Natural Language Understanding for Unsupported Functionality Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing natural language processing systems struggle to identify and support novel combinations of domains, intents, and entities that users frequently request but are not recognized, leading to unsupported functionalities and degraded user experiences.
Innovation Solution
A semi-supervised technique involving functionality clustering and novelty detection using a Bidirectional Encoder Representations from Transformers (BERT) model with multi-stage fine-tuning to generate functionality-based representations, followed by a classifier to label clusters as supported or unsupported.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional speech recognition systems are used, then basic speech-to-text conversion is achieved, but novel combinations of domains, intents, and entities are not recognized leading to unsupported functionalities
Solution Approach 1:
The system performs preliminary clustering of user inputs by functionality before classification. By grouping similar user inputs together and analyzing their characteristics in advance, the system can identify patterns of unsupported functionalities more reliably, rather than attempting to classify each input individually without contextual understanding.
Solution Approach 2:
The patent introduces an intermediary classification layer between speech recognition and functionality execution. This classifier acts as a mediator that determines whether a recognized speech input corresponds to a supported or unsupported functionality, providing a buffer that improves reliability by explicitly identifying novel combinations rather than attempting direct execution.
2Adaptability or versatility
If the system attempts to recognize all possible user requests, then coverage of supported functionalities increases, but system complexity increases making it difficult to identify novel combinations
Solution Approach 1:
The system segments the functionality identification process into distinct stages: clustering user inputs by functionality, extracting characteristics from clusters, and classifying supported versus unsupported functionalities. This segmentation reduces complexity by breaking down the monolithic task of recognizing all possible user requests into manageable, sequential steps.
Solution Approach 2:
Rather than attempting to pre-program recognition for all possible functionality combinations, the system takes a partial action approach by focusing on clustering and identifying patterns in actual user inputs. This allows the system to adaptively recognize novel combinations without requiring exhaustive pre-definition of every possible functionality.
3Speed
If the system processes each user input individually, then processing speed is maintained, but the ability to identify trending functionalities and clusters is reduced
Solution Approach 1:
The system performs preliminary clustering of user inputs before individual processing. By grouping similar inputs together in advance, the system enables faster identification of trending functionalities and patterns, as the clustering structure allows for efficient aggregation and analysis of usage statistics across multiple inputs.
Solution Approach 2:
The patent merges multiple individual user inputs into functionality clusters based on similarity. This combining approach maintains processing efficiency by handling clusters as unified entities while still enabling the identification of trending functionalities through aggregate analysis of cluster characteristics and frequencies.
Data Source
AI summary
Techniques are described for identifying functionalities (i.e., user experiences) that are requested by users but are not supported by natural understanding (NU) processing. Some embodiments may involve identifying functionalities by transforming user inputs to functionality-based representations. The functionality-based representations may be grouped into individual functionalities. The user inputs associated with an individual functionality may be evaluated using an NU component to determine whether the functionality is supported. These techniques may enable discovery at a functionality level, rather than at a user input level, an intent level, or an entity level. These techniques may also be used to group user inputs to determine trending functionalities.