DRAFT: Dense Retrieval Augmented Few-shot Topic classifier Framework
–arXiv.org Artificial Intelligence
With the growing volume of diverse information, the demand for classifying arbitrary topics has become increasingly critical. To address this challenge, we introduce DRAFT, a simple framework designed to train a classifier for few-shot topic classification. DRAFT uses a few examples of a specific topic as queries to construct Customized dataset with a dense retriever model. Multi-query retrieval (MQR) algorithm, which effectively handles multiple queries related to a specific topic, is applied to construct the Customized dataset. Subsequently, we fine-tune a classifier using the Customized dataset to identify the topic. To demonstrate the efficacy of our proposed approach, we conduct evaluations on both widely used classification benchmark datasets and manually constructed datasets with 291 diverse topics, which simulate diverse contents encountered in real-world applications. DRAFT shows competitive or superior performance compared to baselines that use in-context learning, such as GPT-3 175B and InstructGPT 175B, on few-shot topic classification tasks despite having 177 times fewer parameters, demonstrating its effectiveness.
arXiv.org Artificial Intelligence
Dec-5-2023
- Country:
- Oceania > New Zealand (0.04)
- South America
- Pacific Ocean > North Pacific Ocean
- San Francisco Bay > Golden Gate (0.04)
- North America
- Panama (0.14)
- Costa Rica (0.04)
- Puerto Rico (0.04)
- Greenland (0.04)
- Dominican Republic (0.04)
- Honduras (0.04)
- Guatemala (0.04)
- Mexico (0.04)
- Canada (0.04)
- Haiti (0.04)
- United States
- New York (0.04)
- Virginia (0.04)
- South Carolina (0.04)
- Alaska (0.04)
- Maine (0.04)
- Minnesota (0.04)
- Indiana (0.04)
- Texas (0.04)
- Missouri (0.04)
- Wisconsin (0.04)
- Oregon (0.04)
- Maryland (0.04)
- Kansas (0.04)
- Hawaii (0.04)
- North Carolina (0.04)
- Michigan (0.04)
- New Hampshire (0.04)
- Rhode Island (0.04)
- New Mexico (0.04)
- New Jersey (0.04)
- Louisiana (0.04)
- Nebraska (0.04)
- Pennsylvania (0.04)
- Kentucky (0.04)
- Montana (0.04)
- Ohio > Lucas County
- Oregon (0.04)
- Nevada > Clark County
- Las Vegas (0.04)
- Europe
- Germany (0.04)
- Greece (0.04)
- Poland (0.04)
- Spain (0.04)
- Iceland (0.04)
- Switzerland (0.04)
- Denmark (0.04)
- Norway (0.04)
- Russia (0.04)
- Belgium (0.04)
- Sweden (0.04)
- Portugal (0.04)
- Romania > Sud - Muntenia Development Region
- Giurgiu County > Giurgiu (0.04)
- Italy > Calabria
- Catanzaro Province > Catanzaro (0.04)
- France > Île-de-France
- United Kingdom
- Scotland (0.04)
- England > Greater London
- London (0.04)
- Asia
- North Korea (0.04)
- India (0.04)
- Afghanistan (0.04)
- Singapore (0.04)
- Thailand (0.04)
- Russia (0.04)
- Philippines (0.04)
- Vietnam (0.04)
- Mongolia (0.04)
- Pakistan (0.04)
- South Korea > Seoul
- Seoul (0.04)
- China
- Hong Kong (0.04)
- Beijing > Beijing (0.04)
- Anhui Province (0.04)
- Middle East
- Republic of Türkiye (0.04)
- Jordan (0.04)
- Israel (0.04)
- Iran (0.04)
- Japan > Honshū
- Kantō > Tokyo Metropolis Prefecture > Tokyo (0.04)
- Africa
- Middle East > Egypt (0.04)
- Nigeria (0.04)
- Genre:
- Research Report (0.82)
- Industry:
- Technology: