Standard language processing systems make use of a compositionality principle in language which argues that a sentence meaning is derived from its constituents and the way they are grouped. Non sentential utterances pose a challenge to this principle since stand alone constituents (PP or NP) carry full sentential content that can not be derived solely based on the fragment lexical surface.
In this talk we will present approaches that aim at reconstructing meaning from fragmental utterances along with their limitations. We will also present current data from the fruitcarts experiment exhibiting the phenomena we want to address. Preliminary work towards a module that will process fragments and a future plan of research will also be discussed.