You can deploy this model and start earning money today!

ganbon/
novel-bert-base-relationship

Name: novel-bert-base-relationship
Availability: InStock

tohoku-nlp/bert-base-japanese-v3

finetune

apache-2.0

Views

No views yet

Novel-BERT-base-Relationship(beta version)

This model is in beta and subject to change.
This model determines the utterance response relationship in novel dialogue.
This model was constructed for a Japanese Novel Dialogue System.

Dataset

We trained the model using 5,000 annotated samples from 8 romance genre titles and 2 mystery genre titles on the Shosetukani Naro platform.

How to use

Install dependencies
$ pip install transformers fugashi[unidic-lite]
Inference
If the output is True, it is determined that there is a utterance response relationship, and if the output is False, it is determined that there is no relationship.

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification

model_path = "ganbon/novel-bert-base-relationship"
model = AutoModelForSequenceClassification.from_pretrained(model_path, trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_path)
previous_sentence_list = [["「おはよう」",
                       "「おはよう。元気ですか？」",
                       "太郎が花子を気にかける。"
                       "昨日体調が悪そうだったからだ。"
                       ],
                       [
                           "「早く学校に行こうか」"
                           "「そうだね！」"
                           "そういいながら、学校に行った。",
                           "学校ではすでに次郎がいた。"
                       ]]
next_sentence_list = ["「はい、元気です。お気遣いありがとう。」","「おはよう、次郎」"]
for previous, next_sentence in zip(previous_sentence_list, next_sentence_list):
   previous_sentences = "[SEP]".join(previous)
   outputs = model(**tokenizer(previous_sentences, next_sentence, return_tensors='pt'))
   if torch.argmax(outputs["logits"]).item() == 0:
       print(True)
   else:
       print(False)

Citation

When using this model, please cite one of the following.

@article{iwamoto-2024-novel_relation-ja,
  title={小説を用いた対話コーパス構築のための言語モデルによる台詞間の発話応答関係の判定手法},
  author={岩本 和真 and 安藤 一秋},
  journal={人工知能学会全国大会論文集},
  volume={JSAI2024},
  pages={2G4GS605-2G4GS605},
  year={2024},
}
@inproceedings{iwamoto-2024-novel_relation,
	author	=	"Iwamoto, Kazuma and Ando, Kazuaki",
	title =	"A Method for Determining Utterance-Response Relationships Between Japanese Novel Lines for Constructing a Daily Dialogue Corpus",
	booktitle = 	"2024 16th IIAI International Congress on Advanced Applied Informatics (IIAI-AAI)",
	year =		2024,
}