Hello,

you could use the name finder to detect the time, location
and food in the query.

For the classification of the order type you could train a maxent model
which detects the order type.

You should do the training with a couple of hundred samples to start
with per language.

Jörn

On 11/20/11 8:34 AM, Eldad Yamin wrote:
  HI,
I'm wondering if anyone had an experience with using OpenNLP as a
search engine input parser.

For example, let's imagine that a restaurant wants to have a
search engine in their site. their users can write things like:
"Pizza to 12:00, delivery to my house at London" ->  food: pizza, time:
12:00, order type: delivery, location: London"
"Pasta, Apple pie, reservation, 14:00 (outside/smoking)." ->  food: pasta,
apple pie, time:14:00, order type: reservation, location: outside"
"Smoking area, reservation to 12:00" ->  "order type: reservation: , time:
12:00, location: outside/smoking area"

In addition, it should be multilingual (I.e. German, French etc).
I have the entities translated into different languages (I.e EN:reservation
->  FR:réservation).

Please advise.
Thanks!


Reply via email to