{"id":1413,"date":"2022-08-05T05:17:21","date_gmt":"2022-08-05T04:17:21","guid":{"rendered":"http:\/\/cafe2sach.com\/?p=1413"},"modified":"2022-08-05T05:18:41","modified_gmt":"2022-08-05T04:18:41","slug":"dkn-mang-hoc-sau-de-xuat-tin-tuc","status":"publish","type":"post","link":"https:\/\/cafe2sach.com\/index.php\/2022\/08\/05\/dkn-mang-hoc-sau-de-xuat-tin-tuc\/","title":{"rendered":"DKN – m\u1ea1ng h\u1ecdc s\u00e2u \u0111\u1ec1 xu\u1ea5t tin t\u1ee9c"},"content":{"rendered":"\n

DKN[1] ( Deep Knowledge-Aware Network l\u00e0 m\u1ed9t m\u00f4 h\u00ecnh h\u1ecdc s\u00e2u k\u1ebft h\u1ee3p th\u00f4ng tin t\u1eeb bi\u1ec3u \u0111\u1ed3 tri th\u1ee9c \u0111\u1ec3 \u0111\u1ec1 xu\u1ea5t tin t\u1ee9c. C\u1ee5 th\u1ec3, DKN s\u1eed d\u1ee5ng ph\u01b0\u01a1ng ph\u00e1p TransX[2] \u0111\u1ec3 h\u1ecdc bi\u1ec3u di\u1ec5n \u0111\u1ed3 th\u1ecb tri th\u1ee9c, sau \u0111\u00f3 \u00e1p d\u1ee5ng m\u00f4 h\u00ecnh CNN, \u0111\u1eb7t t\u00ean l\u00e0 KCNN, \u0111\u1ec3 k\u1ebft h\u1ee3p c\u00e1c th\u1ef1c th\u1ec3 t\u1eeb v\u00e0 xinh vector cu\u1ed1i c\u00f9ng cho m\u1ed9t tin t\u1ee9c m\u1edbi. <\/p>\n\n\n\n

C\u00e1c \u0111\u1eb7c tr\u1ef1c c\u1ee7a DKN <\/h3>\n– DKN la fmootj m\u00f4 h\u00ecnh h\u1ecdc sau d\u1ef1a tr\u00ean n\u1ed9i dung cho CTR d\u1ef1 \u0111o\u00e1n thay v\u00ec ph\u01b0\u01a1ng ph\u00e1p l\u1ecdc d\u1ef1a tr\u00ean ID truy\u1ec1n th\u1ed1ng\n– N\u00f3 s\u1eed d\u1ee5ng c\u00e1c th\u1ef1c th\u1ec3 tri th\u1ee9c chung trong n\u00f4i dung tin t\u1ee9c th\u00f4ng qua vi\u1ec7c h\u1ecdc t\u1eeb semantic-level v\u00e0 knowlege-level c\u1ee7a c\u00e1c b\u00e0i b\u00e1o tin t\u1ee9c\n– DKN s\u1eed d\u1ee5ng module attention \u0111\u1ec3 t\u00ednh to\u00e1n c\u00e1c x\u00e1c nh\u1eadn l\u1ecbch s\u1eed t\u1ed5ng h\u1ee3p c\u1ee7a ng\u01b0\u1eddi d\u00f9ng\n\n\n\n

Th\u1ef1c thi h\u1ec7 th\u1ed1ng khuy\u1ebfn ngh\u1ecb v\u1edbi MIND data <\/h3> \nMIND dataset[3] l\u00e0 m\u1ed9t t\u1eadp d\u1eef li\u1ec7u tin t\u1ee9c ti\u1ebfng Anh quy m\u00f4 l\u1edbn. N\u00f3 \u0111\u01b0\u1ee3c thu th\u1eadp t\u1eeb nh\u1eadt k\u00fd h\u00e0nh vi \u1ea9n danh c\u1ee7a trang web Microsoft News. MIND ch\u01b0a 1.000.000 ng\u01b0\u1eddi d\u00f9ng, 161.013 tin b\u00e0i v\u00e0 15.777.377 nh\u1eadt k\u00fd hi\u1ec3n th\u1ecb. M\u1ecdi tin b\u00e0y \u0111\u1ec1u ch\u01b0a n\u1ed9i dung phong ph\u00fa bao g\u1ed3m, ti\u00eau \u0111\u1ec1, t\u1ed1m t\u1eaft, n\u1ed9i dung, danh m\u1ee5c v\u00e0 c\u00e1c th\u1ef1c th\u1ec3. M\u1ed7i b\u1ea3n ghi hi\u1ec7n th\u1ecb ch\u1ee9c c\u00e1c s\u1ef1 ki\u1ec7n nh\u1ea5p chu\u1ed9t, s\u1ef1 ki\u00ean kh\u00f4ng \u0111\u01b0\u1ee3c nh\u1ea5p chu\u1ed9t v\u00e0 hanh vi nh\u1ea5p chu\u1ed9t tin t\u1ee9c l\u1ecbch s\u1eed c\u1ee7a ng\u01b0\u1eddi d\u00f9ng n\u00e0y tr\u01b0\u1edbc l\u1ea7n hi\u1ec3n th\u1ecb n\u00e0y. \n\n\n\n\n\n

