There are five main game types to serve as templates for building dialogue. While the generic game is a general-purpose behavior for conditional dialogue logic, the other game-types each combine unique dialogue behaviors to address specific tasks.
This type of dialog behavior allows a variety of contextualized responses, where a defined trigger intent is directly linked to a response of the bot. The specific reaction of the bot can depend on knowledge about the previous conversation, about the user, as well as about underlying data.
The form-filling game is the conversational equivalent to forms on websites. It is used to collect information necessary to query a provided API. While search behavior is explorative and driven by both the user’s (often underspecified) wishes and the bot’s strategy to narrow down the search space to the most relevant matches, form filling is mainly driven by the input required by an API, which the bot needs to elicit. While it’s usually the user that guides a search, it’s usually the bot that guides a form filling.
This behavior is used to explore a dataset, aiming to find a specific data object or a selection of relevant data objects in dialog with the user. Data objects can be products such as cars or financial products, but also content such as recipes, movies, articles or any other structured information. The bot's search behavior is dynamic: the bot can decide to narrow down or widen the search space, depending on the underlying data, and will flexibly adapt the search query in order to match the known constraints and preferences of the user. In doing so, it keeps track of which query parameters have been expressed over the course of a search, and which ones are currently most relevant.
Content objects, as identified by the dynamic search dialog mentioned above, often have a variety of properties. This dialog behavior is used to answer questions about these properties. It leverages reference resolution techniques in order to handle referring expressions (like “the first one”), elliptical utterances (like “And this?”) and other types of context-dependence and vagueness that commonly occur when users inquire about properties of objects.
To provide customer care and help users resolve specific problems beyond simple FAQs it is important to understand the problem, find out which possible solutions to the problem are applicable and then guide the user through a step by step solution path. This is exactly what the solution game does. It gives a structured approach to manage the solution logic and all conversational moves from qualification to resolution.