External Annotation Services for Speech Parsing Rule Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems face challenges in efficiently processing and parsing speech commands due to the large number of entity instances, such as person and business names, which require frequent updates and maintenance of parsing rules.
Innovation Solution
The system utilizes external annotation services to identify n-grams in speech commands, allowing for the formation of compact parsing rules without enumerating all instances of non-terminals, and performs a bottom-up parse to determine actions based on successful grammars.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If parsing rules include an enumerated list of all instances of non-terminals (e.g., all person names, business names), then the parser can accurately identify entities, but the rule sets become extremely large and difficult to maintain
Solution Approach 1:
The patent extracts the entity identification function from the parsing rules and relocates it to external annotation services. Instead of maintaining comprehensive lists of entities within the parser, the system uses external services (such as named entity recognition services) to provide annotations for entities in the input sentence. This separation reduces the complexity of parsing rules while maintaining accurate entity identification.
Solution Approach 2:
The patent introduces external annotation services as intermediaries between the speech recognition system and the parsing rules. These services act as mediators that provide entity information without requiring the parsing rules to contain exhaustive lists of entities. The annotation services receive the input sentence, identify entities, and return annotations that can be used by the parser without expanding the rule set complexity.
2Adaptability or versatility
If parsing rules are updated to include new entity instances, then the system can recognize new entities, but the maintenance and updating of rule sets becomes frequent and cumbersome
Solution Approach 1:
The patent extracts the entity identification functionality from the parsing rules and places it in external annotation services. This means that when new entities need to be recognized, the parsing rules do not need to be updated or recompiled. Instead, the external services provide annotations for new entities dynamically, allowing the system to adapt to new entities without maintenance of the parsing rule sets.
Solution Approach 2:
The patent performs entity identification in advance through external annotation services before the parsing process. By pre-annotating entities in the input sentence using external services, the system prepares entity information beforehand, eliminating the need to update parsing rules when new entities emerge. This preliminary action separates entity recognition from parsing rule maintenance.
3Device complexity
If the system uses external annotation services to identify entities, then the parsing rules become compact and easier to maintain, but the system architecture becomes more complex with additional external dependencies
Solution Approach 1:
The patent extracts the complex entity identification logic from the parsing system and places it in external specialized services. This extraction simplifies the parsing rules and parser architecture, as they no longer need to contain comprehensive entity lists or complex matching logic. The trade-off is the introduction of external service dependencies, but this is accepted as a manageable architectural complexity compared to maintaining massive parsing rule sets.
Data Source
AI summary
A language processing system uses annotation services that are external to the language processing system to identify n-grams that identify entities in an input sentence. The n-grams are annotated by the annotation services. The annotations are used to determine which n-grams, if any, correspond to instances of an entity type (e.g., values for a variable or terminals for a non-terminal). After determining which n-grams correspond to entity types, parse initializations are generated for parsing rules and parses for each rule are attempted. The rules that successfully parse are used to determine whether the input sentence invokes a specific action, and if so, what arguments are to be passed to the invocation of the action.