AI 용어집
인공지능 완전 사전
Content Vectorization
The process of converting textual or multimedia content into numerical vectors in a multidimensional space to enable mathematical similarity calculations between items.
User Profile
A structured representation of a user's preferences and interests, built from their interaction history and used to generate personalized recommendations.
Item Profile
A vectorized description of an item's intrinsic characteristics (attributes, content, metadata) used as a basis to identify similar items in the recommendation system.
Feature Extraction
The algorithmic process of identifying and isolating relevant attributes from raw content to build numerical representations usable by recommendation algorithms.
Bag of Words
A text representation model that ignores grammar and word order to focus solely on the frequency of term occurrences within a document.
Cold Start Problem
The inherent difficulty for recommendation systems when they need to generate suggestions for new users or items without sufficient interaction history.
Semantic Embeddings
Dense vector representations that capture semantic and contextual relationships between words, phrases, or documents in a continuous, low-dimensional space.
Term-Document Matrix
A mathematical structure where each row represents a vocabulary term and each column a document, with cells containing the weights or frequencies of terms in the documents.
Latent Semantic Indexing
Semantic analysis technique that uses singular value decomposition to discover hidden relationships between terms and documents in a corpus.
Term Weighting
Strategy for assigning numerical weights to terms in a document to reflect their relative importance, influencing the relevance of content-based recommendations.
Explicit Feedback
Direct and intentional ratings provided by users (ratings, likes, comments) used to refine preference profiles in recommender systems.
Implicit Feedback
Indirect behavioral signals (view time, clicks, shares) inferred to deduce user preferences without direct evaluation of the content.
Jaccard Distance
Similarity metric for sets defined as the size of the intersection divided by the size of the union, used to compare interest profiles between users.
Context Window
Parameter defining the range of surrounding words considered during semantic analysis to capture local relationships and improve the quality of embeddings.