proposed solution:
use another similarity measure in post processing to get more accurate result
problem: single database record existing in multiple parts of the utternace
utterance: "i'm from Algeria, Algeria and my name is ..."
{ "Algeria": [ {"token": "from Algeria"}, {"token": "Algeria Algeria"}, {"token": "Algeria to"} ] }
scenario:
{ "cairo": [ {"token": "from cairo"}, {"token": "cairo"}, {"token": "cairo to"} ] }
problem is, this should only be cairo
proposed solution: use another similarity measure in post processing to get more accurate result problem: single database record existing in multiple parts of the utternace utterance: "i'm from Algeria, Algeria and my name is ..."
{ "Algeria": [ {"token": "from Algeria"}, {"token": "Algeria Algeria"}, {"token": "Algeria to"} ] }
we need to resolve two entities not just one!