Behaviour-based language-interactive speaking systems (BLISS)
Download as PPTX, PDF0 likes85 views
The document describes the BLISS project which aims to develop a behavior-based language-interactive speaking system to help elderly people with health problems like dementia remain independent at home. The system will use natural language processing techniques on big data like text, speech, and video from interviews to build a personalized model of individual clients and facilitate their self-management of health, well-being, and happiness. It will be evaluated on a large scale to assess how well it can increase client health outcomes.
Behaviour-based language-interactive speaking systems (BLISS)
- 1. Radboud University Nijmegen NWO Data2Person project BLISS: Behaviour-based Language-Interactive Speaking Systems Helmer Strik NovoLanguage BV Radboud Univ. Nijmegen CLS : Centre for Language Studies CLST : Centre for Language and Speech Technology AI – Big Data
- 2. Radboud University Nijmegen NWO Data2Person project: BLISS BLISS: Behaviour-based Language-Interactive Speaking Systems Applicants: Helmer Strik, Catia Cucchiarini Consortium: CLS, CLST (RU): Louis ten Bosch, Iris Hendrickx HMI (Twente Univ.) Mariët Theune ReadSpeaker Staffan Meij, Esther Judd-Klabbers Games for Health Projects Jurriaan van Rijswijk, Roland Goetgeluk, Rob Tieben • Health care organisation & clients http://hstrik.ruhosting.nl/bliss/ CLS all: BLISS - Reading Tutor 2 AI – Big Data
- 3. Radboud University Nijmegen Now Increasing: Elderly, with (healthcare) problems, e.g. dementia • remain at home, self-management Big data: text, speech Games for Health Projects GfH_EU - BLISS 3 QuestionR and HealthyR – text InterviewR – speech / spoken data
- 4. Radboud University Nijmegen NLG TTS ASR NLU Dial. Manager Speech SDS Speech Client ClientStandard SDSKnowledge Bases Domain Ontology Info NLG: Natural Language Generation TTS: Text-To-Speech NLU: Natural Language Understanding ASR: Autom. Speech Recognition SDS: Spoken Dialog System GfH_EU - BLISS
- 5. Radboud University Nijmegen NLG TTS Dialogue History Big Data: Text - Sp. ASR NLU Dial. Manager Speech BLISS Speech Client Client-- BLISS --Knowledge Bases Domain Ontology IE Framing IE: Info. Extraction NLG: Natural Language Generation TTS: Text-To-Speech NLU: Natural Language Understanding ASR: Autom. Speech Recognition Goal: Big Data + IE => personalized SDS: BLISS GfH_EU - BLISS Happiness model
- 6. Radboud University Nijmegen Innovative aspects - Research Questions IE on speech (interviews, dialogs, ...) using Deep Neural Nets (DNNs); combined with IE on texts. Speech contains much more information than just the words, verbal info. RQ1. How to extract relevant knowledge about the client’s health, wellbeing and happiness? Big Data personalized SDS that communicates with clients ... ... to facilitate their self-management of health, wellbeing & happiness. Happiness and health are related - Happiness Model (Veenhuizen, 2008) RQ2. How to incorporate all knowledge in a ‘Model of Happiness’? RQ3. ... and use it to personalize the SDS? Large scale use by clients. Integrated in applications on agenda management that use gamification. Happiness ~ The ideal schedule - sequence of daily activities. RQ4. To what extent does the SDS help increase the client’s health, wellbeing and happiness? CLS all: BLISS - Reading Tutor 6
- 7. Radboud University Nijmegen Now & Future Increasing: Elderly, with problems, e.g. dementia • remain at home, self-management Big data: text, speech & video, wearables • Speech & video - not only words => monitoring Games for Health Projects GfH_EU - BLISS 7 CalendR –ideal day schedule InterviewR – speech / spoken data
- 8. Radboud University Nijmegen Thanks for your attention. Questions ? For more information see: http://hstrik.ruhosting.nl/ http://hstrik.ruhosting.nl/contact/ http://hstrik.ruhosting.nl/projects/ • http://hstrik.ruhosting.nl/bliss/ GfH_EU - BLISS 8