Natural Language Model for Predicting User Activity Events
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional recommendation systems are inadequate for content management systems due to decreased accuracy in predicting user actions and inflexibility in handling complex data representations, leading to inapplicable suggestions and inefficient resource utilization.
Innovation Solution
A natural language model-based system that analyzes historical user activity events to generate event tokens and predict next user actions by leveraging a trained model to rank candidate sequences of activity events, providing more accurate and contextually relevant recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional recommendation systems are used, then the system structure is simple, but the prediction accuracy of user actions decreases
Solution Approach 1:
The patent transforms the recommendation system into a sequence prediction model by changing the fundamental parameter of how user actions are modeled - from simple association rules to natural language sequence processing. This allows the system to capture temporal dependencies and contextual relationships in user behavior, significantly improving prediction accuracy while maintaining reasonable system complexity through the use of pre-trained language models.
Solution Approach 2:
The patent replaces traditional recommendation algorithms with a natural language processing-based sequence prediction model. By treating user action sequences as linguistic data that can be processed by language models, the system achieves higher prediction accuracy without proportionally increasing mechanical complexity, leveraging the power of pre-trained neural networks.
2Adaptability or versatility
If conventional recommendation systems are used, then the model is inflexible, but the handling of complex data representations is limited
Solution Approach 1:
The patent makes the recommendation system universal by using a natural language model that can process any sequence of user actions regardless of their type or complexity. The same model architecture handles diverse data representations (file operations, folder manipulations, sharing actions, etc.) uniformly, greatly enhancing model flexibility and adaptability to complex data without requiring separate specialized components.
3Device complexity
If a limited subset of user activity is considered, then the system complexity is low, but the accuracy of predictions decreases
Solution Approach 1:
The patent implements a dynamic approach where the system automatically adapts to consider relevant user activities based on the current context and historical patterns. The sequence prediction model dynamically weights and selects which user actions are most predictive, allowing the system to effectively consider a comprehensive set of user activities without requiring manual configuration or excessive system complexity.
4Productivity
If the system provides inapplicable suggestions, then resource utilization is inefficient, but the user experience deteriorates
Solution Approach 1:
The sequence prediction model inherently incorporates feedback mechanisms by learning from historical user behavior patterns. The model continuously refines its predictions based on observed user actions, ensuring that suggestions are contextually appropriate and applicable. This feedback-driven approach improves both resource utilization by avoiding irrelevant recommendations and user experience by providing actionable suggestions.
Data Source
AI summary
The present disclosure relates to systems, methods, and non-transitory computer-readable media that can leverage a natural language model to determine a most probable candidate sequence of tokens and thereby generate a predicted user activity. In particular, the disclosed systems can tokenize activity event vectors to generate a series of sequential tokens that correspond to recent user activity of one or more user accounts. In addition, the disclosed systems can, for each candidate (e.g., hypothetical) user activity, augment the series of sequential tokens to include a corresponding token. Based on respective probability scores for each of the augmented series of sequential tokens, the disclosed systems can identify as the predicted user activity, a candidate user activity corresponding to one of the augmented series of sequential tokens associated with a highest probability score. Based on the predicted user activity, the disclosed systems can surface one or more suggestions to a client device.