|Topic:||Personalizovani asistenti (chatboti) s persistentni pameti pro monitorovani vyvoje dusevniho stavu & ziskavani zpetne vazby|
|Supervisor:||Ing. Marek Otáhal|
|Announce as:||Diplomová práce, Bakalářská práce, Semestrální projekt|
- Most state of the art voice-assistants, chatbots, dialogue-systems are state-less, in a sense that after you end a current conversation, and and resume talking to the bot later (the other day) the bot is unaware of your previous encounter - and will ask you the same questions, oblivious to your preferences, etc.
- This can improve methods of information retrieval, feedback, and datamining. Esp. in cases of long-term period of observation (months), time-consuming questionnaires (takes too long time to answer all at once), and statistical avaraging of random effects (unrelated factors: weather, morning mood,..).
Key new ideas:
- design ways of storing personalized info extracted from conversation about the user
- ability to identify a (returning) user
- design a strategy for asking questions/dialogue contextualy-aware to the user (knowledge-base) and (determined) conditions (ie. "Don't ask Pepa 'How are you feeling today' at 6am until he had his coffee"; or "If
- semantic labeling of questions (in dataset, questionnaire) so we can ask those context-specific
feedback, chatbot, wellbeing, mood, monitoring, personalisation, persistent, knowledge-base, topic classification, NLP