Finite State Transducer Rule Code Embedding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional finite state transducers (FSTs) in automatic speech recognition (ASR) processing lose hierarchical designations such as rules and sub-rules due to their 'flat' nature, leading to incomplete output strings that fail to capture the application of rules during speech processing.
Innovation Solution
An enhanced FST is constructed by encoding rule codes along its arcs, ensuring that these codes are included in the output string during ASR processing, thereby preserving and utilizing hierarchical information for improved ASR and post-ASR functionality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional finite state transducers are used in ASR processing, then the processing speed and simplicity are improved, but hierarchical designations such as rules and sub-rules are lost due to the flat nature of FST
Solution Approach 1:
The patent embeds hierarchical rule information within the flat FST structure by nesting rule codes and hierarchical designations inside the state transitions and arcs of the FST. This allows the FST to maintain its processing efficiency while containing nested hierarchical information that can be extracted when needed.
Solution Approach 2:
The patent introduces an intermediary layer that maps between the flat FST representation and the hierarchical rule structure. This intermediary mechanism allows hierarchical designations to be preserved and retrieved without disrupting the efficient flat processing of the FST.
2Device complexity
If traditional finite state transducers are used, then device complexity is reduced, but the output string fails to capture rule application information
Solution Approach 1:
The patent merges the rule application information directly into the FST output by combining the speech recognition results with rule codes and hierarchical designations in a single integrated output string. This eliminates the need for separate rule tracking mechanisms.
Solution Approach 2:
The patent performs preliminary encoding of rule codes and hierarchical designations into the FST structure during the recognition process, so that the complete information is already embedded in the output string before any post-processing occurs.
3Loss of information
If rule codes are encoded along FST arcs, then hierarchical information is preserved in the output string, but device complexity increases
Solution Approach 1:
The patent makes the FST structure multi-functional by enabling it to simultaneously perform speech recognition and hierarchical rule encoding. The same FST arcs that recognize speech patterns also carry rule codes, eliminating the need for separate rule encoding mechanisms.
Solution Approach 2:
The patent changes the parameters of the FST arcs to include additional rule code information alongside the traditional speech recognition data. This parameter expansion allows hierarchical information to be carried without requiring a fundamentally different structure.
Data Source
AI summary
In automatic speech recognition, certain parsing information, such as rules and tags, may be embedded into a finite state transducer (FST) to produce FST output that includes speech recognition results along with codes indicating parsing results of the recognized speech. The codes in the FST output may be formatted using a markup language, such as XML or JSON, for processing by a later application. The FST may be constructed according to a grammar defining the parsing information. The codes for inclusion in the FST output may be embedded into arcs of the FST and then included in the FST output when the speech recognition engine traverses the arcs of the FST.