Word Recommendation Embeddings for Foreign Vocabulary Knowledge Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems lack efficient and effective methods for recommending words to foreign language learners using artificial intelligence, making it difficult for users to manually enter unknown words and implement automatic word recommendation systems.
Innovation Solution
A deep learning-based system that utilizes a trained AI model to recommend words by generating user and word embedding vectors, determining their proximity, and outputting a result value for predicting word knowledge, using a user embedding model and word embedding model to encode vectors closely related to each other.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual entry of unknown words is required, then users can create flashcards, but user burden increases and efficiency decreases
Solution Approach 1:
The system automatically tracks user knowledge states and generates word recommendations without requiring manual input. The AI model self-updates based on user interactions, eliminating the need for users to manually enter unknown words or create flashcards.
Solution Approach 2:
The patent replaces the mechanical manual entry process with an automated AI-based system. Instead of users physically typing or writing words, the system uses deep learning models to automatically identify and recommend words based on user behavior patterns.
2Ease of operation
If automatic word recommendation system is implemented, then user burden decreases, but system complexity increases
Solution Approach 1:
The patent introduces embedding models as intermediary components that transform complex user behavior data into simplified vector representations. These embeddings serve as a bridge between raw interaction data and the recommendation engine, reducing the complexity of the overall system while maintaining automation.
Solution Approach 2:
The system transforms complex user knowledge states into simplified numerical parameters (embedding vectors). By representing user knowledge and word characteristics as vectors in a continuous space, the system manages complexity through parameter transformation rather than complex rule-based logic.
3Measurement precision
If AI model is trained with user data, then recommendation accuracy improves, but training time and computational resources increase
Solution Approach 1:
The patent pre-trains embedding models on large corpora before deployment. This preliminary action allows the models to capture general language patterns and user behavior characteristics in advance, reducing the time needed for fine-tuning on specific user data during actual use.
Solution Approach 2:
The system uses pre-trained embedding models that capture more information than immediately necessary. These over-parameterized models are then fine-tuned with user data, allowing the system to achieve high accuracy without requiring extensive training time on user-specific data alone.
Data Source
AI summary
A method in which a user terminal recommends a word to a user according to the present specification, includes generating a user embedding vector by inputting a user vector to a user embedding model; generating a word embedding vector by inputting a word vector to a word embedding model; inputting the user embedding vector and the word embedding vector to a function; outputting a result value for predicting whether the user knows a word related to the word vector from the function; and displaying recommended word information through a display of the user terminal based on the result value, wherein the function output the result value on the basis of proximity of the user embedding vector and the word embedding vector in a user-word joint embedding space.