Meaning Generation Using Restatement Model for Word Importance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional meaning generation techniques struggle with accurately identifying important words in sentences due to the misclassification of unimportant words that occur frequently, leading to incorrect conversion of utterances to meaning labels.
Innovation Solution
A method that involves acquiring first meaning training data and restatement training data, learning the association between utterance sentences and meaning information, and using a common neural network model to determine the importance of words, thereby improving the accuracy of meaning generation by leveraging both types of data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional meaning generation techniques use frequency-based word importance learning, then the learning process is simple, but unimportant words occurring at high frequency are incorrectly recognized as important words
Solution Approach 1:
The patent segments the learning process into two distinct components: a common model that learns general associations between utterance sentences and meaning information, and a restatement model that specifically learns to identify important words by generating restatements. This segmentation allows each model to specialize in its respective task, improving word importance identification accuracy while managing complexity through modular design.
Solution Approach 2:
The patent introduces restatement sentences as an intermediary element that mediates between the utterance sentence and the meaning information. The restatement model generates simplified restatements that highlight important words, and this intermediary representation helps the common model more accurately identify word importance without being misled by high-frequency unimportant words.
2Measurement precision
If a common model learns both meaning information and restatement sentence associations, then meaning generation accuracy is improved, but the learning complexity increases
Solution Approach 1:
The patent merges the learning of two different associations (utterance-sentence to meaning information, and utterance-sentence to restatement sentence) into a single common model. This combining approach allows the model to learn complementary patterns simultaneously, improving meaning generation accuracy while avoiding the need for separate training processes that would increase learning time.
Solution Approach 2:
The common model is designed with multi-functionality, serving both to learn the association between utterance sentences and meaning information and to learn the association between utterance sentences and restatement sentences. This universal model approach improves meaning generation accuracy through dual learning objectives while maintaining a single model structure that efficiently utilizes training data.
Data Source
AI summary
A meaning generation method, in a meaning generation apparatus, includes acquiring meaning training data including text data of a sentence that can be an utterance sentence and meaning information indicating a meaning of the sentence and associated with the text data of the sentence, acquiring restatement training data including the text data of the sentence and text data of a restatement sentence of the sentence, and learning association between the utterance sentence and the meaning information and the restatement sentence. The learning includes learning of a degree of importance of a word included in the utterance sentence, and the learning is performed by applying the meaning training data and the restatement training data to a common model, and storing a result of the learning as learning result information.