LLM Query Answering for Crowd-Sourced Comment Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional text search techniques for processing large volumes of unstructured data, such as user feedback, are inefficient and require significant manual effort due to numerous documents matching a search request.
Innovation Solution
A system utilizing a machine learning based language model processes crowd-sourced information from multiple client devices, generating aggregate representations and responses to user queries by incorporating user feedback into prompts for the model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional text search techniques are used to process large volumes of unstructured data, then the system can search through documents, but the efficiency deteriorates and manual effort increases significantly
Solution Approach 1:
The patent replaces conventional mechanical text search techniques with machine learning-based language models that automatically process, understand, and summarize unstructured data. The system uses LLMs to generate insights and answers directly from crowd-sourced information, eliminating the need for manual document review and significantly improving search efficiency while reducing time loss.
2Productivity
If machine learning based language models are used to process crowd-sourced information, then query answering efficiency improves, but system complexity increases
Solution Approach 1:
The patent introduces an intermediary layer that bridges users and raw crowd-sourced data through machine learning models. The system generates aggregated representations and insights as intermediaries, translating complex unstructured data into actionable answers. This intermediary approach simplifies the user interface while leveraging the computational power of LLMs in the background.
Solution Approach 2:
The system performs preliminary processing of crowd-sourced information by pre-computing aggregate representations and storing them in an index. This preliminary action enables faster query responses by avoiding the need to process all raw data from scratch, thus improving query answering efficiency while managing system complexity through proactive data preparation.
3Loss of information
If conventional text search is used for large corpora of small documents, then the system can retrieve matching documents, but the quality of information processing deteriorates due to manual review requirements
Solution Approach 1:
The patent replaces manual document review with automated machine learning-based information processing. The system uses LLMs to understand, summarize, and extract insights from large corpora of small documents automatically, significantly improving information processing quality while eliminating the burden of manual review effort.
Solution Approach 2:
The system merges multiple small documents into aggregated representations through machine learning models. By combining and synthesizing information from numerous individual documents, the system creates comprehensive insights that maintain high information processing quality while reducing the operational effort required compared to manually reviewing each document separately.
Data Source
AI summary
A system, for example, an online system uses a machine learning based language model, for example, a large language model (LLM) to process crowd-sourced information provided by users. The crowd-sourced information may include comments from users represented as unstructured text. The system further receives queries from users and answers the queries based on the crowd-sourced information collected by the system. The system generates a prompt for input to a machine-learned language model based on the query. The system provides the prompt to the machine-learned language model for execution and receives a response from the machine-learned language model. The response comprises the insight on the topic and evidence for the insight. The evidence identifies one or more comments used to obtain the insight.