Implementasi Teknik Word Embedding Untuk Rekomendasi Hasil Pencarian Katalog Online Menggunakan Algoritma WORD2VEC

Abstract

The purpose of this study is to apply the word2vec algorithm to recommend search results in online catalogs. The reason for taking this title is because, based on the results of observations and the observations of researchers, the data search process, especially in online catalogs, only reaches the syntactic level. So that the results are given only up to the syntactic level. Therefore, researchers utilize the word2vec algorithm, which has the ability to represent words at the semantic level, to carry out a search process where the results of this process are used as alternative search results or recommendations for search results. The data that the researchers used was data on 12,701 book titles in the Raja Ali Haji Maritime University library. To evaluate the recommendations for the search results obtained, the researcher tested the recommendation system for search results using several scenarios, and then the results were measured using a precision test on the k document (P@k). From the results of the precision test measurements on the k document, various results were found. Scenarios 1 and 2 show a fairly high precision value with a value of 0.53 and 0.59, while for testing in scenarios 3, 4, and 5, the resulting precision value is relatively low with a value of 0.50, 0.42, and 0.14.