Search Tips and Best Practices

This guide shows you how to get the most out of DeepPaper AI's search feature.

Picking the Right Search Type

DeepPaper AI has different search types. Choose the one that fits your needs:

General Search

• When to use: Looking for all papers on a topic
• Example: machine learning finds all papers with "machine learning"

Survey Search

• When to use: Wanting overview papers in a field
• Example: Search computer vision with Survey type to find review papers
• Tip: Great for starting research or learning a field

Benchmark Search

• When to use: Looking for papers with standard tests
• Example: Search natural language processing with Benchmark type for papers with standard tests
• Tip: Good for comparing methods or finding test standards

Tutorial Search

• When to use: Learning specific methods
• Example: Search deep reinforcement learning with Tutorial type for how-to papers
• Tip: Perfect for learning new techniques

Perspective Search

• When to use: Wanting opinion pieces or future views
• Example: Search AI ethics with Perspective type for opinion papers
• Tip: Good for understanding debates or future trends

Dataset Search

• When to use: Looking for papers about datasets
• Example: Search image classification with Dataset type for papers about datasets
• Tip: Great for finding data for experiments

Choosing Good Keywords

Keyword Tips

1. Be Specific

   • Use convolutional neural network segmentation not just deep learning
   • Specific keywords give better results

2. Use Field Terms

   • Use BERT not language model
   • Field terms find more relevant papers

3. Use Current Terms

   • Use popular technical names
   • Field terms change over time

Combining Keywords

While DeepPaper AI doesn't support complex searches, you can still improve results:

1. Use Clear Phrases

   • Use self attention neural networks not self attention networks neural
   • Clear phrases match how papers are written

2. Try Different Words

   • If graph neural networks doesn't work, try GNN or graph convolutional networks
   • Different papers use different terms

Working with Results

Finding Good Papers

After searching, you can find the best papers by:

1. Looking at Dates

   • Focus on recent or classic papers
   • New papers have latest methods, older ones are well-tested

2. Checking Authors

   • Papers from top researchers or labs are usually good
   • Multiple papers from same author means expertise

3. Using Highlights

   • Look at keyword context in abstracts
   • Papers with keywords in important parts are usually relevant

Using arXiv IDs

arXiv IDs tell you about papers:

1. Field Info

   • ID prefixes show the field
   • Like: cs.AI (AI), cs.CV (Computer Vision), cs.CL (Language)

2. Date Info

   • Numbers show when published
   • Like: 2301.xxxxx means January 2023

Managing Searches

Planning Searches

For complex topics, try this approach:

1. Start Broad

   • Use general keywords first
   • Then get more specific based on results

2. Mix Search Types

   • Start with Survey type for overview
   • Then use other types for details

3. Save Good Keywords

   • Keep track of keywords that work
   • Make different sets for different stages

Keeping Up to Date

Research keeps moving. Stay current by:

1. Regular Searches

   • Repeat important searches weekly or monthly
   • Focus on new papers in fast-moving fields

2. Saving Searches

   • Save important search URLs
   • Easy to check later or share with team

Common Problems

Not Enough Results

If you get too few results, try:

1. More General Terms

   • Change graph convolutional attention networks to graph neural networks
   • Use fewer, simpler terms

2. Different Search Types

   • If Survey type doesn't work, try general search
   • Different types work for different needs

3. Check Spelling

   • Make sure terms are spelled right
   • Try more common field terms

Too Many Results

If you get too many or wrong results, try:

1. More Specific Terms

   • Change deep learning to deep learning medical image segmentation
   • Add field or method details

2. More Technical Terms

   • Change AI prediction to time series forecasting LSTM
   • Technical terms give better results

3. Specific Search Types

   • Use Tutorial or Survey types
   • These types narrow results