In this paper we investigate if the bigram graph-based analysis can be applied to identify the most common subjects of discussion in English - spoken social media. We will present the construction of directed graph of bigrams, data filtration model and we will perform validation on three real (not artificially created)datasets, each containing 1 000 000 microblogging posts from Twitter platform. We will compare our findings with our earlier researches that used hashtag - based analysis and simple NLP method namely lemmatization. Those methods were evaluated on the same dataset as we use in this paper. With bigrams network it is possible to obtain some limited information about subjects the users are talking about however less than when the analysis uses hashtags. Bigrams also work better than simple lemmatization that in our previous experiment did not produce any useful data. The approach that supplies us with most useful data describing content of microblogging posts seems to be hashtag filtration. However this last type of analysis requires the presence of additional annotation data in user messages. Terms generated by bigrams and hashtag analysis might not be identical and sometimes one of them might be valuable complement of the other. Bigrams network might show the general idea of the content of the information gathered in microblogging posts. It is recommended to visualize a filtered part of this graph (for example limited by weights of edges)to see the general idea of data consisted in this network.