{"id":156375,"date":"2009-04-01T00:00:00","date_gmt":"2009-04-01T00:00:00","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/msr-research-item\/cross-lingual-speech-recognition-under-run-time-resource-constraints\/"},"modified":"2018-10-16T20:22:49","modified_gmt":"2018-10-17T03:22:49","slug":"cross-lingual-speech-recognition-under-run-time-resource-constraints","status":"publish","type":"msr-research-item","link":"https:\/\/www.microsoft.com\/en-us\/research\/publication\/cross-lingual-speech-recognition-under-run-time-resource-constraints\/","title":{"rendered":"Cross-lingual speech recognition under run-time resource constraints"},"content":{"rendered":"<div class=\"asset-content\">\n<p>This paper proposes and compares four cross-lingual and bilingual automatic speech recognition techniques under the constraints of limited memory size and CPU speed. The first three techniques fall into the category of lexicon conversion where each phoneme sequence (PHS) in the foreign language (FL) lexicon is mapped into the native language (NL) phoneme sequence. The first technique determines the PHS mapping through the international phonetic alphabet (IPA) features; The second and third techniques are data-driven. They determine the mapping by converting the PHS into corresponding context-independent and context-dependent hidden Markov models (HMMs) respectively and searching for the NL PHS with the least Kullback-Leibler divergence (KLD) between the HMMs. The fourth technique falls into the category of acoustic-model (AM) merging where the FL\u2019s AM is merged into the NL\u2019s AM by mapping each senone in the FL\u2019s AM to the senone in the NL\u2019s AM with the minimum KLD. We discuss the strengths and limitations of each technique developed, report empirical evaluation results on recognizing English utterances with a Korean recognizer, and demonstrate the high correlation between the average KLD and the word error rate (WER). The results show that the AM merging technique performs the best, achieving 60% relative WER reduction over the IPA-based technique.<\/p>\n<\/div>\n<p><!-- .asset-content --><\/p>\n","protected":false},"excerpt":{"rendered":"<p>This paper proposes and compares four cross-lingual and bilingual automatic speech recognition techniques under the constraints of limited memory size and CPU speed. The first three techniques fall into the category of lexicon conversion where each phoneme sequence (PHS) in the foreign language (FL) lexicon is mapped into the native language (NL) phoneme sequence. The [&hellip;]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","footnotes":""},"msr-content-type":[3],"msr-research-highlight":[],"research-area":[13556,13545,13554],"msr-publication-type":[193716],"msr-product-type":[],"msr-focus-area":[],"msr-platform":[],"msr-download-source":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-156375","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-human-language-technologies","msr-research-area-human-computer-interaction","msr-locale-en_us"],"msr_publishername":"Institute of Electrical and Electronics Engineers, Inc.","msr_edition":"Proceedings of the ICASSP","msr_affiliation":"","msr_published_date":"2009-04-01","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":1,"msr_main_download":"207734","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"file","title":"Cross-Lingual-ICASSP2009.pdf","viewUrl":"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/Cross-Lingual-ICASSP2009.pdf","id":207734,"label_id":0}],"msr_related_uploader":"","msr_attachments":[],"msr-author-ordering":[{"type":"user_nicename","value":"dongyu","user_id":31667,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=dongyu"},{"type":"user_nicename","value":"deng","user_id":31602,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=deng"},{"type":"user_nicename","value":"pengliu","user_id":33227,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=pengliu"},{"type":"text","value":"Jian Wu","user_id":0,"rest_url":false},{"type":"user_nicename","value":"ygong","user_id":34994,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=ygong"},{"type":"user_nicename","value":"alexac","user_id":30932,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=alexac"}],"msr_impact_theme":[],"msr_research_lab":[],"msr_event":[],"msr_group":[],"msr_project":[169434],"publication":[],"video":[],"download":[],"msr_publication_type":"inproceedings","related_content":{"projects":[{"ID":169434,"post_title":"Acoustic Modeling","post_name":"acoustic-modeling","post_type":"msr-project","post_date":"2004-01-29 16:42:42","post_modified":"2019-08-14 14:50:04","post_status":"publish","permalink":"https:\/\/www.microsoft.com\/en-us\/research\/project\/acoustic-modeling\/","post_excerpt":"Acoustic modeling of speech typically refers to the process of\u00a0establishing statistical\u00a0representations for the feature vector sequences\u00a0computed from the speech waveform. Hidden Markov Model (HMM) is one most common type of acoustuc models. Other acosutic models include segmental models, super-segmental models (including hidden dynamic models), neural networks, maximum entropy models, and (hidden) conditional random fields, etc. Acoustic modeling also encompasses \"pronunciation modeling\", which describes how a sequence or multi-sequences of fundamental speech units\u00a0(such as phones or&hellip;","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/169434"}]}}]},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/156375"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":1,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/156375\/revisions"}],"predecessor-version":[{"id":433107,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/156375\/revisions\/433107"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=156375"}],"wp:term":[{"taxonomy":"msr-content-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-content-type?post=156375"},{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=156375"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=156375"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=156375"},{"taxonomy":"msr-product-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-product-type?post=156375"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=156375"},{"taxonomy":"msr-platform","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-platform?post=156375"},{"taxonomy":"msr-download-source","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-download-source?post=156375"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=156375"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=156375"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=156375"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=156375"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=156375"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=156375"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=156375"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}