Neural Network Word Score Calculation for New Words
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional word score calculation devices cannot calculate word scores for strings that include new words not present in their lexicon, limiting their effectiveness in voice recognition systems.
Innovation Solution
A word score calculation device that uses a neural network to calculate word scores by inputting lexicon features of known and new words, with the ability to identify part-of-speech features for new words and calculate scores using conditional probabilities and softmax functions, allowing for the inclusion of new words in the scoring process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a conventional word score calculation device holds a list of words called a lexicon, then it can calculate word scores for known words, but it cannot calculate word scores for new words not included in the lexicon
Solution Approach 1:
The patent transforms the lexicon from a static list of words to a dynamic feature-based representation. Instead of relying on exact word matches in a predefined lexicon, the system extracts features (such as character n-grams, part-of-speech tags, and contextual information) from words and uses these features as inputs to a neural network. This parameter transformation enables the system to handle both known and new words by comparing their feature representations rather than requiring exact lexical matches.
Solution Approach 2:
The patent introduces a neural network as an intermediary between the input words and the word score calculation. The neural network acts as a mediator that processes feature representations of words (including new words not in the lexicon) and outputs probability scores. This intermediary component bridges the gap between the traditional lexicon-based approach and the need to handle unseen words, enabling the system to generalize to new vocabulary while maintaining reliable scoring.
2Reliability
If a voice recognition system uses a fixed lexicon for word score calculation, then it maintains consistency and reliability, but it cannot adapt to new vocabulary or evolving language
Solution Approach 1:
The patent makes the word scoring system dynamic by using a neural network that can process and adapt to new feature patterns. Unlike a static lexicon that requires manual updates, the neural network dynamically learns from training data and can generalize to new words based on their feature representations. This dynamic approach allows the system to maintain reliability through learned patterns while simultaneously adapting to new vocabulary and language evolution.
Solution Approach 2:
The patent creates a universal word scoring mechanism that handles both known and new words through a single neural network model. The system extracts features that are universally applicable to any word, regardless of whether it appears in the training data. This multi-functional approach allows the same system to reliably score established vocabulary while also adapting to new words, eliminating the need for separate handling mechanisms.
Data Source
AI summary
A word score calculation device according to an embodiment includes an input unit and a calculating unit. The input unit receives input of a string of words including known words, which are known already, and new words, which are not yet known. The calculating unit inputs the lexicon feature of a word, which is included in the string of words, in a neural network in which one or more unit groups for inputting the lexicon features of words are included in an input layer, and calculates a word score of the target word included in the string of words.