natural language query understanding

Vector similarity calculation

Use Cases

Recommendation system

RAG-style movie recommendation

Serving as a high-quality vector retrieval component in the retrieval-augmented generation pipeline

Recall@1 reaches 0.456, significantly better than 0.214 of the base model

Semantic filtering of movie catalogs

Semantic-based filtering and retrieval of large movie catalogs

Query-document re-ranking

Used for re-ranking queries and documents in the retrieval pipeline

🚀 Fine-Tuned Movie Retriever (Rich Semantic & Metadata Queries + Smart Negatives)

This custom fine-tuned sentence-transformer model is designed for movie and TV recommendation systems. It's optimized for high-quality vector retrieval in a movie and TV show recommendation RAG pipeline. The fine-tuning process used approximately 32K synthetic natural language queries across metadata and vibe-based prompts, including enriched vibe-style and metadata-based natural language queries, smarter negative sampling, and a dataset of over 32,000 triplets.

🚀 Quick Start

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("jjtsao/fine-tuned_movie_retriever-bge-base-en-v1.5")
query_embedding = model.encode("mind-bending sci-fi thrillers from the 2000s about identity")

✨ Features

Enriched Queries: Supports enriched vibe-style natural language queries (e.g., Emotionally powerful space exploration film with themes of love and sacrifice) and metadata-based natural language queries (e.g., Any crime movies from the 1990s directed by Quentin Tarantino about heist).
Smarter Negative Sampling: Utilizes smarter negative sampling methods such as genre contrast, theme mismatch, and star - topic confusion.
Large Dataset: Trained on a dataset of over 32,000 triplets (query, positive doc, negative doc).

📦 Installation

No specific installation steps are provided in the original README. If you want to use this model, you can install the sentence-transformers library using pip install sentence-transformers.

💻 Usage Examples

Basic Usage

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("jjtsao/fine-tuned_movie_retriever-bge-base-en-v1.5")
query_embedding = model.encode("mind-bending sci-fi thrillers from the 2000s about identity")

📚 Documentation

Training Details

Base model: BAAI/bge-base-en-v1.5
Loss function: MultipleNegativesRankingLoss
Epochs: 4
Optimized for: top - k semantic retrieval in RAG systems

Evaluation: Fine - Tuned vs Base Model

Metric	Fine - Tuned Model Score	Base Model Score
Recall@1	0.456	0.214
Recall@3	0.693	0.361
Recall@5	0.758	0.422
Recall@10	0.836	0.500
MRR	0.595	0.315

Evaluation setup:

Dataset: 3,598 held - out metadata and vibe - style natural queries
Method: Top - k ranking using cosine similarity between query and positive documents
Goal: Assess top - k retrieval quality in recommendation - like settings

Ideal Use Cases

RAG - style movie recommendation apps
Semantic filtering of large movie catalogs
Query - document reranking pipelines

📄 License

Apache 2.0 — open for personal and commercial use.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご