Intelligent word recognition, or IWR, is the recognition of unconstrained handwritten words. IWR recognizes entire words or phrases instead of individual characters Optical Mark Recognition, or OCR is using. For instance, when trying to extract the word 'paper' using OCR system, it will recognize all characters one by one ('p', 'a', 'p', 'e', 'e', and 'r'). On the other hand, the IWR will match letters to a user-defined dictionary to extract the whole word 'paper'. IWR significantly reduces character errors.

How to use it it?

To use IWR in Papersurvey.io go to Surveys -> Verify section and click the Dictionary tab. This will be helpful when you are expecting to have multiple repeating text answers that may similarly match (e.g. Course Name, Event Name, Location, Country, etc.).

To add a new verification rule click the 'Add Custom Rule' button. In the modal, select the question for which the rule should be applied (Only works with 'Short Text' and 'Number' questions).

Then select one of two methods - 'Text is similar to' or 'Text is equal to'. We recommend using 'Text is similar to' method with 'Number' questions and 'Text is equal to' with 'Short Text' questions to get the highest accuracy rates.

Enter text value which should be always accepted as a correct answer and would not need additional verification. You may add multiple values by separating each value with a line break.

After clicking 'Add', similar responses will be automatically verified.

What about handwriting recognition?

We use handwriting recognition (HTR) to read the text. It is build on modern technologies and utilizes deep neural network approach in recognizing the handwriting and combines all ideas behind handwriting recognition (IWR, OCR, ICR) into one. Therefore, IWR is not necessarily needed but helpful to reduce the number of false positives (e.g. Software detects that user wrote "New ork" but using IWR, we can automatically correct it to "New York" if we know all possible combinations).