Log in / Register
Home arrow Computer Science arrow Social Informatics
< Prev   CONTENTS   Next >

3.3 Semantic Analysis

Understanding the content of the posts, and in particular the key topics of interest for the users, is important to understand engagement. For this purpose we have semantically annotated the tweets of our dataset by using TextRazor.[1] This annotator provides us with the entities from all seed and non-seed posts in our dataset, thereby returning a mapping between each post and a list of DBPedia URIs. We can then identify the concepts that are referred to within a post by looking up each entity's rdf:type in the DBPedia ontology and recording these concepts in a list for each post.

Table 3. Top entities/concepts for seeds vs. non-seed posts

Top Entity [Types] Seed Posts

Top Entity [Types] NonSeed Posts

Dorset [Place, PopulatedPlace] Bournemouth [Place, PopulatedPlace]

England [Place, Country, PopulatedPlace] Flood


Weymouth,_Dorset [Place, PopulatedPlace] Poole [Place, PopulatedPlace]


A31_road [Place, Road] Collision

South_West_England [Place, PopulatedPlace] Dorchester,_Dorset [Place, PopulatedPlace Volvo_XC90 [Automobile] Severe_weather [WeatherHazards, Danger] A35_road [Place, Road]

Bournemouth [Place, PopulatedPlace] Weymouth,_Dorset [Place, PopulatedPlace] Dorset [Place, PopulatedPlace]

Burglary [Crime]

Dorset_Police [LawEnforcementAgency] Closed-circuit_television

Poole [Place, PopulatedPlace]

Twitter [Organisation, Company]

Bridport [Place, PopulatedPlace] Driving_under_the_influence

Assault [Crime] 999_(emergency_telephone_number) Traffic

Robbery [Crime] Property_damage [Crime]

Table 3 presents the top entities/concepts for the seed and non-seed posts respectively (top entities are the most frequent ones within our dataset). Note that only the URL label has been selected for better visualisation. However, each of those entity labels corresponds to a specific Wikipedia page, e.g. ( social_behaviour). Also note that not all the entities identified by TextRazor have associated an rdf:type concept in the DBPedia ontology.

Locations, such as Dorset, Bournemouth, Poole or Weymouth are constant across the two groups of posts. However, seed posts include less focalised locations, such as England and South West England. Additionally, seed posts include entities related with weather (snow, severe weather, flood) as well as road an infrastructures (A31 road, A35 road, etc.). Non seed posts, on the other hand, talk about crimes such as burglary, assault or driving under the influence of alcohol.

As we can see from this first overview, semantic entities help us to understand those topics of interests for the citizens, and to differentiate some of the key themes attracting their attention (e.g., road problems or weather conditions). A further analysis should be performed to investigate deeper which combinations of entities spike higher attention levels, in which context they appear (semantic relations with other tweets), and how they differ from the information explicitly provided by hashtags. These are part of our future line of work.

  • [1]
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science