Handwriting recognition works well for most responses, but certain questions have a limited set of expected answers, such as city names, course titles, or product names. Intelligent Word Recognition (IWR) takes advantage of this by matching written responses against your predefined dictionary, dramatically improving accuracy for those predictable fields.

How IWR improves accuracy

Unlike standard character-by-character recognition, IWR matches entire words against your dictionary. For example, standard recognition might read "New Yrok" due to a character-level error, but IWR matches it to "New York" from your dictionary and automatically verifies the correct answer.

When to use IWR

IWR works best for questions with:

• Limited possible answers such as countries, states, or cities
• Repeated responses like course names or department names
• Common misspellings of technical terms or brand names
• Standardized formats including ID numbers and codes

Setting up IWR

Step 1: Navigate to Dictionary settings

1. Go to Surveys in your dashboard
2. Select your survey
3. Click the Verify tab
4. Select Dictionary

Step 2: Add verification rules

1. Click Add Custom Rule
2. Select the question (works with Short Text and Number questions)
3. Choose your matching method:
   • Fuzzy match (similar spelling) for handwritten text (recommended)
   • Exact match for numbers or codes

Step 3: Enter expected values

Add your dictionary entries, one per line:

Example entries:

New York
Los Angeles
Chicago
Houston
Phoenix

Click Add to save your rule.

Best practices

For text responses

• Use "Fuzzy match" for flexibility
• Include common variations (NY, New York, NYC)
• Add common misspellings you have observed
• Keep lists manageable (under 100 entries)

For numeric responses

• Use "Exact match" for exact matching
• Include formats with and without separators
• Add leading zeros if applicable
• Consider ranges for validation

Dictionary management

• Start with a small dictionary and expand based on actual responses
• Review unmatched responses regularly
• Export verified responses to identify new dictionary entries
• Group similar rules by question type

How IWR works with handwriting recognition

PaperSurvey uses advanced Handwriting Text Recognition (HTR) powered by neural networks. IWR enhances this pipeline by:

1. Processing the handwritten text through HTR first
2. Comparing results against your dictionary
3. Auto-correcting close matches
4. Flagging uncertain matches for review

Common use cases

Location questions

United States
United Kingdom
Canada
Australia
Germany

Course evaluations

Introduction to Psychology
Calculus I
Organic Chemistry
World History
English Literature

Product feedback

Model A-100
Model A-200
Model B-100
Premium Version
Standard Version

Monitoring and improvement

Review verification results

• Check the Verify tab regularly
• Look for patterns in unmatched responses
• Add frequently occurring responses to your dictionary

Optimize your rules

• Remove rarely used entries
• Combine similar variations
• Adjust matching methods based on accuracy

IWR significantly reduces manual verification time while maintaining data accuracy. Start with your most common responses and refine the dictionary as you collect more data.