Intent-Based NLP System for Ambiguous Natural Language Requests
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Natural language data applications face challenges in efficiently processing user requests due to ambiguity, requiring time-consuming and repetitive interrogation processes to determine user intent, which degrades the user experience and increases effort in retrieving and analyzing data from diverse data sources with different domain-specific languages.
Innovation Solution
A computer-implemented method that receives a natural-language request, selects an intent from a pre-defined intent template, determines property field values, generates a query for a field-searchable data source, and displays results, including visualizations filtered by specific criteria, to efficiently retrieve and analyze data across various data sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If natural language data applications use traditional interrogation processes to determine user intent, then they can handle ambiguous NL requests, but the process becomes time-consuming and repetitive, degrading user experience
Solution Approach 1:
The system performs preliminary actions by pre-defining intent templates and property fields before receiving natural language requests. The intent template includes predefined property fields that guide the extraction process, allowing the system to proactively structure and categorize user intents without requiring time-consuming interactive clarification
Solution Approach 2:
The patent introduces an intermediary layer consisting of intent templates and property fields that mediate between the ambiguous natural language input and the structured query output. This intermediary structure captures user intent by mapping NL requests to predefined property fields, reducing the need for direct user-system clarification exchanges
2Adaptability or versatility
If NL data applications extract and curate metadata from diverse data sources, then they can translate NL requests to appropriate DSL requests, but the complexity of handling multiple domain-specific languages increases
Solution Approach 1:
The system implements a universal intent template structure that can accommodate multiple domain-specific languages and data sources. The property fields in the intent template are designed to be language-agnostic, allowing the same template framework to handle SQL, SPL, and other DSLs through a single unified interface
Solution Approach 2:
The patent uses an intermediary metadata layer that sits between diverse data sources and the natural language interface. This metadata layer standardizes the representation of data from different DSLs, allowing the system to translate NL requests to appropriate DSL queries without directly managing the complexity of each individual language
3Productivity
If the system uses pre-defined intent templates with property fields, then it can reduce user effort and time, but the system must accurately determine property field values from ambiguous NL requests
Solution Approach 1:
The system employs feedback mechanisms where the structured intent template and property fields provide constraints and guidance back to the natural language processing stage. This feedback loop helps disambiguate NL requests by comparing extracted values against expected property field types and formats, improving the accuracy of property field value determination
Data Source
AI summary
Various embodiments of the present application set forth a computer-implemented method that includes receiving, from a device, a natural-language (NL) request. The method further includes selecting, using the NL request, an intent from a set of intents, wherein the intent is associated with a pre-defined intent template, the pre-defined intent template including a set of property fields that are associated with one or more portions of the NL request. The method also includes determining, based on the NL request, a set of property field values for the set of property fields. The method further includes generating a query to be executed on a field-searchable data source, wherein the query is based on one or more property field values included in the set of property field values. The method also includes receiving, in response to the query, a result that includes a set of event field values. In addition, the method includes causing the device to display at least a portion of the result.