Build and Visualize Three Kingdoms Character Networks with Python

Introduction

Inspired by a long‑standing interest in natural language processing and social network analysis, the author explores how to combine these fields to uncover the character network hidden within the classic Chinese novel Romance of the Three Kingdoms using Python.

Preparation

First, the full text of the novel is obtained and split into chapters. An entity‑mention dictionary and an entity‑type dictionary are built so that different names referring to the same historical figure (e.g., "刘备", "玄德") can be linked.

chapters = get_sanguo()  # list of chapter texts
print(chapters[0][:106])

entity_mention_dict, entity_type_dict = get_sanguo_entity_dict()
print("刘备的指称有：", entity_mention_dict["刘备"])
print("刘备的类型为", entity_type_dict["刘备"])

Building the Social Network

HarvestText is used to load the entity dictionaries, segment sentences, and build an entity graph where an edge is added whenever two entities co‑occur within two consecutive sentences.

ht = HarvestText()
ht.add_entities(entity_mention_dict, entity_type_dict)  # load model

doc = chapters[0].replace("操", "曹操")
ch1_sentences = ht.cut_sentences(doc)
# create two‑sentence windows
doc_ch01 = [ch1_sentences[i] + ch1_sentences[i+1] for i in range(len(ch1_sentences)-1)]
ht.set_linking_strategy("freq")
G = ht.build_entity_graph(doc_ch01, used_types=["人名"])  # only person names

The resulting graph contains thousands of nodes and edges. A subgraph of the most connected characters (degree ≥ 5) is visualized.

Analysis

Important characters are identified by degree, PageRank, and betweenness centrality. The top‑20 figures are plotted as bar charts.

page_ranks = pd.Series(nx.algorithms.pagerank(G)).sort_values()
page_ranks.tail(20).plot(kind="barh")
plt.show()

between = pd.Series(nx.betweenness_centrality(G)).sort_values()
between.tail(20).plot(kind="barh")
plt.show()

Community Detection

The Louvain algorithm (via the community package) partitions the network into clusters that correspond roughly to the three kingdoms and their major retainers.

import community  # python‑louvain
partition = community.best_partition(G)
comm_dict = defaultdict(list)
for person in partition:
    comm_dict[partition[person]].append(person)

Sample communities are printed and visualized.

Dynamic Network

To illustrate how the social network evolves over the narrative, snapshots are taken every five chapters and combined into an animated GIF using moviepy. Only the most active nodes are kept in each frame.

import moviepy.editor as mpy
from moviepy.video.io.bindings import mplfig_to_npimage
# ... build sliding windows of graphs ...
animation = mpy.VideoClip(make_frame_mpl, duration=duration)
animation.write_gif("./images/三国社交网络变化.gif", fps=1)

Conclusion

The tutorial shows that, despite the novel’s ancient language, modern Python tools can automatically construct and analyze a rich social graph, revealing key figures, their influence, and the shifting alliances throughout the story.