WEKO3
アイテム
言語や文の特徴を考慮した関係情報の活用と抽出
https://doi.org/10.18997/00008348
https://doi.org/10.18997/000083485a07d9c2-b5ed-4795-8740-54ce60bf0cec
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
jou_k_353.pdf (6.9 MB)
|
|
| アイテムタイプ | 学位論文 = Thesis or Dissertation(1) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 公開日 | 2021-06-09 | |||||||||
| 資源タイプ | ||||||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_db06 | |||||||||
| 資源タイプ | doctoral thesis | |||||||||
| タイトル | ||||||||||
| タイトル | Utilization and Extraction of Relation Information Considering Features of Languages and Sentences | |||||||||
| 言語 | en | |||||||||
| タイトル | ||||||||||
| タイトル | 言語や文の特徴を考慮した関係情報の活用と抽出 | |||||||||
| 言語 | ja | |||||||||
| 言語 | ||||||||||
| 言語 | eng | |||||||||
| 著者 |
肥合, 智史
× 肥合, 智史
|
|||||||||
| 抄録 | ||||||||||
| 内容記述タイプ | Abstract | |||||||||
| 内容記述 | This dissertation focuses on the relation information in natural language processing （NLP）. NLP is a technique to process and interpret natural language and an important task in artificial intelligence. In NLP, the relation is often defined as the connections between two words in documents. For example, in the sentence “Apple released new iPhone models,” there is the relation “product-of” between “iPhone” and “Apple.” Documents written in natural language contain relation information and we usually utilize it to interpret natural language. The utilization and extraction of the relation information are important tasks in NLP. First, we discuss the utilization of relation information for natural language interpretation. We deal with sarcasm in this dissertation. Sarcasm presents a negative meaning using positive expressions. Computational sarcasm is important from two perspectives. One perspective is the contribution to sentiment analysis （SA）. Sarcasm often leads to mistakes in the SA task. Therefore, sarcasm detection is important for the SA task. The other perspective is the contribution to the realization of human-like chatbots. The computation of figurative language contributes to the realization of a more natural form of a conversation between humans and machines. We utilize relation information based on the characteristics of sarcasm and verify the effectiveness of the relation information for sarcasm detection. Then, we handle biomedical documents for the relation information extraction. The number of biomedical articles is increasing rapidly. Biomedical relation extraction （RE） techniques determine the existence of a relation between two chemical entities and classify the relation into certain relation types. Since there is a large number of chemicals in biomedical documents, the manual identification of the relations is highly costly. Therefore, biomedical RE is an important task. Recently, the BERT model showed great performance in various NLP tasks. The BERT model is pre-trained on a large scale corpus. BioBERT, which is the BERT model pre-trained on large scale biomedical corpus, improved the performance of biomedical RE. However, the BERT model is a large scale neural network model and requires large-scale computational resources. In this dissertation, we construct a lightweight and high-performance RE model. | |||||||||
| 目次 | ||||||||||
| 内容記述タイプ | TableOfContents | |||||||||
| 内容記述 | 1 Introduction||2 Sarcasm Detection with Relation Information||3 Relation Extraction using Multiple Pre-Trained Models in Biomedical Domain||4 Conclusions | |||||||||
| 備考 | ||||||||||
| 内容記述タイプ | Other | |||||||||
| 内容記述 | 九州工業大学博士学位論文 学位記番号：情工博甲第353号 学位授与年月日：令和3年3月25日 | |||||||||
| キーワード | ||||||||||
| 主題Scheme | Other | |||||||||
| 主題 | Relation Information | |||||||||
| キーワード | ||||||||||
| 主題Scheme | Other | |||||||||
| 主題 | Natural Language Processing | |||||||||
| キーワード | ||||||||||
| 主題Scheme | Other | |||||||||
| 主題 | Sarcasm Detection | |||||||||
| キーワード | ||||||||||
| 主題Scheme | Other | |||||||||
| 主題 | Relation Extraction | |||||||||
| キーワード | ||||||||||
| 主題Scheme | Other | |||||||||
| 主題 | Biomedical Documents | |||||||||
| アドバイザー | ||||||||||
| 嶋田, 和孝 | ||||||||||
| 学位授与番号 | ||||||||||
| 学位授与番号 | 甲第353号 | |||||||||
| 学位名 | ||||||||||
| 学位名 | 博士（情報工学） | |||||||||
| 学位授与年月日 | ||||||||||
| 学位授与年月日 | 2021-03-25 | |||||||||
| 学位授与機関 | ||||||||||
| 学位授与機関識別子Scheme | kakenhi | |||||||||
| 学位授与機関識別子 | 17104 | |||||||||
| 学位授与機関名 | 九州工業大学 | |||||||||
| 学位授与年度 | ||||||||||
| 内容記述タイプ | Other | |||||||||
| 内容記述 | 令和2年度 | |||||||||
| 出版タイプ | ||||||||||
| 出版タイプ | VoR | |||||||||
| 出版タイプResource | http://purl.org/coar/version/c_970fb48d4fbd8a85 | |||||||||
| アクセス権 | ||||||||||
| アクセス権 | open access | |||||||||
| アクセス権URI | http://purl.org/coar/access_right/c_abf2 | |||||||||
| ID登録 | ||||||||||
| ID登録 | 10.18997/00008348 | |||||||||
| ID登録タイプ | JaLC | |||||||||
