I just signed up for the SDK and am testing it out with my product. It works pretty well so far, but I have question regarding the regular expression functionality for processTextField.
One of the fields I need to recognize is a name field. These names are actually the names of my users and therefore I know the entire domain of possible entries. I'd like to pass this domain to the SDK to give it a better chance of a positive match. I'm trying to use the regExp parameter of processTextField to accomplish this, but it seems to have no effect on the outcome. I have tried passing
none of these seem to have an effect on the outcome. Guidance or suggestions, appreciated!
Specifying regular expression doesn't actually force recognition engine to always use it in results. There is a possibility to get output completely different to what is specified in regexp.
When recognizer has several hypotheses how to recognize given word, it checks all of them against regular expression. If given recognition variant conforms to regexp, it has higher probability of being selected as final recognition output. But if there is no variant that matches regular expression, the result cannot be conforming to it.
However, there are other options besides regular expression to improve recognition quality. For example, internally there are many specialized dictionaries for names in different languages.
You can send your image snippet to email@example.com. We'll take a look at it and probably suggest some options or provide something in cloud API to get better results.
answered 30 Mar '12, 11:17
Vasily Panferov ♦♦