Phonetic Sequence Similarity via HMM State Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques fail to derive a phonetic symbol sequence similar to a given phonetic symbol sequence, limiting their effectiveness in speech recognition, word spelling, and document retrieval applications.
Innovation Solution
An information processing device comprising a phonetic converting unit, an HMM converting unit, and a searching unit that converts a phonetic symbol sequence into a hidden Markov model (HMM) state sequence, then uses a similarity matrix to generate a score vector sequence, allowing the device to search for and output a phonetic symbol sequence similar to the input sequence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional techniques are used to calculate similarity between two phonetic symbol sequences, then the similarity between two given sequences can be derived, but the ability to generate a phonetic symbol sequence similar to a given sequence is lost
Solution Approach 1:
The patent introduces a similarity matrix as an intermediary component that stores pre-calculated similarity values between phonetic symbols. This matrix serves as a mediator between the input phonetic sequence and the generated output sequence, enabling efficient similarity-based generation without requiring complex real-time calculations. The similarity matrix allows the system to both measure similarity and generate similar sequences by looking up pre-computed values.
Solution Approach 2:
The patent performs preliminary action by pre-calculating and storing similarity values between all pairs of phonetic symbols in a similarity matrix before the actual sequence generation task. This preliminary computation enables the system to quickly generate similar phonetic sequences by simply querying the pre-computed matrix, rather than performing complex similarity calculations during the generation process.
2Measurement precision
If phonetic symbol sequences are converted to HMM state sequences using alignment, then the conversion can be performed accurately, but the complexity of the conversion process increases
Solution Approach 1:
The patent uses copying by creating a simplified representation of the phonetic sequence in terms of HMM states through alignment. Instead of performing complex transformations, the system copies the essential information from the phonetic sequence into a standardized HMM state sequence format, preserving the semantic meaning while simplifying the representation for downstream processing.
Solution Approach 2:
The patent applies segmentation by dividing the phonetic symbol sequence into discrete HMM states through alignment. The continuous phonetic sequence is segmented into discrete states that can be independently processed and compared using the similarity matrix, making the conversion process more manageable and computationally efficient.
3Adaptability or versatility
If a similarity matrix is introduced to enable phonetic sequence generation, then the ability to derive similar sequences is improved, but the device complexity increases
Solution Approach 1:
The similarity matrix serves as a simple intermediary structure that enables sequence generation capability without requiring complex algorithms. By storing pre-computed similarity values in a matrix format, the system can generate similar phonetic sequences through straightforward matrix queries and lookups, avoiding the need for complex real-time similarity calculations while maintaining high generation accuracy.
Data Source
AI summary
An information processing device includes a phonetic converting unit, an HMM converting unit, and a searching unit. The phonetic converting unit converts a phonetic symbol sequence into a hidden Markov model (HMM) state sequence in which states of an HMM are aligned. The HMM converting unit converts the HMM state sequence into a score vector sequence indicating the degree of similarity to a specific pronunciation using a similarity matrix defining the similarity between the states of the HMM. The searching unit searches for a path having a better score for the score vector sequence than that of the other paths out of paths included in a search network and outputs a phonetic symbol sequence corresponding to the retrieved path.