We have a requirement of recognizing dates, ex: date of birth, date of joining, date of graduation in such cases how to handle the recognizer
Dates(1-31) and months(Jan-Dec) can be handled but how to handle year which can be in a very big range and more combination of numbers (1880-2015)
to maintain a good accuracy as individual digits which we are getting accuracy of (85%)
Regards,
CK
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
If you are using language model you can introduce all years and numbers in a language model training corpus. You can create required sentences with a script.
Once text is recognized you have to create an analysis code to parse the number into digit string and type. For example of such code you can check
As you said if i am using langugae model I will add all sentences, if i am using a grammar what kind of structure i can use to get best accuarcy?
Also I have seen some import functionalites
Hi
We have a requirement of recognizing dates, ex: date of birth, date of joining, date of graduation in such cases how to handle the recognizer
Dates(1-31) and months(Jan-Dec) can be handled but how to handle year which can be in a very big range and more combination of numbers (1880-2015)
to maintain a good accuracy as individual digits which we are getting accuracy of (85%)
Regards,
CK
good question! I have the same problem.
If you are using language model you can introduce all years and numbers in a language model training corpus. You can create required sentences with a script.
Once text is recognized you have to create an analysis code to parse the number into digit string and type. For example of such code you can check
https://github.com/wit-ai/duckling
a source code of Wit.AI startup which specifically targeted parsing problem.
Hi Nickolay,
As you said if i am using langugae model I will add all sentences, if i am using a grammar what kind of structure i can use to get best accuarcy?
Also I have seen some import functionalites
"For example,
import <com.sun.speech.app.index.1stto31st>;
import <com.sun.speech.app.numbers.*>;" </com.sun.speech.app.numbers.*></com.sun.speech.app.index.1stto31st>
in the site "http://www.w3.org/TR/jsgf/#16460" will that help in this situation and in Sphinx for dates recognition
Last edit: CK 2015-05-19