Chatbot Intent Resolution via Negation and Coreference Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional chatbots struggle with accurate intent detection in user utterances that include negation or coreferences, often misinterpreting negations and pronouns, leading to incorrect task execution.
Innovation Solution
The online system employs an intent detection model that modifies utterances by removing negations and replacing pronouns, generating confidence scores to determine the intent, and uses a multi-tenant system architecture to customize chatbot interactions, ensuring accurate task execution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional intent detection models are used, then the system can process user utterances quickly, but the accuracy deteriorates for utterances containing negation or coreferences
Solution Approach 1:
The system performs preliminary processing of the utterance by identifying and removing negations and resolving coreferences before feeding the modified utterance to the intent detection model. This preliminary action prepares the utterance in a form that is easier and more accurately processed by the model, thereby improving intent detection accuracy without requiring the model itself to be more complex.
Solution Approach 2:
The utterance processing is segmented into distinct stages: (1) identifying negations and coreferences, (2) modifying the utterance by removing negations and replacing coreferences, and (3) detecting intent using the modified utterance. This segmentation allows each stage to be optimized independently, with the modification stage simplifying the input for the intent detection model.
2Measurement precision
If the system processes negations and coreferences correctly, then intent detection accuracy improves, but the processing time increases
Solution Approach 1:
The system uses the same intent detection model that processes simple utterances to also process modified utterances that have had negations and coreferences handled. The model serves itself by processing both types of utterances through the same infrastructure, avoiding the need for separate complex processing pipelines and minimizing additional time overhead.
Solution Approach 2:
The system changes the parameters of the utterance (removing negations, replacing coreferences) to transform it into a form that is faster and more accurate to process. By modifying the utterance structure rather than changing the model architecture, the system achieves improved performance without significant time cost.
3Reliability
If the chatbot performs tasks based on potentially incorrect intent detection, then task execution speed is maintained, but task correctness deteriorates
Solution Approach 1:
The system performs preliminary modification of the utterance to ensure correct intent detection before executing any tasks. By preparing the utterance in advance (removing negations, resolving coreferences), the system ensures that the subsequent task execution is both correct and efficient, as the intent detection model can quickly and accurately process the modified utterance.
Data Source
AI summary
A system performs conversations with users using chatbots customized for performing a set of tasks. The system may be a multi-tenant system that allows customization of the chatbots for each tenant. The system processes sentences that may include negation or coreferences. The system determines a confidence score for an input sentence using an intent detection model, for example, a neural network. The system modifies the sentence to generate a modified sentence, for example, by removing a negation or by replacing a pronoun with an entity. The system generates a confidence score for the modified sentence using the intent detection model. The system determines the intent of the sentence based on the confidence scores of the sentence and the modified sentence. The system performs tasks based on the determined intent and performs conversations with users based on the tasks.