{"id":10953,"date":"2020-07-08T10:24:40","date_gmt":"2020-07-08T07:24:40","guid":{"rendered":"https:\/\/umram.bilkent.edu.tr\/?p=10953"},"modified":"2020-07-08T10:24:40","modified_gmt":"2020-07-08T07:24:40","slug":"umramdan-yeni-bir-makale-imparting-interpretability-to-word-embeddings-while-preserving-semantic-structure","status":"publish","type":"post","link":"https:\/\/umram.bilkent.edu.tr\/index.php\/tr\/2020\/07\/08\/umramdan-yeni-bir-makale-imparting-interpretability-to-word-embeddings-while-preserving-semantic-structure\/","title":{"rendered":"UMRAM’dan Yeni Bir Makale: Imparting interpretability to word embeddings while preserving semantic structure"},"content":{"rendered":"

Koc Lab’\u0131n do\u011fal dil i\u015fleme \u00fczerine \u00e7al\u0131\u015fmas\u0131, Cambridge \u00dcniversitesi Yay\u0131nlar\u0131 Do\u011fal Dil M\u00fchendisli\u011fi Dergisi’nde yay\u0131nlanan \u201cAnlamsal yap\u0131y\u0131 korurken kelime d\u00fc\u011f\u00fcnlerine yorumlanabilirlik kazand\u0131rmak\u201d ba\u015fl\u0131kl\u0131 bir makaleyle<\/a> sonu\u00e7land\u0131. Makalenin yazarlar\u0131 L\u00fcfti Kerem \u015eenel<\/a>, \u0130hsan Utlu, Furkan \u015eahinu\u00e7<\/a>, Haldun \u00d6zakta\u015f ve Aykut Ko\u00e7<\/a>.<\/p>\n

Kelime d\u00fc\u011f\u00fcnleri bir\u00e7ok NLP g\u00f6revinde \u00e7ok \u00f6nemli ara\u00e7lard\u0131r. Bununla birlikte, kara kutu do\u011falar\u0131 kelime d\u00fc\u011f\u00fcnlerini a\u00e7\u0131klanabilir ve yorumlanabilir olmaktan al\u0131koyar. Bu \u00e7al\u0131\u015fmada, ilk kez, geleneksel GloVe algoritmas\u0131na harici bir objektif fonksiyon ekleyerek, kelimelerin d\u00fc\u011f\u00fcnlerine yorumlanabilir olma \u00f6zelli\u011fi verilmektedir. Ayr\u0131ca, yorumlanabilir kelime d\u00fc\u011f\u00fcnlerinin ortak \u00f6\u011frenimi, kelime d\u00fc\u011f\u00fcnlerinin alt\u0131nda yatan anlamsal yap\u0131y\u0131 bozmaz.<\/p>\n

Kelime d\u00fc\u011f\u00fcnleri i\u00e7indeki yorumlanabilirlik derecesindeki art\u0131\u015f, insan deneklerin elle de\u011ferlendirilmesinin yan\u0131 s\u0131ra nitel ve nicel testlerle de g\u00f6sterilmi\u015ftir. Elde edilen sonu\u00e7lar, ifade edilen kelime d\u00fc\u011f\u00fcnlerinin literat\u00fcrdeki benzer \u00e7al\u0131\u015fmalardan daha iyi performans g\u00f6sterdi\u011fini g\u00f6stermektedir.<\/p>\n

NLP alan\u0131ndaki a\u00e7\u0131klanabilir AI algoritmalar\u0131n\u0131 geli\u015ftirmek i\u00e7in \u00e7al\u0131\u015fman\u0131n katk\u0131s\u0131 esast\u0131r.<\/p>\n

Abstract\u00a0<\/strong><\/span><\/p>\n

As a ubiquitous method in natural language processing, word embeddings are extensively employed to map semantic properties of words into a dense vector representation. They capture semantic and syntactic relations among words, but the vectors corresponding to the words are only meaningful relative to each other. Neither the vector nor its dimensions have any absolute, interpretable meaning. We introduce an additive modification to the objective function of the embedding learning algorithm that encourages the embedding vectors of words that are semantically related to a predefined concept to take larger values along a specified dimension, while leaving the original semantic learning mechanism mostly unaffected. In other words, we align words that are already determined to be related, along predefined concepts. Therefore, we impart interpretability to the word embedding by assigning meaning to its vector dimensions. The predefined concepts are derived from an external lexical resource, which in this paper is chosen as Roget\u2019s Thesaurus. We observe that alignment along the chosen concepts is not limited to words in the thesaurus and extends to other related words as well. We quantify the extent of interpretability and assignment of meaning from our experimental results. Manual human evaluation results have also been presented to further verify that the proposed method increases interpretability. We also demonstrate the preservation of semantic coherence of the resulting vector space using word-analogy\/word-similarity tests and a downstream task. These tests show that the interpretability-imparted word embeddings that are obtained by the proposed framework do not sacrifice performances in common benchmark tests.<\/p>\n<\/div><\/div><\/div><\/div><\/div><\/section>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":7,"featured_media":10914,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[445],"tags":[],"_links":{"self":[{"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/posts\/10953"}],"collection":[{"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/users\/7"}],"replies":[{"embeddable":true,"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/comments?post=10953"}],"version-history":[{"count":1,"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/posts\/10953\/revisions"}],"predecessor-version":[{"id":10954,"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/posts\/10953\/revisions\/10954"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/media\/10914"}],"wp:attachment":[{"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/media?parent=10953"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/categories?post=10953"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/umram.bilkent.edu.tr\/index.php\/wp-json\/wp\/v2\/tags?post=10953"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}