{"id":793,"date":"2019-04-16T09:46:20","date_gmt":"2019-04-16T09:46:20","guid":{"rendered":"https:\/\/speechlabs.pl\/?p=793"},"modified":"2019-04-16T09:55:02","modified_gmt":"2019-04-16T09:55:02","slug":"arm-1-2nd-in-poleval-2019-asr-competition","status":"publish","type":"post","link":"https:\/\/speechlabs.pl\/en\/arm-1-2nd-in-poleval-2019-asr-competition\/","title":{"rendered":"ARM-1 2nd in PolEval 2019 ASR competition"},"content":{"rendered":"<p>\r\n<a href=\"https:\/\/speechlabs.pl\/en\/our-offer\/arm\/\">ARM-1 engine<\/a> was one of the systems competing within Task 5: Automatic Speech Recognition at <a href=\"http:\/\/poleval.pl\/tasks\/task5\">PolEval 2019<\/a>. PolEval is a series of competitions organized every year for natural language processing tools for Polish. The event was inspired by <a href=\"http:\/\/alt.qcri.org\/semeval2019\/\">SemEval <\/a> International Workshop on Semantic Evaluation.\r\n<\/p><br\/>\r\n<img decoding=\"async\" src=\"https:\/\/speechlabs.pl\/wp-content\/uploads\/2019\/04\/IMG_2408.jpg\" alt=\"\" width=\"650\"  class=\"alignnone size-full wp-image-800\" srcset=\"https:\/\/speechlabs.pl\/wp-content\/uploads\/2019\/04\/IMG_2408.jpg 4600w, https:\/\/speechlabs.pl\/wp-content\/uploads\/2019\/04\/IMG_2408-300x200.jpg 300w, https:\/\/speechlabs.pl\/wp-content\/uploads\/2019\/04\/IMG_2408-768x512.jpg 768w, https:\/\/speechlabs.pl\/wp-content\/uploads\/2019\/04\/IMG_2408-1024x683.jpg 1024w\" sizes=\"(max-width: 4600px) 100vw, 4600px\" \/>\r\n<p\/>\r\n<p>\r\nVarious tools and system compete against each other within certain tasks, selected each year by the organizers. This year one of the tasks was to <b>transcribe audio recordings of speech in a noisy environment<\/b> and was designed for Large-Vocabulary \r\nContinuous Speech Recognition (LVSCR) systems. ARM-1 was submitted in open competition subtask for systems that could use any training data, as opposed to fixed competition for systems which could use only a few corpora listed by the organizers. In both cases, this year speech recordings from lower and higher house of the Polish Parliament were excluded from data used for system training, since a subset of these recordings was used as a test set. During evaluation phase only systems developed by the competitors could have been used. <\/p>\r\n<p>\r\n<img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/speechlabs.pl\/wp-content\/uploads\/2019\/04\/graph.png\" alt=\"\" width=\"650\" height=\"353\" class=\"alignnone size-full wp-image-806\" srcset=\"https:\/\/speechlabs.pl\/wp-content\/uploads\/2019\/04\/graph.png 650w, https:\/\/speechlabs.pl\/wp-content\/uploads\/2019\/04\/graph-300x163.png 300w\" sizes=\"auto, (max-width: 650px) 100vw, 650px\" \/>\r\n<\/p>\r\n<p>\r\nThe test set consisted of 29 recordings of various speakers with a total duration of nearly 48 minutes and the results were evaluated by determining WER (Word Error Rate). ARM-1 had the best <a href=\"http:\/\/poleval.pl\/results\">result<\/a> in its subtask with WER = 26,4% and correctness of 77% over all recordings, and took second place among all competing systems.\r\n<\/p>","protected":false},"excerpt":{"rendered":"ARM-1 engine was one of the systems competing within Task 5: Automatic Speech Recognition at PolEval 2019. PolEval is a series of competitions organized every year for natural language processing tools for Polish. The event was inspired by SemEval International Workshop on Semantic Evaluation. Various tools and system compete against each other within certain tasks, [&hellip;]","protected":false},"author":2,"featured_media":814,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_locale":"en_US","_original_post":"793","footnotes":""},"categories":[1],"tags":[],"class_list":["post-793","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized","en-US"],"meta_box":[],"_links":{"self":[{"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/posts\/793","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/comments?post=793"}],"version-history":[{"count":13,"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/posts\/793\/revisions"}],"predecessor-version":[{"id":808,"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/posts\/793\/revisions\/808"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/media\/814"}],"wp:attachment":[{"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/media?parent=793"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/categories?post=793"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/speechlabs.pl\/wp-json\/wp\/v2\/tags?post=793"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}