There is an ongoing debate about whether the Internet is like a public sphere or an echo chamber. Among many forms of social media, Twitter is one of the most crucial online places for political debate. Most of the previous studies focus on the formal structure of the Twitter political field, such as its homophilic tendency, or otherwise limit the analysis to a few topics. In order to explore whether Twitter functions as an echo chamber in general, however, we have to investigate not only the structure but also the contents of Twitter's political field. Accordingly, we conducted both large-scale social network analysis and natural language processing. We firstly applied a community detection method to the reciprocal following network in Twitter and found five politically distinct communities in the field. We further examined dominant topics discussed therein by employing a topic model in analyzing the content of the tweets, and we found that a topic related to xenophobia is circulated solely in right-wing communities. To our knowledge, this is the first study to address echo chambers in Japanese Twitter political field and to examine the formal structure and the contents of tweets with the combination of large-scale social network analysis and natural language processing.