vec = Vectoriser(docs, base=self.base_length)
rl_agent = RLAgent(vec,
summaries,
strict_para=self.rl_strict,
train_round=self.train_episode)
summary = rl_agent(rewards)
return summary
if __name__ == '__main__':
# read source documents
reader = CorpusReader('data/topic_1')
source_docs = reader()
# generate summaries, with summary max length 100 tokens
supert = Supert(source_docs)
rl_summarizer = RLSummarizer(reward_func=supert)
summary = rl_summarizer.summarize(source_docs, summ_max_len=100)
print('\n=====Generated Summary=====')
print(summary)
# (Optional) Evaluate the quality of the summary using ROUGE metrics
if os.path.isdir('./rouge/ROUGE-RELEASE-1.5.5'):
refs = reader.readReferences(
) # make sure you have put the references in data/topic_1/references
avg_rouge_score = {}
for ref in refs:
rouge_scores = evaluate_summary_rouge(summary, ref)
add_result(avg_rouge_score, rouge_scores)
print('\n=====ROUGE scores against {} references====='.format(
len(refs)))