Hasil Pencarian  ::  Simpan CSV :: Kembali

Abstrak :
Saat mencari artikel yang diterbitkan dalam periode waktu yang panjang, pengguna biasanya membutuhkan dokumen yang tidak hanya relevan terhadap topik tetapi juga relevan terhadap waktu. Tesis ini membahas tentang pemeringkatan dokumen dengan konsep waktu atau temporal, di mana dokumen dengan topik dan waktu yang dekat dengan query harus diberikan peringkat yang lebih tinggi. Untuk mengetahui waktu yang sesuai dengan query pengguna, tesis ini mengembangkan teknik pemeringkatan temporal yang diperoleh dari distribusi keterkaitan kata dari waktu ke waktu yang dipelajari pada sebuah arsip berita dalam Bahasa Indonesia. Keterkaitan kata dipelajari menggunakan Dynamic Embeddings yaitu Word2Vec yang dipelajari terpisah dari waktu ke waktu, OrthoTrans-Word2Vec dan Dynamic Bernoulli Embeddings. Dalam menangkap relevansi secara topikal, model yang diusulkan menggunakan Dual Embedding Space Model (DESM) yang dibangun dengan teknik temporal sesuai dengan waktu pembuatan dokumen. Untuk meningkatkan nilai presisi, model tersebut juga menggunakan sebuah klasifikasi temporal yang dipelajari menggunakan Support Vector Machine (SVM) dan Basis Threshold. Skor tertinggi dicapai ketika membangun model menggunakan Word2Vec yaitu 66% pada presisi rata- rata dan 68% pada presisi awal. Model tersebut juga terbukti efektif pada query temporal yang memiliki pola seperti tren, periodisitas, dan musiman. ......When searching for articles published over time, users usually require documents that are not only topically relevant but also created during relevant time periods. This thesis studied about document ranking with temporal concept, where documents with topic and time that closely match with the queries should be ranking higher. In order to capturing the time of user query intent, the models developed with temporal ranking technique from distribution of word relatedness over time learned from news archive in Bahasa Indonesia. Word relatedness captured by using Dynamic embeddings, such as Word2Vec learned separately over time, OrthoTrans-Word2Vec dan Dynamic Bernoulli Embeddings. For capturing topical relevance, the proposed model used Dual Embedding Space Model (DESM) in the temporal technique according to document timestamp. The model also combined with temporal classification using Support Vector Machine (SVM) and threshold-based strategy. The highest score was achieved by a model using Word2Vec, which is 66% in average precision and 68% in early precision. The result also showed that the model is effective in capturing temporal patterns such as spikes, periodicity, and seasonality

Abstrak :
ABSTRAK
OneSearch merupakan alat penemuan informasi terkini di Indonesia yang dikembangkan atas tuntutan pelayanan informasi yang cepat, dinamis, dan mengikuti perkembangan teknologi informasi. Sayangnya belum ada penelitian empiris yang mengkaji bagaimana pemanfaatan OneSearch dari sisi pengguna. Kajian ini menganalisis data log transaksi server Indonesia OneSearch berjumlah 256.953 rekaman klik yang diuraikan, dikodekan dan dianalisis pola pergerakan dan pola kueri pengguna. Hasilnya, mayoritas pengguna sangat bergantung dengan pencarian sederhana dan fitur suggestion records, serta rendahnya penggunaan pencarian tingkat lanjut dan laman perpustakaan pribadi. Temuan lainnya menunjukan bahwa sebagian besar kueri merupakan kueri sederhana dan rendahnya penggunaan berbagai operator pencarian yang disediakan. Penelitian ini mengusulkan agar pada waktu mendatang, OneSearch fokus meningkatkan model algoritma relevansi hasil pencarian sederhana, mengembangkan fitur rekomendasi agregasi pada laman perpustakaan pribadi, dan peningkatan formulasi kueri interaktif untuk mengurangi kegagalan pencarian.

---

ABSTRACT
AbstractIndonesia OneSearch is a current digital portal that operates under the National Library of Indonesia to fulfill many demands for a more rapidly dynamic and up to date information searching services on a nationwide scale as it indexes the metadata of resources spread throughout its partner organizations. However, there has not been one empirical research that analyzes the actual data coming from the users on their usage. The 256,953 clickstreams logs are being cleaned, parsed, coded, then analyzed using integrated clickstream data analysis framework. Consequently, it is found that the majority of users depend on a single box search and suggestion records feature, while only a small part user utilize the advanced search and personal library page. Another finding includes that most of the queries categorized as a simple query, and the use of search operators is still low. This study proposes that in the forthcoming years, OneSearch must optimize the search algorithms, adding an information recommendation feature on user rsquo s personal library page, devise a help search page to assist users as a part of information literacy, and make an improvement on faceted search features as well as interactive query formulations to reduce any search failures.

Abstrak :
Subdivide and conquer! "Magic search: getting the best results from your catalog and beyond" showcases how to increase the power of Library of Congress Subject Heading (LCSH) subdivisions to produce astonishing results from your searches. Rebecca S. Kornegay and Heidi E. Buchanan, experienced reference librarians, and Hildegard B. Morgan, an expert cataloger, explain how, when used wisely, LCSH subdivisions can save time and provide a new level of precision in information retrieval for patrons of the library. "Magic Search" presents the 467 best-performing LCSH subdivisions that speak to the kinds of research questions librarians handle every day. This quick reference format, along with a handy index, offers a useful tool to keep for quick reference rather than a cumbersome tome to be read from cover to cover. In addition, this book provides: a thematic arrangement of LC subdivisions that yield the most successful search; chapters on discipline-specific subdivisions to hone effective search terms; and, precise, professional vocabulary useful in searches and explained in easy-to-understand language. Grasping the importance and having command of LC subdivisions, now appearing in unexpected places beyond the library catalog, is key in this rapidly evolving, 21st-century information environment. No other work explores the LCSH subdivisions is such detail or with such commitment, making this book vital to every reference desk.