Referring to Objects Thr oughSub-Contexts in Multimodal Human-Computer Interaction

There is no one-to-one relation betweenreferentialtermsandtypesof accessto the referents(referring modes) in multimodalhuman-computer interaction. We proposea classificationof referringmodesimplying sub-contexts. Wedescribein detailthenatureof these contextual subsetsand we show their importancefor eachtypeof referringaction. Thenwe definea relationbetween termsandmodes,andwededucealist of disambiguationprinciplesfor the computationof referentialtermsin orderto identify the correctreferringmode,the correctsub-context, andthecorrectreferent.