Chatbot Intent Prediction Engine for Transaction Progress
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing chatbot applications often fail to facilitate natural and efficient user-driven conversations that lead to fulfilling transactions, as they tend to be complex and limited, resembling choose-your-own-adventure narratives rather than dynamic dialogues.
Innovation Solution
A computer-system implemented method for a chatbot application that uses a trained intent classifier and next intent predictor to generate suggested questions based on user inputs, with a re-ranker scoring intents relative to transaction progress, enabling the chatbot to provide productive next questions to users during live chat conversations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing chatbot applications use complex predetermined conversation flows, then transaction completion can be achieved, but the conversation becomes limited and unnatural like choose-your-own-adventure narratives
Solution Approach 1:
The chatbot dynamically generates suggested next questions based on real-time classification of user intent and prediction of user needs, rather than following fixed predetermined paths. The system adapts the conversation flow dynamically by analyzing the current state and generating context-relevant suggestions, making the interaction feel more natural while maintaining transaction progress.
Solution Approach 2:
The chatbot autonomously generates and presents suggested next questions to users without requiring users to navigate complex menus or follow rigid scripts. The system serves itself by automatically analyzing conversation state, predicting user intent, and generating appropriate follow-up questions, thereby simplifying the user experience while driving transaction completion.
2Productivity
If the chatbot provides multiple suggested questions to users, then conversation efficiency improves, but the system complexity increases due to multiple AI components
Solution Approach 1:
The system is divided into distinct functional modules: an intent classifier that analyzes user input, a next-question predictor that generates candidate questions, a re-ranker that prioritizes suggestions based on transaction progress, and a question selection engine that presents final suggestions. This segmentation allows each component to specialize in a specific task, improving overall efficiency while managing complexity through modular design.
Solution Approach 2:
The re-ranker acts as an intermediary between the predictor and the user, filtering and prioritizing predicted questions based on their relevance to transaction progress. This intermediary layer manages the flow of information between complex AI components and the user interface, presenting only the most relevant suggestions while hiding the underlying system complexity.
3Productivity
If the chatbot suggests questions that advance transaction stages, then transaction progress improves, but the system must accurately predict user intent which increases computational complexity
Solution Approach 1:
The system pre-trains the intent classifier and question predictor on historical conversation data before deployment. This preliminary training allows the models to learn patterns of user intent and effective question sequences in advance, reducing the computational complexity required during real-time interactions while maintaining accurate transaction progress prediction.
Solution Approach 2:
The re-ranker uses feedback from transaction stage analysis to prioritize suggested questions. By continuously monitoring which transaction stage the conversation is in and comparing it against the predicted impact of suggested questions, the system can efficiently select questions that advance transaction progress without requiring overly complex real-time analysis of all possible user responses.
Data Source
AI summary
When a chatbot application is in a mode of live chat conversation with a user, a trained intent classifier determines an intent that underlies a first live question received by the trained intent classifier from the user. A trained next predictor receives the intent from the intent classifier. The trained next intent predictor generates a set of predicted next intents responsive to receiving the intent. A trained re-ranker selects at least one of the predicted next intents responsive to receiving the set of predicted next intents. A question selection engine sends at least one suggested question to the user responsive to receiving the at least one predicted next intent. As a result of the above, the chatbot application provides to the user at least one suggested next question the user may wish to ask in response to the first question from the user.