# The open-source model of comet-t5-base-japanese - Free to use for generating common-sense inference results of Japanese events

Home

Comet T5 Base Japanese

Developed by nlp-waseda

T5 model fine-tuned on the Japanese ATOMIC dataset for generating common sense reasoning results of events

Large Language Model

Transformers

Japanese#Japanese event reasoning #Common sense knowledge generation #Multi-relation prediction

Downloads 158

Release Time : 11/12/2022

Model Overview

This model is a Japanese text generation model based on the T5 architecture, specifically designed for common sense reasoning tasks of events. It can generate possible outcomes, preconditions, motivations, and emotional reactions based on given events.

Model Features

Japanese event common sense reasoning

Specifically designed for Japanese, capable of understanding and generating common sense reasoning related to events

Support for multiple relation types

Supports various relation types such as xNeed (preconditions), xEffect (subsequent events), xIntent (motivations), and xReact (emotional reactions)

Prompt-based generation

Uses prompt template methods to guide the model in generating different types of results through various prompt templates

Model Capabilities

Prediction of event outcomes

Reasoning of event preconditions

Analysis of behavioral motivations

Prediction of emotional reactions

Japanese text generation

Use Cases

Common sense reasoning

Event outcome prediction

Predict possible subsequent situations after an event occurs

Generate multiple possible subsequent events

Behavioral motivation analysis

Analyze the underlying motivations behind a behavior

Generate possible motivation descriptions

Dialogue systems

Dialogue response generation

Provide more common sense responses for dialogue systems

Generate dialogue responses that align with human common sense

🚀 COMET-T5 ja

A fine-tuned T5 model on ATOMIC ja using a text-to-text language modeling objective, introduced in this paper.

🚀 Quick Start

You can use this model directly with a pipeline for text2text generation. Since the generation relies on some randomness, we set a seed for reproducibility:

>>> from transformers import pipeline, set_seed
>>> generator = pipeline('text2text-generation', model='nlp-waseda/comet-t5-base-japanese')
>>> set_seed(42)
>>> generator("次の出来事の後に起こりうることは何ですか: Xが友人に電話する", max_length=30, num_return_sequences=5, do_sample=True)

[{'generated_text': 'Xが友人から返事を得る'},
 {'generated_text': 'Xが会話する'},
 {'generated_text': 'Xが友人に怒られる'},
 {'generated_text': 'Xが退屈しそうな雰囲気になる'},
 {'generated_text': 'Xが友人と会う'}]

💻 Usage Examples

Basic Usage

# The original code example for basic usage
>>> from transformers import pipeline, set_seed
>>> generator = pipeline('text2text-generation', model='nlp-waseda/comet-t5-base-japanese')
>>> set_seed(42)
>>> generator("次の出来事の後に起こりうることは何ですか: Xが友人に電話する", max_length=30, num_return_sequences=5, do_sample=True)

[{'generated_text': 'Xが友人から返事を得る'},
 {'generated_text': 'Xが会話する'},
 {'generated_text': 'Xが友人に怒られる'},
 {'generated_text': 'Xが退屈しそうな雰囲気になる'},
 {'generated_text': 'Xが友人と会う'}]

🔧 Technical Details

Preprocessing

The prompts are different for each relation:

Property	Details
xNeed	次の出来事に必要な前提条件は何ですか:
xEffect	次の出来事の後に起こりうることは何ですか:
xIntent	次の出来事が起こった動機は何ですか:
xReact	次の出来事の後に感じることは何ですか:

📚 Documentation

Evaluation results

The model achieves the following results:

BLEU	BERTScore
39.85	82.37

BibTeX entry and citation info

@InProceedings{ide_nlp2023_event,
    author =    "井手竜也 and 村田栄樹 and 堀尾海斗 and 河原大輔 and 山崎天 and 李聖哲 and 新里顕大 and 佐藤敏紀",
    title =     "人間と言語モデルに対するプロンプトを用いたゼロからのイベント常識知識グラフ構築",
    booktitle = "言語処理学会第29回年次大会",
    year =      "2023",
    url =       "https://www.anlp.jp/proceedings/annual_meeting/2023/pdf_dir/B2-5.pdf"
}

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご