Granularizing Compound Natural Language Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Natural language queries that include multiple requests for information within a single query are difficult to interpret, as existing technologies treat each word as text input, leading to irrelevant results, and simple text searches fail to extract the desired information effectively.
Innovation Solution
A method for processing natural language queries involves parsing the query to create an argument tree, allowing for foldable, splittable, and sequence-based nodes, with validity rules to derive granular query components that accurately represent the query intent, enabling the extraction of multiple pieces of information from a single query.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If basic search engines treat each word as text input to find results including every word, then the search process is simple, but the results are not relevant and do not provide the actual information desired
Solution Approach 1:
The patent segments a complex natural language query into multiple granular query components by parsing the query structure. For example, the query 'what was the population of Japan in 1900 and 2000?' is divided into separate components: population of Japan in 1900, and population of Japan in 2000. This segmentation allows the system to process each component individually and return precise information rather than irrelevant sources.
2Device complexity
If simple text searches eliminate common words like 'a' or 'the', then the search processing is simplified, but the results rarely change and remain ineffective
Solution Approach 1:
The patent performs preliminary parsing of the natural language query to identify its structural components before executing the search. The system analyzes the query to determine the subject, attributes, qualifiers, and relationships between elements. This preliminary action enables the system to construct meaningful search components that preserve the original query's intent, rather than simply eliminating common words.
3Adaptability or versatility
If rudimentary systems attempt to answer simple phrases, then they can handle basic queries, but they cannot deal with queries requiring multiple answers or complex information extraction
Solution Approach 1:
The patent adds a structural dimension to query processing by introducing an argument tree representation. This tree structure organizes query components hierarchically, showing relationships between subjects, attributes, and qualifiers. By transforming the flat text query into a multi-dimensional tree structure, the system can systematically process complex queries with multiple answers while maintaining the adaptability to handle simple queries through the same framework.
Data Source
AI summary
A method for processing a natural language query. The method includes receiving a text query, the query referring to a plurality of objects, attributes, qualifiers and other arguments and parsing the query to produce an argument tree representing the substance and structure of the query. The method also includes the capability to define qualifiers as being possibly projectable onto other arguments and indicate their direction of projectability and the capability to denote nodes of the argument tree as foldable, as splittable, or as containing sequences of qualifier arguments. The method additionally includes defining validity rules for a domain of knowledge, used to determine whether a list of arguments form a valid granular query component and processing of the argument tree, in view of the above in order to derive a corresponding plurality of granular query components that collectively request the plurality of pieces of information representing the intent of the query.