東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Transfer Learning Techniques for Seq...

Singh, Amanpreet.

FindBook

Google Book

Amazon

博客來

Transfer Learning Techniques for Sequence Labeling in Network File System Specifications.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Transfer Learning Techniques for Sequence Labeling in Network File System Specifications./
作者:	Singh, Amanpreet.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2021,
面頁冊數:	54 p.
附註:	Source: Masters Abstracts International, Volume: 83-01.
Contained By:	Masters Abstracts International83-01.
標題:	Computer science. -
電子資源:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28499444
ISBN:	9798516079740

Transfer Learning Techniques for Sequence Labeling in Network File System Specifications.
Singh, Amanpreet.

Transfer Learning Techniques for Sequence Labeling in Network File System Specifications. - Ann Arbor : ProQuest Dissertations & Theses, 2021 - 54 p.

Source: Masters Abstracts International, Volume: 83-01.

Thesis (M.Sc.)--State University of New York at Stony Brook, 2021.

This item must not be sold to any third party vendors.

Creating a formal model for the purpose of verifying a large scale system requires a lot of effort and expertise. Specification documents of these systems, written in natural language, provide a blueprint of their behavior. In this work, we explore the possibility of modeling the behavior of the Network File System (NFS) protocol using NLP techniques. To that end, we define SpecIR, an efficient representation scheme for NFS specifications and SpecNFS, a dataset of 1200 sentences annotated with this scheme. The conversion of a sentence into SpecIR is designed as a combination of two classification tasks - sequence labeling to identify tokens that can potentially occur in an NFS implementation and link prediction to identify the relation between those tokens. We focus on the first sub-task and benchmark the performance of state-of-the-art language models. Thereafter, we experiment with 3 transfer learning strategies - domain and task adaptation, intermediate task transfer, and label aware learning. For the domain adaptation strategy, we create an additional unlabeled corpus from multiple sources related to our task domain for effective knowledge transfer. We show that for low resource tasks such as ours, transfer learning can lead to continuous performance improvements when applied incrementally. Our best model achieves an F1 score of 63%, an improvement of 1.1% over the baselines.

ISBN: 9798516079740Subjects--Topical Terms:

523869
Computer science.
Subjects--Index Terms:

Artificial Intelligence

Transfer Learning Techniques for Sequence Labeling in Network File System Specifications.
LDR:02570nmm a2200373 4500 001 2282537
005 20211012150152.5
008 220723s2021 ||||||||||||||||| ||eng d
020 $a 9798516079740
035 $a (MiAaPQ)AAI28499444
035 $a AAI28499444
040 $a MiAaPQ $c MiAaPQ
100 1 $a Singh, Amanpreet. $3 3561338
245 1 0 $a Transfer Learning Techniques for Sequence Labeling in Network File System Specifications.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2021
300 $a 54 p.
500 $a Source: Masters Abstracts International, Volume: 83-01.
500 $a Advisor: Balasubramanian, Niranjan.
502 $a Thesis (M.Sc.)--State University of New York at Stony Brook, 2021.
506 $a This item must not be sold to any third party vendors.
520 $a Creating a formal model for the purpose of verifying a large scale system requires a lot of effort and expertise. Specification documents of these systems, written in natural language, provide a blueprint of their behavior. In this work, we explore the possibility of modeling the behavior of the Network File System (NFS) protocol using NLP techniques. To that end, we define SpecIR, an efficient representation scheme for NFS specifications and SpecNFS, a dataset of 1200 sentences annotated with this scheme. The conversion of a sentence into SpecIR is designed as a combination of two classification tasks - sequence labeling to identify tokens that can potentially occur in an NFS implementation and link prediction to identify the relation between those tokens. We focus on the first sub-task and benchmark the performance of state-of-the-art language models. Thereafter, we experiment with 3 transfer learning strategies - domain and task adaptation, intermediate task transfer, and label aware learning. For the domain adaptation strategy, we create an additional unlabeled corpus from multiple sources related to our task domain for effective knowledge transfer. We show that for low resource tasks such as ours, transfer learning can lead to continuous performance improvements when applied incrementally. Our best model achieves an F1 score of 63%, an improvement of 1.1% over the baselines.
590 $a School code: 0771.
650 4 $a Computer science. $3 523869
650 4 $a Artificial intelligence. $3 516317
653 $a Artificial Intelligence
653 $a BERT
653 $a Formal verification
653 $a NFS
653 $a NLP
653 $a Transfer learning
690 $a 0984
690 $a 0800
710 2 $a State University of New York at Stony Brook. $b Computer Science. $3 1674709
773 0 $t Masters Abstracts International $g 83-01.
790 $a 0771
791 $a M.Sc.
792 $a 2021
793 $a English
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28499444