Voice Bot Development Platform Using ML Training Instances
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice bot technologies are limited by pre-defined intent schemas, requiring extensive computational resources for manual definition and storage, and are not scalable to learn the nuances of human speech, as they need a large quantity of intent schemas to be defined and stored.
Innovation Solution
A voice bot development platform that trains voice bots using machine learning (ML) layers based on a plurality of training instances, including user input and feature emphasis inputs, allowing the voice bot to conduct conversations for telephone calls associated with a third-party, using protocols like VoIP and PSTN, and enabling efficient training and deployment without the need for extensive intent schema definition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pre-defined intent schemas are used to enable voice bot functionality, then the bot can perform specific actions, but the bot fails when encountering intents not defined in the schemas and requires extensive computational resources for manual definition and storage
Solution Approach 1:
The patent replaces the manual mechanical process of defining and maintaining intent schemas with an automated machine learning system. The voice bot uses trained ML models to automatically understand and respond to user intents without requiring manual schema definition, thereby substituting the mechanical intent schema system with an automated learning-based system.
Solution Approach 2:
The voice bot performs self-learning through continuous training with conversation data. The system automatically improves its intent recognition capabilities by processing real conversation examples, eliminating the need for external manual intervention to update intent schemas. The bot serves itself by automatically adapting to new intents through the training pipeline.
2Adaptability or versatility
If a large quantity of intent schemas are defined to make bots robust to various nuances of human speech, then the bot can handle more diverse intents, but a large amount of memory is required to store and utilize the schemas
Solution Approach 1:
The patent changes the fundamental parameter of how intents are represented and stored. Instead of storing extensive textual intent schemas, the system uses compact ML model parameters and embeddings that capture intent semantics. This parameter transformation allows the bot to maintain high intent coverage while significantly reducing memory requirements through efficient vector representations.
Solution Approach 2:
The system uses embedding vectors as compact copies of intent information. Rather than storing full intent schema definitions, the ML model creates condensed vector representations that capture the essential semantic information, allowing the bot to access intent meanings with minimal memory consumption while maintaining versatility.
3Measurement precision
If extensive computational resources are allocated to manually define and refine intent schemas, then the bot can achieve better intent recognition, but the process is not practically scalable to learn the nuances of human speech
Solution Approach 1:
The patent replaces the manual mechanical process of intent schema refinement with automated machine learning training. The system uses computational resources to train ML models on conversation data, automatically learning intent patterns without human intervention. This substitution maintains high recognition accuracy while dramatically improving productivity through automation.
Solution Approach 2:
The system enables continuous learning through an automated training pipeline that can process conversation data and update the voice bot model continuously. This continuous action allows the bot to progressively improve its intent recognition accuracy over time without interruption, maintaining high productivity while achieving precision through ongoing training iterations.
Data Source
AI summary
Implementations are directed to providing a voice bot development platform that enables a third-party developer to train a voice bot based on training instance(s). The training instance(s) can each include training input and training output. The training input can include a portion of a corresponding conversation and a prior context of the corresponding conversation. The training output can include a corresponding ground truth response to the portion of the corresponding conversation. Subsequent to training, the voice bot can be deployed for conducting conversations on behalf of a third-party. In some implementations, the voice bot is further trained based on a corresponding feature emphasis input that attentions the voice bot to a particular feature of the portion of the corresponding conversation. In some additional or alternative implementations, the voice bot is further trained to interact with third-party system(s) via remote procedure calls (RPCs).