Notebook training h\u1ec7 th\u1ed1ng khuy\u1ebfn ngh\u1ecb s\u1eed d\u1ee5ng t\u1eadp d\u1ef1 li\u1ec7u MIND <\/h3>\n\u0110\u1ea7u v\u00e0o c\u1ee7a model DKN g\u1ed3m c\u00e1c file sau: \n– D\u1eef li\u1ec7u \u0111\u01b0\u1ee3c chia th\u00e0nh training\/validation v\u00e0 testing files.\n– c\u00e1c file d\u1eef li\u1ec7u l\u1ecbch s\u1eed c\u1ee7a ng\u01b0\u1eddi d\u00f9ng\n– t\u00e0i li\u1ec7u \u0111\u1eb7c t\u1ea3 feature \n– word embedding\/entity embedding\/ context embedding files\n\n\n\n\n

C\u00e0i \u0111\u1eb7t v\u00e0 import th\u01b0 vi\u1ec7n:<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Khi ch\u1ea1y c\u00e1c b\u1ea1n c\u00f3 th\u1ec3 b\u1ecb thi\u1ebfu m\u1ed9t s\u1ed1 th\u01b0 vi\u1ec7n, b\u1ea1n c\u00f3 th\u1ec3 d\u00f9ng l\u1ec7nh Pip \u0111\u1ec3 c\u00e0i \u0111\u1eb7t nhanh nh\u01b0 sau:<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

T\u01b0\u01a1ng t\u1ef1 v\u1edbi recommenders : <\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

T\u1ea3i v\u00e0 load d\u1eef li\u1ec7u <\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Kh\u1edfi t\u1ea1o c\u00e1c parameters, c\u0169ng gi\u1ed1ng nh\u01b0 c\u00e1c m\u00f4 h\u00ecnh kh\u00e1c c\u00e1c b\u00ean c\u00f3 th\u1ec3 t\u00f9y ch\u1ec9nh c\u00e1c tham s\u1ed1 nh\u01b0 epoch, batch_size….<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n
\"\"<\/figure>\n\n\n\n

Traing model KKN <\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n
\"\"<\/figure>\n\n\n\n

\u0110\u00e1nh gi\u00e1 DKN model <\/p>\n\n\n\n

Nh\u01b0 \u0111\u00e3 l\u00e0m \u1edf ph\u1ea7n x\u1eed l\u00fd data, ch\u00fang ta c\u00f3 chia d\u1eef li\u1ec7u th\u00e0nh data train\/validation\/test. B\u00e2y gi\u1edd b\u1ea1n s\u1ebd ki\u1ec3m tra k\u1ebft qu\u1ea3 tr\u00ean test set <\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

T\u00e0i li\u1ec7u tham kh\u1ea3o: <\/p>\n\n\n\n

[1] Wang, Hongwei, et al. “DKN: Deep Knowledge-Aware Network for News Recommendation.” Proceedings of the 2018 World Wide Web Conference on World Wide Web. International World Wide Web Conferences Steering Committee, 2018.
[2] Knowledge Graph Embeddings including TransE, TransH, TransR and PTransE. https:\/\/github.com\/thunlp\/KB2E<\/a>
[3] Wu, Fangzhao, et al. “MIND: A Large-scale Dataset for News Recommendation” Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. https:\/\/msnews.github.io\/competition.html<\/a>
[4] GloVe: Global Vectors for Word Representation. https:\/\/nlp.stanford.edu\/projects\/glove\/<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

DKN[1] ( Deep Knowledge-Aware Network l\u00e0 m\u1ed9t m\u00f4 h\u00ecnh h\u1ecdc s\u00e2u k\u1ebft h\u1ee3p th\u00f4ng tin t\u1eeb bi\u1ec3u \u0111\u1ed3 tri th\u1ee9c […]<\/p>\n","protected":false},"author":1,"featured_media":1425,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[565],"tags":[],"class_list":["post-1413","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-he-thong-khuyen-nghi-ai"],"_links":{"self":[{"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/posts\/1413","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/comments?post=1413"}],"version-history":[{"count":2,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/posts\/1413\/revisions"}],"predecessor-version":[{"id":1427,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/posts\/1413\/revisions\/1427"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/media\/1425"}],"wp:attachment":[{"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/media?parent=1413"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/categories?post=1413"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/cafe2sach.com\/index.php\/wp-json\/wp\/v2\/tags?post=1413"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}