In the last decades," distant reading" became one of the most promising methodologies in the domain of computer-assisted literary analysis. This paper looks at an alternative path for gathering new insights about the literary texts, that of the" shared reading", or" crowd reading", consisting in multiple readings of the same text, performed by various readers and analysed so as to uncover new theoretical insights about the literary phenomenon, based on the observation of similarities and divergences amongst readers in text interpretation and text understanding. More precisely, this paper presents an experience of crowd reading applied to the modeling of literary characters, in view of a semi-automated analysis of these literary objects.