def draw_weight(self, sentence):
attn_data = self.__get_attention(self.model, sentence)
gv = GraphVisualizer()
x = attn_data["text"]
y = x
x_data = []
y_data = []
z_data = []
for x_index in range(len(x)):
x_data.append("(" + str(x_index) + ")" + x[x_index])
z_data.append(attn_data["attn"][11][11][x_index][x_index])
data_meta_list = []
data_meta = {
"graph_type": "histogram",
"data_name": "Y",
"x_data": x_data,
"y_data": z_data,
"y_axis": "y2"
}
data_meta_list.append(data_meta)
graph_meta = {
"title": "BERT NER WEIGHT",
"x_tickangle": -45,
"y1_tickangle": 0,
"y2_tickangle": 0,
"x_name": "TOKEN",
"y1_name": "WEIGHT",
"y2_name": "WEIGHT",
}
return gv.draw_histogram(data_meta_list, graph_meta)
def get_plotly_graph(self, max_words=100):
gv = GraphVisualizer()
x = self.get_word_list()[:max_words]
y = [score for word, score in self.get_tf_list()][:max_words]
z = [score for word, score in self.get_tfidf_list()][:max_words]
data_meta_list = []
data_meta = {
"graph_type": "histogram",
"data_name": "TF",
"x_data": x,
"y_data": y,
"y_axis": "y1",
}
data_meta_list.append(data_meta)
data_meta = {
"graph_type": "scatter",
"data_name": "TF-IDF",
"x_data": x,
"y_data": z,
"y_axis": "y2"
}
data_meta_list.append(data_meta)
graph_meta = {
"title": "단어빈도 및 TF-IDF (TF & TF-IDF)",
"x_tickangle": -45,
"y1_tickangle": 0,
"y2_tickangle": 0,
"x_name": "단어 (WORD)",
"y1_name": "빈도 (TF)",
"y2_name": "TF-IDF",
}
return gv.draw_histogram(data_meta_list, graph_meta)
def get_inertia_transition_graph(self, inertia_list):
print(
".get_inertia_transition_graph() Will be replaced by .get_kmeans_inertia_transition_graph()"
)
gv = GraphVisualizer()
gv.set_plotly()
x = [i for i in range(1, len(inertia_list) + 1)]
y = inertia_list
data_meta_list = []
data_meta = {
"graph_type": "scatter",
"data_name": "Y",
"x_data": x,
"y_data": y,
"y_axis": "y1",
}
data_meta_list.append(data_meta)
graph_meta = {
"title": "K-Means Clutering Inertia Transition Graph",
"x_tickangle": 0,
"y1_tickangle": 0,
"y2_tickangle": 0,
"x_name": "NUMBER of CLUSTER",
"y1_name": "INERTIA",
"y2_name": "Y2",
}
return gv.draw_line_graph(data_meta_list, graph_meta)
def get_cluster_graph(self, df_result, label):
gv = GraphVisualizer()
gv.set_plotly()
data_meta_list = []
for i in OrderedDict.fromkeys(df_result[label]):
content_label_list = []
for content_label in df_result[df_result.predict == i]["content"]:
if len(content_label) > 30:
content_label = content_label[:30] + "..."
content_label_list.append(content_label)
else:
content_label_list.append(content_label)
data_meta = {
"data_name": i,
"x_data": df_result[df_result[label] == i]["x"],
"y_data": df_result[df_result[label] == i]["y"],
"label": content_label_list
}
data_meta_list.append(data_meta)
graph_meta = {
"title": "Cluter Graph - " + label,
"x_name": "TSNE X",
"y_name": "TSNE Y"
}
return gv.draw_scatter(data_meta_list, graph_meta)
def draw_weight(self, sentence):
gv = GraphVisualizer()
gv.set_plotly()
token_list, weight_list = self.get_weight(sentence)
x_data = [
"(" + str(token_index) + ")" + token
for token_index, token in enumerate(token_list)
]
z_data = [weight for weight in weight_list]
data_meta_list = []
data_meta = {
"graph_type": "histogram",
"data_name": "Y",
"x_data": x_data,
"y_data": z_data,
"y_axis": "y2"
}
data_meta_list.append(data_meta)
graph_meta = {
"title": "NER WEIGHT",
"x_tickangle": -45,
"y1_tickangle": 0,
"y2_tickangle": 0,
"x_name": "TOKEN",
"y1_name": "WEIGHT",
"y2_name": "WEIGHT",
}
return gv.draw_histogram(data_meta_list, graph_meta)
def get_kmeans_graph(self, df_result, label):
gv = GraphVisualizer()
gv.set_plotly()
data_meta_list = []
for predict in list(OrderedDict.fromkeys(df_result[label])):
data_meta = {
"data_name": predict,
"x_data": df_result[df_result[label]==predict]["x"],
"y_data": df_result[df_result[label]==predict]["y"],
"label": predict
}
data_meta_list.append(data_meta)
graph_meta = {
"title": "K-Means Clutering Graph - " + label,
"x_name": "TSNE X",
"y_name": "TSNE Y"
}
return gv.draw_scatter(data_meta_list, graph_meta)
def get_weight(self, sentence):
attn_data = self.__get_attention(self.model, sentence)
gv = GraphVisualizer()
token_list = attn_data["text"]
weight_list = []
for token_index in range(len(token_list)):
weight_list.append(
attn_data["attn"][11][11][token_index][token_index])
return {"token_list": token_list, "weight_list": weight_list}
def get_co_matrix_graph(self, max_words):
gv = GraphVisualizer()
gv.set_plotly()
co_word_list = self.result_list
x_data = []
max_freq = 0
for co_word, freq in co_word_list[:max_words]:
first_word = co_word[0]
second_word = co_word[1]
if max_freq < freq:
max_freq = freq
if first_word not in x_data:
x_data.append(first_word)
if second_word not in x_data:
x_data.append(second_word)
y_data = x_data
z_data = [[0 for freq in x_data] for freq in y_data]
for co_word, freq in co_word_list[:max_words]:
first_word = co_word[0]
second_word = co_word[1]
z_data[x_data.index(first_word)][y_data.index(second_word)] = freq
data_meta = {
"colorbar_title": "동시출현빈도 (CO-WORD FREQUENCY)",
"x_data": x_data,
"y_data": y_data,
"z_data": z_data
}
graph_meta = {
"title": "동시출현빈도 매트릭스 (CO-WORD MATRIX)",
"height": 1000,
"width": 1000,
"y_tickangle": -45,
"y_name": "Y",
"x_tickangle": -45,
"x_name": "X",
}
return gv.draw_matrix(data_meta, graph_meta)
def draw_sentence_tree(self, sentence, label_list, edge_list):
gv = GraphVisualizer()
gv.set_plotly()
return gv.draw_sentence_tree(sentence, label_list, edge_list)