TY - GEN
T1 - Applying a Vector Search Method in Reference Service Question-Answer Retrieval Systems
AU - Yang, Te Lun
AU - Huang, Guan Lun
AU - Tseng, Yuen Hsien
N1 - Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2023
Y1 - 2023
N2 - We have developed a reference service question-answer retrieval system prototype based on dense vector search technology. When librarians receive reference questions, they can utilize reference service question-answer systems to determine if similar questions have been asked before and how they were resolved. Readers can also search the system to see if their questions have been answered. However, traditional reference service question-answer retrieval systems primarily rely on keyword-based searches, which may not capture questions with different phrasing but identical meaning. Vector search has emerged as a promising approach to address this limitation, and it has been widely researched and applied in various domains. Nevertheless, there is a lack of research on applying vector search in reference service question-answer retrieval systems. Thus, we conducted initial experiments in this area. Leveraging ElasticSearch vector search technology, which can efficiently search through large datasets, we adopted this technology in our study. For building our system, we collected question-answer pairs from the Taipei Public Library’s Online Reference Service Platform, one of Taiwan’s most prominent online reference services. We report the data set we used and the system architecture.
AB - We have developed a reference service question-answer retrieval system prototype based on dense vector search technology. When librarians receive reference questions, they can utilize reference service question-answer systems to determine if similar questions have been asked before and how they were resolved. Readers can also search the system to see if their questions have been answered. However, traditional reference service question-answer retrieval systems primarily rely on keyword-based searches, which may not capture questions with different phrasing but identical meaning. Vector search has emerged as a promising approach to address this limitation, and it has been widely researched and applied in various domains. Nevertheless, there is a lack of research on applying vector search in reference service question-answer retrieval systems. Thus, we conducted initial experiments in this area. Leveraging ElasticSearch vector search technology, which can efficiently search through large datasets, we adopted this technology in our study. For building our system, we collected question-answer pairs from the Taipei Public Library’s Online Reference Service Platform, one of Taiwan’s most prominent online reference services. We report the data set we used and the system architecture.
KW - Dense vector search
KW - Full-text search engine
KW - Question-answer system
KW - Reference service
KW - Taipei public library
UR - http://www.scopus.com/inward/record.url?scp=85180158085&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85180158085&partnerID=8YFLogxK
U2 - 10.1007/978-981-99-8085-7_18
DO - 10.1007/978-981-99-8085-7_18
M3 - Conference contribution
AN - SCOPUS:85180158085
SN - 9789819980840
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 204
EP - 209
BT - Leveraging Generative Intelligence in Digital Libraries
A2 - Goh, Dion H.
A2 - Chen, Shu-Jiun
A2 - Tuarob, Suppawong
PB - Springer Science and Business Media Deutschland GmbH
T2 - 25th International Conference on Asia-Pacific Digital Libraries, ICADL 2023
Y2 - 4 December 2023 through 7 December 2023
ER -