{"id":854880,"date":"2022-06-21T11:51:52","date_gmt":"2022-06-21T18:51:52","guid":{"rendered":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/"},"modified":"2023-03-01T14:57:15","modified_gmt":"2023-03-01T22:57:15","slug":"toxic-speech-and-speech-emotions-investigations-of-audio-based-modeling-and-intercorrelations","status":"publish","type":"msr-research-item","link":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/publication\/toxic-speech-and-speech-emotions-investigations-of-audio-based-modeling-and-intercorrelations\/","title":{"rendered":"Toxic Speech and Speech Emotions: Investigations of Audio-based Modeling and Intercorrelations"},"content":{"rendered":"

Content<\/span> moderation<\/span> (CM)<\/span> systems<\/span> have<\/span> become <\/span>essential following the monumental increase in multimodal and <\/span>online social platforms; and while increasingly published work <\/span>focuses on text-based solutions, there is still limited work on <\/span>audio-based methods. In this study we aim to explore relation<\/span>ships between speech emotions and toxic speech, as part of a <\/span>CM scenario. We first investigate an appropriate framework for <\/span>combining speech emotion recognition (SER) and audio-based <\/span>CM models. We then investigate which emotional aspects (i.e., <\/span>attribute, sentiment, or attitude) could contribute the most in <\/span>facilitating audio-based CM recognition platforms. Our experi<\/span>mental results indicate that conventional shared feature encoder <\/span>approaches may fail to capture additional discriminative features <\/span>for boosting audio-based CM tasks while utilizing SER learning. <\/span>We<\/span> further<\/span> investigate<\/span> performance<\/span> trade-offs<\/span> of<\/span> late-fusion <\/span>frameworks for combining SER and CM information. We argue <\/span>that these observations could be attributed to an emotionally- <\/span>biased distribution in the CM scenario, concluding that SER <\/span>could indeed play a role in content moderation frameworks, <\/span>given added application-specific emotional information.<\/span><\/p>\n

\"Accuracy<\/p>\n

Table. Recognition performance of the Content Moderation (CM) model, compared to emotional-based recognition models.
\nAttr-1D: Emotional regressor model trained on arousal and valence attributes (IEMOCAP)
\nSenti-1D: Categorical classifier for 3-class sentiment classes Pos\/Neu\/Neg (IEMOCAP)
\nSenti-5D: Categorical classifier for 3-class sentiment classes Pos\/Neu\/Neg (5 corpora).
\nAtti-1D: Categorical classifier for 3-class sentiment classes Pos\/Neu\/Neg (Cust Support Calls Attitude corpora)<\/p>\n

\"Content

Figure: Trade-off performance trend for different training data size (left); attention weights distribution during testing (right). CM: audio content moderation model. CM+SER: CM model with incorporated emotional information. The distribution of the model’s attention weights shows that SER contributes significantly less compared to the CM features w.r.t. the overall audio-based CM model (i.e., most attention weights for the SER channel fall below 0.2). The major contributor to this may be the negative sentiment-bias in the online CM scenario or the more controlled recording settings of the SER domain when compared to the CM corpus.<\/p><\/div>\n","protected":false},"excerpt":{"rendered":"

Content moderation (CM) systems have become essential following the monumental increase in multimodal and online social platforms; and while increasingly published work focuses on text-based solutions, there is still limited work on audio-based methods. In this study we aim to explore relationships between speech emotions and toxic speech, as part of a CM scenario. We […]<\/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":"","msr-author-ordering":null,"msr_publishername":"IEEE","msr_publisher_other":"","msr_booktitle":"","msr_chapter":"","msr_edition":"","msr_editors":"","msr_how_published":"","msr_isbn":"","msr_issue":"","msr_journal":"","msr_number":"","msr_organization":"EURASIP","msr_pages_string":"","msr_page_range_start":"","msr_page_range_end":"","msr_series":"","msr_volume":"","msr_copyright":"","msr_conference_name":"European Signal Processing Conference (EUSIPCO)","msr_doi":"","msr_arxiv_id":"","msr_s2_paper_id":"","msr_mag_id":"","msr_pubmed_id":"","msr_other_authors":"","msr_other_contributors":"","msr_speaker":"","msr_award":"","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_original_fields_of_study":"","msr_release_tracker_id":"","msr_s2_match_type":"","msr_citation_count_updated":"","msr_published_date":"2022-8-1","msr_highlight_text":"","msr_notes":"","msr_longbiography":"","msr_publicationurl":"","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"https:\/\/2022.eusipco.org\/","msr_journal_url":"","msr_s2_pdf_url":"","msr_year":0,"msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":0,"msr_s2_match_confidence":0,"msr_microsoftintellectualproperty":true,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[],"msr_hide_image_in_river":0,"footnotes":""},"msr-research-highlight":[],"research-area":[243062],"msr-publication-type":[193716],"msr-publisher":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[264253,247741,264840],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-854880","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-audio-acoustics","msr-locale-en_us","msr-field-of-study-audio-and-speech-processing","msr-field-of-study-audio-signal-processing","msr-field-of-study-content-moderation"],"msr_publishername":"IEEE","msr_edition":"","msr_affiliation":"","msr_published_date":"2022-8-1","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":"EURASIP","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":"","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"file","viewUrl":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-content\/uploads\/2022\/06\/Toxic_Speech_and_Speech_Emotions_EUSIPCO_2022.pdf","id":"854892","title":"toxic_speech_and_speech_emotions_eusipco_2022","label_id":"243109","label":0}],"msr_related_uploader":"","msr_citation_count":0,"msr_citation_count_updated":"","msr_s2_paper_id":"","msr_influential_citations":0,"msr_reference_count":0,"msr_arxiv_id":"","msr_s2_author_ids":[],"msr_s2_open_access":false,"msr_s2_pdf_url":null,"msr_attachments":[{"id":854892,"url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-content\/uploads\/2022\/06\/Toxic_Speech_and_Speech_Emotions_EUSIPCO_2022.pdf"}],"msr-author-ordering":[{"type":"guest","value":"wei-cheng-lin","user_id":814636,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=wei-cheng-lin"},{"type":"user_nicename","value":"Dimitra Emmanouilidou","user_id":37461,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Dimitra Emmanouilidou"}],"msr_impact_theme":[],"msr_research_lab":[199565],"msr_event":[],"msr_group":[144923,702211],"msr_project":[559086],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"inproceedings","related_content":{"projects":[{"ID":559086,"post_title":"Audio Analytics","post_name":"audio-analytics","post_type":"msr-project","post_date":"2019-02-08 15:57:54","post_modified":"2023-01-13 13:28:08","post_status":"publish","permalink":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/project\/audio-analytics\/","post_excerpt":"Audio analytics is about analyzing and understanding audio signals captured by digital devices, with numerous applications in enterprise, healthcare, productivity, and smart cities.","_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/559086"}]}}]},"_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/854880","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":5,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/854880\/revisions"}],"predecessor-version":[{"id":924207,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/854880\/revisions\/924207"}],"wp:attachment":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/media?parent=854880"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=854880"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=854880"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=854880"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=854880"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=854880"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=854880"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=854880"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=854880"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=854880"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=854880"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=854880"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=854880"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}