Core Concepts
Detecting salient information in space innovation ideas using a new dataset.
Abstract
1. Introduction:
Detecting salient text parts to mitigate information overflow.
Sequential sentence classification used for this task.
2. Related Work:
Datasets for role sentence classification mainly from academic publications.
3. SPACE-IDEAS Dataset:
Roles identified: Challenge, Proposal, Elaboration, Benefits, Context.
Annotation process with high-quality annotations.
3.1. SPACE-IDEAS+:
Larger dataset annotated with OpenAI's gpt-3.5-turbo model.
3.2. Fields and Label Distribution:
Knowledge fields distribution in the datasets.
3.3. Relation to Existing Datasets:
SPACE-IDEAS covers innovative ideas in the space domain with various writing styles.
4. Detecting Salient Information:
Multi-class classification task assigning labels to sentences in idea descriptions.
5. Experiments:
Evaluation results of different classifiers trained on the datasets.
6. Conclusion:
Introduction of SPACE-IDEAS dataset and its impact on sequential sentence classification.
Stats
空間イノベーションアイデアにおける顕著な情報検出のための新しいデータセット。
主要なメッセージは、宇宙イノベーションアイデアで顕著な情報を検出するための新しいデータセットを使用することです。