T5 Base Korean Summarization

🚀 t5-base-korean-summarization

This is a T5 model designed for Korean text summarization. It addresses the need for efficient summarization of Korean texts, offering a practical solution for users dealing with large volumes of Korean language data.

✨ Features

• This model is finetuned based on the 'paust/pko-t5-base' model.
• It is finetuned using 3 datasets, which are as follows:
  • Korean Paper Summarization Dataset(논문자료 요약)
  • Korean Book Summarization Dataset(도서자료 요약)
  • Korean Summary statement and Report Generation Dataset(요약문 및 레포트 생성 데이터)

🚀 Quick Start

Prerequisites

Make sure you have nltk and transformers installed. You can install them using the following commands:

pip install nltk
pip install transformers

Example Code

import nltk
nltk.download('punkt')
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

model = AutoModelForSeq2SeqLM.from_pretrained('eenzeenee/t5-base-korean-summarization')
tokenizer = AutoTokenizer.from_pretrained('eenzeenee/t5-base-korean-summarization')

prefix = "summarize: "
sample = """
    안녕하세요? 우리 (2학년)/(이 학년) 친구들 우리 친구들 학교에 가서 진짜 (2학년)/(이 학년) 이 되고 싶었는데 학교에 못 가고 있어서 답답하죠? 
    그래도 우리 친구들의 안전과 건강이 최우선이니까요 오늘부터 선생님이랑 매일 매일 국어 여행을 떠나보도록 해요. 
    어/ 시간이 벌써 이렇게 됐나요? 늦었어요. 늦었어요. 빨리 국어 여행을 떠나야 돼요. 
    그런데 어/ 국어여행을 떠나기 전에 우리가 준비물을 챙겨야 되겠죠? 국어 여행을 떠날 준비물, 교안을 어떻게 받을 수 있는지 선생님이 설명을 해줄게요. 
    (EBS)/(이비에스) 초등을 검색해서 들어가면요 첫화면이 이렇게 나와요. 
    자/ 그러면요 여기 (X)/(엑스) 눌러주(고요)/(구요). 저기 (동그라미)/(똥그라미) (EBS)/(이비에스) (2주)/(이 주) 라이브특강이라고 되어있죠? 
    거기를 바로 가기를 누릅니다. 자/ (누르면요)/(눌르면요). 어떻게 되냐? b/ 밑으로 내려요 내려요 내려요 쭉 내려요. 
    우리 몇 학년이죠? 아/ (2학년)/(이 학년) 이죠 (2학년)/(이 학년)의 무슨 과목? 국어. 
    이번주는 (1주)/(일 주) 차니까요 여기 교안. 다음주는 여기서 다운을 받으면 돼요. 
    이 교안을 클릭을 하면, 짜잔/. 이렇게 교재가 나옵니다 .이 교안을 (다운)/(따운)받아서 우리 국어여행을 떠날 수가 있어요. 
    그럼 우리 진짜로 국어 여행을 한번 떠나보도록 해요? 국어여행 출발. 자/ (1단원)/(일 단원) 제목이 뭔가요? 한번 찾아봐요. 
    시를 즐겨요 에요. 그냥 시를 읽어요 가 아니에요. 시를 즐겨야 돼요 즐겨야 돼. 어떻게 즐길까? 일단은 내내 시를 즐기는 방법에 대해서 공부를 할 건데요. 
    그럼 오늘은요 어떻게 즐길까요? 오늘 공부할 내용은요 시를 여러 가지 방법으로 읽기를 공부할겁니다. 
    어떻게 여러가지 방법으로 읽을까 우리 공부해 보도록 해요. 오늘의 시 나와라 짜잔/! 시가 나왔습니다 시의 제목이 뭔가요? 다툰 날이에요 다툰 날. 
    누구랑 다퉜나 동생이랑 다퉜나 언니랑 친구랑? 누구랑 다퉜는지 선생님이 시를 읽어 줄 테니까 한번 생각을 해보도록 해요."""

inputs = [prefix + sample]

inputs = tokenizer(inputs, max_length=512, truncation=True, return_tensors="pt")
output = model.generate(**inputs, num_beams=3, do_sample=True, min_length=10, max_length=64)
decoded_output = tokenizer.batch_decode(output, skip_special_tokens=True)[0]
result = nltk.sent_tokenize(decoded_output.strip())[0]

print('RESULT >>', result)

📚 Documentation

Evalutation Result

• Korean Paper Summarization Dataset(논문자료 요약)

  ROUGE-2-R 0.09868624890432466
  ROUGE-2-P 0.9666714545849712
  ROUGE-2-F 0.17250881441169427
  

• Korean Book Summarization Dataset(도서자료 요약)

  ROUGE-2-R 0.1575686156943213
  ROUGE-2-P 0.9718318136896944
  ROUGE-2-F 0.26548116834852586
  

• Korean Summary statement and Report Generation Dataset(요약문 및 레포트 생성 데이터)

  ROUGE-2-R 0.0987891733555808
  ROUGE-2-P 0.9276946867981899
  ROUGE-2-F 0.17726493110448185
  

Training

The model was trained with the following parameters:

Seq2SeqTrainingArguments(
    per_device_train_batch_size=8,
    per_device_eval_batch_size=8,
    auto_find_batch_size=False,
    weight_decay=0.01,
    learning_rate=4e-05,
    lr_scheduler_type=linear,
    num_train_epochs=3,
    fp16=True)

