Ten years of experience in the research and development of conversational interfaces (aka dialogue systems, voice assistants and chatbots), has led us to know the nuts and bolts of the technology and user experience underneath these conversational systems.
Personality is the key to the success of dialogue systems, as people require natural conversations to fulfil their needs or get things done. Besides, implementing emotional intelligence leads to empathetic systems that provide value in certain circumstances.
If you’re going to make bots today, give them personality. Brands have a soul, and their bots should have a soul too.
Conversational interfaces need to find their killer use cases yet, but providing them with personality will make them more usable creating experiences and bonds with users. Not only necessary if a company wants to build a link with their customers but effective to avoid certain situations or to support and help people in crucial moments.
It’s been already proven that having a personality helps increase your engagement with your users and reduces some problems related to misleading expectations. Nowadays the situation is solved with tailor-made solutions from agencies that create personalities ad-hoc.
NAIZ automates the design of personalities and the management of different users and contexts. Take care of the core functionalities in your chatbot (built with other tools) and leave the management of the personality to NAIZ. NAIZ models traits that are interesting in customer service (patience and politeness) and chit chat situations (friendly).
Our core is a dialogue system that comprises:
- Understanding of user smalltalk intention and emotions (Natural Language Understanding).
- Management of these emotions and intention inside the context of the conversation (Dialog Management).
- Answering based on NAIZ own intention, personalizing the response (Response Generation).
Understanding and modeling personality traits and emotions is complex. In our algorithms and models, built with Machine Learning and Natural Language Processing tecniques, we have taken into account several psychological theories of personality, traits and emotions.
NATURAL LANGUAGE UNDERSTANDING
Models to detect user intentions from smalltalk conversations, emotions detection from words and emojis used, and mood classification.
Our dialogue system manages conversations with users and contexts, emotions from users (sad, anger, happiness, etc.) and emotions that the dialogue system feels itself due to interactions with the users: it didn’t understand the user intention so feels confused or sad.
Traits affect how the dialogue system manages these emotions and the type of answer it will provide. The bot has its own intent (we called it NAIZ intent) to answer the user. For example, if a chatbot user says “thanks”, our dialogue system will reply “not at all”.
Depending on the characteristics of the personality, the answers can be improved with emojis, GIFs or images. A sentence answered for a NAIZ intent is classified by its personality traits and mood.
0.1 (current beta)
Basic personality traits.
Understands basic user intents.
Basic emotional management.
More personality traits.
Understands complex user intents.
Complex emotional management by trait.
Supervised learning of new expressions.
Corpus-based learning of traits.
Fine-tune of NAIZ intents.