Comparison of measurements by the betweenness centrality and subjective experiment on the word priority of tweets

In recent years, it is very popular to use Twitter as an interaction tool and for information transmission and sharing as well. It is more difficult for us to understand the interaction flow on Twitter than the usual dialogue and conversation. But it will become easier to catch the interaction flow if we can focus on the priority words of Tweets. As we know, the word priority of Tweets can be measured by the betweenness centrality based on the word co-occurrence network. However, it is unknown whether the word priority measured by the betweenness centrality is the same to what the Twitter users think or not, which has not yet been well studied. In this study, we design an experiment to compare the measurement calculated by the betweenness centrality and the evaluation given by subjects. We further analyze their difference by statistical methods. Our result shows that the measurement determined by the betweenness centrality does not always match the subjective evaluation, but they tend to have a positive correlation. Particularly, proper nouns and words with long characters are more likely to be regarded higher priority than the betweenness centrality.