{"id":136,"date":"2025-06-18T16:23:16","date_gmt":"2025-06-18T16:23:16","guid":{"rendered":"https:\/\/blog.chataignon.org\/joseph\/?p=136"},"modified":"2025-06-18T16:26:38","modified_gmt":"2025-06-18T16:26:38","slug":"visualiseur-de-vecteurs-semantiques","status":"publish","type":"post","link":"https:\/\/blog.chataignon.org\/joseph\/fr\/post-136\/visualiseur-de-vecteurs-semantiques\/","title":{"rendered":"Visualiseur de vecteurs s\u00e9mantiques"},"content":{"rendered":"\n

Le visualiseur: https:\/\/word-embeddings.wbkolleg.unibe.ch\/<\/a>
Code: https:\/\/gitlab.com\/JChataigne\/embeddings-visualiser<\/a><\/p>\n\n\n\n

En tant qu’\u00e9tudiant ing\u00e9nieur d\u00e9couvrant les r\u00e9seaux de neurones en 2017, avant l’introduction des Transformers; l’une des choses les plus intrigantes et amusantes que j’ai d\u00e9couvertes \u00e9tait que des op\u00e9rations arithm\u00e9tiques pouvaient \u00eatre appliqu\u00e9es aux vecteurs s\u00e9mantiques (word embeddings<\/em> en anglais) et que cela avait effectivement un sens. Pour citer le papier<\/a> qui a introduit cette notion :<\/p>\n\n\n\n

\n

De mani\u00e8re surprenante, nous avons constat\u00e9 que la similarit\u00e9 des repr\u00e9sentations de mots va au-del\u00e0 des simples r\u00e9gularit\u00e9s syntaxiques. En utilisant une technique de d\u00e9calage de mots o\u00f9 des op\u00e9rations alg\u00e9briques simples sont effectu\u00e9es sur les vecteurs de mots, il a \u00e9t\u00e9 montr\u00e9, par exemple, que vecteur(‘Roi’) – vecteur(‘Homme’) + vecteur(‘Femme’) donne un vecteur qui est le plus proche de la repr\u00e9sentation vectorielle du mot ‘Reine’. Dans ce papier, nous essayons de maximiser la pr\u00e9cision de ces op\u00e9rations vectorielles en d\u00e9veloppant de nouvelles architectures de mod\u00e8les qui pr\u00e9servent les r\u00e9gularit\u00e9s lin\u00e9aires entre les mots.<\/em><\/p>\n<\/blockquote>\n\n\n\n

C’est une mani\u00e8re intuitive et agr\u00e9able de montrer comment les vecteurs peuvent capturer la s\u00e9mantique. Cet exemple particulier (\u00ab\u00a0Roi\u00a0\u00bb – \u00ab\u00a0Homme\u00a0\u00bb + \u00ab\u00a0Femme\u00a0\u00bb \u2248 \u00ab\u00a0Reine\u00a0\u00bb) est encore tr\u00e8s souvent utilis\u00e9 dans les cours d’introduction aux r\u00e9seaux de neurones.<\/p>\n\n\n\n

Mais en suivant (et plus tard en donnant) de tels cours, j’ai toujours \u00e9t\u00e9 d\u00e9\u00e7u que seuls des exemples fixes soient donn\u00e9s, car (1) ils auraient pu \u00eatre s\u00e9lectionn\u00e9s pour faire para\u00eetre le ph\u00e9nom\u00e8ne plus important et (2) je voulais pouvoir explorer d’autres d’exemples. La seule fa\u00e7on de voir plus d’exemples \u00e9tait d’ex\u00e9cuter le code soi-m\u00eame, ce qui n\u00e9cessite soit un temps, soit des comp\u00e9tences que la plupart des gens n’ont pas.<\/p>\n\n\n\n

J’ai donc pris quelques jours pour construire un visualiseur de vecteurs s\u00e9mantiques et je l’ai rendu public. Vous pouvez cliquer sur ce lien<\/a> pour l’essayer.<\/p>\n\n\n\n

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

J’ai \u00e9galement cr\u00e9\u00e9 une version pour visualiser les vecteurs\/embeddings<\/em> de documents, mais je n’ai pas de serveur pour celle-ci pour le moment. Vous pouvez toujours l’ex\u00e9cuter localement, le code est public.<\/p>\n\n\n\n

<\/p>\n","protected":false},"excerpt":{"rendered":"

Le visualiseur: https:\/\/word-embeddings.wbkolleg.unibe.ch\/Code: https:\/\/gitlab.com\/JChataigne\/embeddings-visualiser En tant qu’\u00e9tudiant ing\u00e9nieur d\u00e9couvrant les r\u00e9seaux de neurones en 2017, avant l’introduction des Transformers; l’une […]<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[11,32],"tags":[],"class_list":["post-136","post","type-post","status-publish","format-standard","hentry","category-ia","category-projet"],"_links":{"self":[{"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/posts\/136","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/comments?post=136"}],"version-history":[{"count":3,"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/posts\/136\/revisions"}],"predecessor-version":[{"id":142,"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/posts\/136\/revisions\/142"}],"wp:attachment":[{"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/media?parent=136"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/categories?post=136"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blog.chataignon.org\/joseph\/wp-json\/wp\/v2\/tags?post=136"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}