Model Architecture

T5ForConditionalGeneration(                                                                                                                                             
  (shared): Embedding(50358, 768)                                                                                                                                       
  (encoder): T5Stack(                                                                                                                                                   
    (embed_tokens): Embedding(50358, 768)                                                                                                                               
    (block): ModuleList(                                                                                                                                                
      (0): T5Block(                                                                                                                                                     
        (layer): ModuleList(                                                                                                                                            
          (0): T5LayerSelfAttention(                                                                                                                                    
            (SelfAttention): T5Attention(                                                                                                                               
              (q): Linear(in_features=768, out_features=768, bias=False)
              (k): Linear(in_features=768, out_features=768, bias=False)
              (v): Linear(in_features=768, out_features=768, bias=False)
              (o): Linear(in_features=768, out_features=768, bias=False)
              (relative_attention_bias): Embedding(32, 12)
            )        
            (layer_norm): T5LayerNorm()  
            (dropout): Dropout(p=0.1, inplace=False)
          )        
          (1): T5LayerFF(   
            (DenseReluDense): T5DenseGatedActDense(
              (wi_0): Linear(in_features=768, out_features=2048, bias=False)
              (wi_1): Linear(in_features=768, out_features=2048, bias=False)
              (wo): Linear(in_features=2048, out_features=768, bias=False)
              (dropout): Dropout(p=0.1, inplace=False)
              (act): NewGELUActivation()
            )
            (layer_norm): T5LayerNorm()
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
      )
      (1~11): T5Block(
        (layer): ModuleList(
          (0): T5LayerSelfAttention(
            (SelfAttention): T5Attention(
              (q): Linear(in_features=768, out_features=768, bias=False)
              (k): Linear(in_features=768, out_features=768, bias=False)
              (v): Linear(in_features=768, out_features=768, bias=False)
              (o): Linear(in_features=768, out_features=768, bias=False)
            )
            (layer_norm): T5LayerNorm()
            (dropout): Dropout(p=0.1, inplace=False)
          )
          (1): T5LayerFF(
            (DenseReluDense): T5DenseGatedActDense(
              (wi_0): Linear(in_features=768, out_features=2048, bias=False)
              (wi_1): Linear(in_features=768, out_features=2048, bias=False)
              (wo): Linear(in_features=2048, out_features=768, bias=False)
              (dropout): Dropout(p=0.1, inplace=False)
              (act): NewGELUActivation()
            )
            (layer_norm): T5LayerNorm()
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
      )
    )
    (final_layer_norm): T5LayerNorm()
    (dropout): Dropout(p=0.1, inplace=False)
  )
  (decoder): T5Stack(
    (embed_tokens): Embedding(50358, 768)
    (block): ModuleList(
      (0): T5Block(
        (layer): ModuleList(
          (0): T5LayerSelfAttention(
            (SelfAttention): T5Attention(
              (q): Linear(in_features=768, out_features=768, bias=False)
              (k): Linear(in_features=768, out_features=768, bias=False)
              (v): Linear(in_features=768, out_features=768, bias=False)
              (o): Linear(in_features=768, out_features=768, bias=False)
              (relative_attention_bias): Embedding(32, 12)
            )
            (layer_norm): T5LayerNorm()
            (dropout): Dropout(p=0.1, inplace=False)
          )
          (1): T5LayerCrossAttention(
            (EncDecAttention): T5Attention(
              (q): Linear(in_features=768, out_features=768, bias=False)
              (k): Linear(in_features=768, out_features=768, bias=False)
              (v): Linear(in_features=768, out_features=768, bias=False)
              (o): Linear(in_features=768, out_features=768, bias=False)
            )
            (layer_norm): T5LayerNorm()
            (dropout): Dropout(p=0.1, inplace=False)
          )
          (2): T5LayerFF(
            (DenseReluDense): T5DenseGatedActDense(
              (wi_0): Linear(in_features=768, out_features=2048, bias=False)
              (wi_1): Linear(in_features=768, out_features=2048, bias=False)
              (wo): Linear(in_features=2048, out_features=768, bias=False)
              (dropout): Dropout(p=0.1, inplace=False)
              (act): NewGELUActivation()
            )
            (layer_norm): T5LayerNorm()
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
      )
      (1~11): T5Block(
        (layer): ModuleList(
          (0): T5LayerSelfAttention(
            (SelfAttention): T5Attention(
              (q): Linear(in_features=768, out_features=768, bias=False)
              (k): Linear(in_features=768, out_features=768, bias=False)
              (v): Linear(in_features=768, out_features=768, bias=False)
              (o): Linear(in_features=768, out_features=768, bias=False)
            )
            (layer_norm): T5LayerNorm()
            (dropout): Dropout(p=0.1, inplace=False)
          )
          (1): T5LayerCrossAttention(
            (EncDecAttention): T5Attention(
              (q): Linear(in_features=768, out_features=768, bias=False)
              (k): Linear(in_features=768, out_features=768, bias=False)
              (v): Linear(in_features=768, out_features=768, bias=False)
              (o): Linear(in_features=768, out_features=768, bias=False)
            )
            (layer_norm): T5LayerNorm()
            (dropout): Dropout(p=0.1, inplace=False)
          )
          (2): T5LayerFF(
            (DenseReluDense): T5DenseGatedActDense(
              (wi_0): Linear(in_features=768, out_features=2048, bias=False)
              (wi_1): Linear(in_features=768, out_features=2048, bias=False)
              (wo): Linear(in_features=2048, out_features=768, bias=False)
              (dropout): Dropout(p=0.1, inplace=False)
              (act): NewGELUActivation()
            )
            (layer_norm): T5LayerNorm()
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
      )
      (final_layer_norm): T5LayerNorm()                              
    (dropout): Dropout(p=0.1, inplace=False)                                                                                           
  )                                                                
  (lm_head): Linear(in_features=768, out_features=50358, bias=False)                                                                   
)

📄 License

No license information provided in the original document.

📚 Citation

• Raffel, Colin, et al. "Exploring the limits of transfer learning with a unified text-to-text transformer." J. Mach. Learn. Res. 21.140 (2020): 1-67.