Speech and natural language processing for the assessment of customer satisfaction and neuro-degenerative diseases

dc.contributorOrozco Arroyave, Juan Rafael
dc.contributorNöth, Elmar
dc.contributorBocklet, Tobias
dc.creatorPérez Toro, Paula Andrea
dc.date2021-03-04T13:42:22Z
dc.date2021-03-04T13:42:22Z
dc.date2021
dc.date.accessioned2023-08-28T19:33:55Z
dc.date.available2023-08-28T19:33:55Z
dc.identifierhttp://hdl.handle.net/10495/18789
dc.identifier.urihttps://repositorioslatinoamericanos.uchile.cl/handle/2250/8463802
dc.descriptionABSTRACT: Nowadays, the interest in the automatic analysis of speech and text in different scenarios have been increasing. Currently, acoustic analysis is frequently used to extract non-verbal information related to para-linguistic aspects such as articulation and prosody. The linguistic analysis focuses on capturing verbal information from written sources, which can be suitable to evaluate customer satisfaction, or in health-care applications to assess the state of patients under depression or other cognitive states. In the case of call-centers many of the speech recordings collected are related to the opinion of the customers in different industry sectors. Only a small proportion of these calls are evaluated, whereby these processes can be automated using acoustic and linguistic analysis. In the assessment of neuro-degenerative diseases such as Alzheimer's Disease (AD) and Parkinson's Disease (PD), the symptoms are progressive, directly linked to dementia, cognitive decline, and motor impairments. This implies a continuous evaluation of the neurological state since the patients become dependent and need intensive care, showing a decrease of the ability from individual activities of daily life. This thesis proposes methodologies for acoustic and linguistic analyses in different scenarios related to customer satisfaction, cognitive disorders in AD, and depression in PD. The experiments include the evaluation of customer satisfaction, the assessment of genetic AD, linguistic analysis to discriminate PD, depression assessment in PD, and user state modeling based on the arousal-plane for the evaluation of customer satisfaction, AD, and depression in PD. The acoustic features are mainly focused on articulation and prosody analyses, while linguistic features are based on natural language processing techniques. Deep learning approaches based on convolutional and recurrent neural networks are also considered in this thesis.
dc.format151
dc.formatapplication/pdf
dc.formatapplication/pdf
dc.languageeng
dc.publisherGrupo de Investigación en Telecomunicaciones Aplicadas (GITA)
dc.publisherMedellín, Colombia
dc.rightsAtribución-NoComercial-CompartirIgual (CC BY-NC-SA)
dc.rightsinfo:eu-repo/semantics/openAccess
dc.rightshttp://creativecommons.org/licenses/by-nc-sa/2.5/co/
dc.rightshttp://purl.org/coar/access_right/c_abf2
dc.rightshttps://creativecommons.org/licenses/by-nc-sa/4.0/
dc.subjectLanguages
dc.subjectLengua
dc.subjectSpeech
dc.subjectHabla
dc.subjectOral expression
dc.subjectExpresión oral
dc.subjectNervous system diseases
dc.subjectEnfermedad del sistema nervioso
dc.subjectLinguistics
dc.subjectLingüística
dc.subjectAlzheimer's Disease
dc.subjectCustomer Satisfaction
dc.subjectDeep Learning
dc.subjectEmotion Modeling
dc.subjectMachine Learning
dc.subjectNatural Language Processing
dc.subjectParkinson's Disease
dc.subjectSpeech Analysis
dc.subjecthttp://vocabularies.unesco.org/thesaurus/concept308
dc.subjecthttp://vocabularies.unesco.org/thesaurus/concept5828
dc.subjecthttp://vocabularies.unesco.org/thesaurus/concept10648
dc.subjecthttp://vocabularies.unesco.org/thesaurus/concept8193
dc.subjecthttp://vocabularies.unesco.org/thesaurus/concept310
dc.titleSpeech and natural language processing for the assessment of customer satisfaction and neuro-degenerative diseases
dc.typeinfo:eu-repo/semantics/masterThesis
dc.typeinfo:eu-repo/semantics/draft
dc.typehttp://purl.org/coar/resource_type/c_bdcc
dc.typehttps://purl.org/redcol/resource_type/TM
dc.typeTesis/Trabajo de grado - Monografía - Maestría


Este ítem pertenece a la siguiente institución