class ConversationSegmenter:
def __init__(self, messages, windowSize, cosineSimilarityThreshold,
tokenizer):
self.messages = messages
self.window = Window(windowSize)
self.similarTopicCalculator = SimilarTopicCalculator(
self.window, messages, tokenizer)
self.replyObjectPredictor = ReplyObjectPredictor(
self.window, cosineSimilarityThreshold,
self.similarTopicCalculator, tokenizer)
def segment(self):
topics = [None for i in self.messages]
topicSet = []
for i, message in enumerate(self.messages):
print("Processing message id: " + str(message.getID()))
if i > 0 and self.messages[i -
1].getAuthor() == message.getAuthor():
topics[i] = topics[i - 1]
topics[i].appendMessage(message, 'same author')
else:
(replied_topic,
reason) = self.replyObjectPredictor.predict(message)
if replied_topic is None:
topic = Topic(message, reason)
topicSet.append(topic)
else:
topic = replied_topic
topic.appendMessage(message, reason)
self.window.addTopic(topic)
topics[i] = topic
# eliminate mistopics
i = 0
while i < len(topicSet):
if topicSet[i].size() > 2 or i == 0:
pass
otherMessages = topicSet[i - 1].getMessages()
messages = topicSet[i].getMessages()
if messages[0].getID() > otherMessages[0].getID(
) and messages[-1].getID() < otherMessages[-1].getID():
topicSet[i - 1].absorve(topicSet[i])
topicSet.remove(topicSet[i])
else:
i = i + 1
return topicSet
