Relational Language Segmentation for Intent Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Intelligent Virtual Assistants (IVAs) face challenges in understanding user intentions due to the use of relational strategies, which can lead to confusion and degradation of user experience as unnecessary background information obfuscates primary intent.
Innovation Solution
A system and method that determine secondary intents within user input by classifying relational language using audio, emojis, and sentiment shifts, allowing for the generation of responses based on the primary intent while ignoring non-essential information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If relational strategies (self-exposure, justification, small talk) are used by users to build trust with IVAs, then human-like interaction and trust are improved, but user intent becomes obfuscated and confusion increases
Solution Approach 1:
The patent segments user input into different relational strategies (small talk, self-exposure, justification) and task-oriented content. By dividing the input stream into distinct segments and classifying each segment's purpose, the system can identify which portions contain actionable intent and which are purely relational, enabling accurate intent recognition despite the presence of human-like interaction strategies
Solution Approach 2:
The patent extracts and isolates the core task-oriented intent from the surrounding relational language. By taking out only the essential intent information and separating it from the relational strategies (small talk, self-exposure, justification), the system can process the intent accurately without being confused by the human-like conversational elements
2Loss of information
If IVAs process all user input including relational language, then complete understanding of user context is achieved, but processing time increases and response accuracy decreases
Solution Approach 1:
The patent extracts only the essential intent information from user input and discards or sets aside the relational language portions. This extraction process allows the system to retain necessary context understanding while eliminating time-consuming processing of irrelevant relational strategies, achieving both efficient processing and accurate intent recognition
Solution Approach 2:
The patent performs preliminary classification of input segments to identify relational language before full processing. By preliminarily sorting input into relational vs. task-oriented categories, the system can quickly determine which portions require detailed processing and which can be handled more efficiently, reducing overall processing time while maintaining context understanding
3Productivity
If IVAs use computerized methodologies to generate responses, then response generation speed is improved, but user experience degrades due to clarification questions and wrong information
Solution Approach 1:
The patent incorporates feedback mechanisms that allow the IVA to detect when relational strategies are being used and adjust its response accordingly. By receiving feedback about the user's relational strategies (small talk, self-exposure, justification), the system can modify its processing approach to better handle these scenarios, reducing clarification questions and wrong information while maintaining fast response generation
Solution Approach 2:
The patent makes the response generation process dynamic by adjusting the level of processing based on the detected relational strategies. When relational language is detected, the system dynamically adapts its processing depth and response formulation, ensuring accurate responses tailored to the user's relational intentions while maintaining efficient processing speeds
Data Source
AI summary
Features, libraries, and techniques are provided herein for determining the kinds of relational language that are present. Applying audio, emojis, and sentiment shifts as features may be used to determine whether the customer is providing backstory, whether there is ranting, etc. Textual features may be considered, as well as audio features may be considered.