• receptive field
• term cloud
• stigmergy
• time series similarity
• microblog analysis

To achieve the goal of distinguishing event-specific topics in social discussions a stigmergy-based architecture has been used. The stigmergy represents a type of spontaneous arrangement, derived from ants behavior inside colonies. To allow this architecture to work properly we collected a huge amount of textual data from a dataset containing tweets about a very important event. Those data were properly cleaned and filtered to obtain time series that could be managed from the stigmergic architecture. Each time series has been divided in temporal windows, each of them labeled as a specific archetype. The core of the proposed system is composed basically of two levels: the first one aims to assign to each temporal window the most similar archetype, while the second one receives for each time series a sequence of the associated archetypes and tries to group them according to the similarities between these sequences. This technique revealed very powerful in grouping sentence terms correctly in clusters.