Using Rejecto together with RuleORama

[Pltpx]

I would like to use the RuleORama plugin along with Rejecto. How can I generate a language model that combines the two?

[Halle Winkler]

Welcome,

Sorry, they are separate approaches – Rejecto creates statistical language models and what RuleORama creates is effectively a rules-based grammar.

[Pltpx]

Thank you. Being able to combine the different languageModels would help leverage the unique features of the rules-based grammar plus the reject non-dictionary words. This would make for an extendible and a chainable languageModel.

Would such a use-case be on the roadmap? Probably, it would require more computational resources for an iPhone, eh?

[Halle Winkler]

Yep, it would create search spaces of truly remarkable size :) .

However, if your RuleORama grammar has a bit of complexity it ought to already be ignoring intermittent noises and utterances.

I.e., if the grammar has very short possible hypotheses it may sometimes respond to random stuff, but if the minimum allowable utterance has at least a few syllables it ought to be ignoring shorter intermittent sounds and utterances pretty decently. If you have a very over-reactive grammar you might also want to investigate whether you have the right vadThreshold value – it could be too low for your use case.

[Author]

Posts