Empower your virtual agents to have complex and meaningful conversations with users at scale – without lengthy development cycles. Kore.ai accelerates the bot building process by enabling you to perform intent discovery on transcripts to create dialogs that match the flow of natural, human conversations.
“Gartner predicts that 25% of customer service operations will use virtual customer assistants by 2020”
Virtual agents are becoming critical in transforming customer service to meet new demands for speed, accuracy, and choice. While enterprises are beginning to realize the importance of chatbots, they also understand that time is one of their most valuable assets. The complexities inherent in building human-like dialog flows and the need for development speed will only increase exponentially as the underlying technology becomes more sophisticated. Traditional methods and strategies for designing, training, and deploying chatbots will no longer be sufficient – precisely because they take too long.
Kore.ai can automatically identify intents and discourse patterns from chat transcripts to generate human-like dialogs for your virtual agents - in a fraction of the time it used to take.
Traditional Bot Building Process vs. Kore.ai’s Auto Dialog Generation
Building a successful chatbot can be challenging, but it doesn’t have to be. Enterprises who use the traditional model must generally follow a long and complicated series of steps, including developing use case blueprints, designing conversational experiences, training NLP models, and deployment. This process can take months of hard work.
Kore.ai’s auto dialog generation system, which uses transcripts from past chat and voice conversations to uncover customer interaction patterns, can now be leveraged to model chatbots. This approach offers enterprises significant advantages compared to the traditional method.
Auto Dialog Generation in Action
The Kore.ai Bots Platform follows a step-by-step approach to accurately extract dialog and train your virtual agents from chat transcripts.
Identify relevant intents
Identifying relevant intents is the first step in the dialog generation process. Conversation transcripts can be complex with primary intents, supplemental intents occuring in the same conversation. Kore.ai normalizes the utterances, extracts relevant utterances from the conversations and discover a set of special features that are used to cluster semantically similar utterances. Unsupervised machine learning algorithms and models are built which helps surface groups which have not been explicitly labeled in the data.
Discover intent discourse patterns
Identifying discourse patterns for each intent is the second step in the dialog generation process. A conversation typically will have a collection of sub-tasks which can be categorized as inputs/responses required to fulfill the task. Kore.ai extracts various sub-tasks and valid ordering of these in an unsupervised manner that captures the variability across conversations.
Generate bot training data
Generate training data set to recognize intents, supplemental intents, and also populate bot response messages for entity prompts, confirmations and other messages.
Model bot dialog tasks
The final step in the auto dialog generation process involves extracting and assigning entities to each identified intent. A dialog is generated after sequentially arranging all the combined intent and entity groups, and is based on their occurrence. The bot model is then trained with the new dialogs before being imported into the Kore.ai dialog builder.
Import into Kore.ai bot builder
Bot modeled for conversation flow and along with training utterances can be imported into Kore.ai bot builder. From here, developers can visualize and review conversation flows and branches using the visual dialog editor. This provides a solid foundation from which to build – or expand on – your virtual agent’s conversational capabilities.