Current Result Document :
| 한글제목(Korean Title) |
BERT-PRF: An Efficient Approach for Intent Detection from Users Search Query Tangina Sultana, Young-Koo Lee |
| 영문제목(English Title) |
BERT-PRF: An Efficient Approach for Intent Detection from Users Search Query Tangina Sultana, Young-Koo Lee |
| 저자(Author) |
Tangina Sultana
Young-Koo Lee
|
| 원문수록처(Citation) |
VOL 48 NO. 02 PP. 0120 ~ 0122 (2021. 12) |
한글내용
(Korean Abstract) |
The exponentially increasing documents on the web demand an efficient approach to search the necessary documents based on the users’ requirement by the intelligent analysis of the user search query. However, the users’ queries are ambiguous because of the characteristics of the natural languages (NL). Therefore, understanding the intent of the users’ query helps the modern search engine to retrieve more relevant documents. Detecting the intent suffers from the lack of sufficient labeled data. Currently, BERT (Bidirectional Encoder Representation from Transformers) has received great attention for processing a wide variety of NL tasks after fine-tuning. There is not much research work exist that explores BERT for intent detection. In this study, we have proposed a novel approach named BERT-PRF for detecting the intent from the users’ search query by using BERT and Pseudo-Relevance Feedback (PRF). We semantically analyze the query by using BERT and improve the performance of the system by using PRF. Experimental results affirm that our proposed approach outperforms the existing scheme in terms of accuracy for detecting the intent.
|
영문내용
(English Abstract) |
|
| 키워드(Keyword) |
BERT
fine-tuning
intent detection
natural language
PRF
|
| 파일첨부 |
PDF 다운로드
|
