Web Session Summarization Using LLM Event Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Website providers lack a clear understanding of how users interact with web documents due to varying network conditions, making it difficult to accurately capture, analyze, and present user interactions effectively.
Innovation Solution
A system utilizing a large language model (LLM) to analyze captured user interactions, generating summaries of web sessions that describe user intent, encountered difficulties, and outcomes by obtaining interaction data, identifying events, and providing textual descriptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional network monitoring methods are used to capture user interactions, then data can be collected, but the accuracy and timeliness of analysis is insufficient due to varying network conditions and large permutations of user interactions
Solution Approach 1:
The patent introduces an intermediary system comprising a capture agent on the user device, an event processing server, and a language model. The capture agent collects raw interaction data locally, the event processing server processes and structures the data, and the language model generates meaningful summaries. This intermediary architecture bridges the gap between raw data collection and accurate analysis, resolving the contradiction by adding structured processing layers that improve measurement precision without overwhelming system complexity.
Solution Approach 2:
The system segments the complex task of user interaction analysis into distinct components: (1) capture agent for data collection, (2) event processing for data structuring and filtering, (3) language model for semantic interpretation. By dividing the analysis process into manageable segments, the system achieves high measurement precision while keeping each component's complexity controlled and manageable.
2Loss of information
If comprehensive user interaction data is captured across all devices and network conditions, then user experience understanding improves, but the time and resources required for processing increases
Solution Approach 1:
The event processing component performs preliminary actions by pre-processing, structuring, and filtering interaction data before it reaches the language model. Events are categorized, timestamps are recorded, and data is organized into a standardized format in advance. This preliminary processing reduces the computational burden during final analysis, maintaining complete information while reducing processing time.
Solution Approach 2:
The patent replaces traditional mechanical data processing methods with a language model that uses natural language understanding. Instead of complex algorithms manually processing each interaction permutation, the language model semantically interprets the structured event data, dramatically reducing processing time while maintaining information completeness.
3Measurement precision
If detailed analysis of each user interaction permutation is performed, then accurate insights are obtained, but the processing complexity and computational resources increase significantly
Solution Approach 1:
The system extracts only the most relevant features from comprehensive interaction data through the event processing layer. Instead of analyzing every raw interaction permutation, the system extracts key events, user actions, and outcome metrics that are most indicative of user experience quality. This extraction maintains measurement precision by focusing on critical insights while dramatically improving processing efficiency by eliminating redundant analysis.
4Speed
If real-time analysis of user interactions is implemented, then timely insights are provided, but the computational load and system complexity increase
Solution Approach 1:
The system performs preliminary structuring and organization of interaction data as it is collected, creating a ready-to-analyze format before real-time processing is needed. Event timestamps are recorded, data is categorized, and structures are prepared in advance, enabling rapid real-time analysis without the full computational burden of processing raw data from scratch.
Solution Approach 2:
The language model replaces complex mechanical processing systems with a more efficient natural language processing approach. The model can rapidly interpret structured event data and generate insights in real-time, reducing both computational load and system complexity compared to traditional real-time analysis methods.
Data Source
AI summary
The method includes: obtaining, from a capture agent on a user device, captured data for a set of captured user interactions with the network site during the network session, wherein the set of captured user interactions includes movements between portions of the network site; analyzing the captured data to identify a set of events; extracting event data corresponding to the set of events; receiving a request for a textual description of the network session; generating a prompt for a language model, using the request and the event data; providing the prompt as an input to the language model; and receiving the textual description from the language model. The textual description is provided to a computer associated with the network site.