AI-ordlista
Den kompletta ordlistan över AI
Sentiment Analysis
NLP technique to identify and quantify opinions, emotions, and attitudes expressed in text, typically classified as positive, negative, or neutral.
Polarity Detection
Process of determining the sentimental orientation of a text by assigning a numerical value representing its degree of positivity or negativity.
Subjectivity Analysis
Classification of a text between subjective content (opinions, personal feelings) and objective content (facts, factual information) without evaluating polarity.
Emotion Detection
Identification and classification of specific emotions (joy, anger, sadness, fear, surprise, disgust) expressed in text beyond simple positive/negative polarity.
ABSA (Aspect-Based Sentiment Analysis)
Granular approach analyzing sentiments associated with specific aspects or entities in text, allowing detailed evaluation by feature or attribute.
Opinion Extraction
Process of identifying and structuredly extracting opinions, targets, and polarities from unstructured texts to create opinion knowledge bases.
Sentiment Lexicon
Specialized dictionary containing words or expressions with their pre-assigned sentiment scores, used as a resource for rule-based sentiment analysis.
VADER
Rule-based sentiment analysis algorithm specifically designed for social media texts, sensitive to intensifiers, punctuation, and emoticons.
BERT for sentiment analysis
Application of the pre-trained language model BERT for sentiment analysis, leveraging its contextual understanding capabilities for accurate classification.
Transformer Models
Neural network architecture based on attention mechanisms, becoming the standard for sentiment analysis due to its advanced contextual understanding.
Fine-grained sentiment analysis
Classification of sentiments on a detailed scale (e.g., very positive, positive, neutral, negative, very negative) rather than a simple binary or ternary classification.
Sentiment classification
Machine learning task consisting of automatically assigning predefined sentiment labels to unlabeled text segments.
Sentiment scoring
Assignment of a continuous numerical score (usually between -1 and 1) representing the sentimental intensity of a text, enabling quantitative comparisons.
Aspect extraction
Automatic identification of specific entities or features on which opinions focus in a text, preliminary step of ABSA.
Target-dependent sentiment analysis
Technique evaluating the sentiment of a text based on the specific entity or target mentioned, recognizing that the same word can have different polarities depending on context.
Multilingual sentiment analysis
Ability to analyze sentiments in multiple languages, either through language-specific models or through cross-lingual transfer approaches.
Domain adaptation in sentiment analysis
Techniques for adapting sentiment analysis models trained on a source domain to work effectively on a different target domain with little labeled data.
Transfer learning for sentiment
Approach that leverages knowledge learned from large corpora to improve performance on specific sentiment analysis tasks with less training data.
Multimodal sentiment analysis
Simultaneous integration and analysis of multiple modalities (text, images, audio, video) to determine the overall sentiment, capturing nuances not accessible through text alone.
Sarcasm detection
Identification of sarcasm and irony in texts, where the literal meaning differs from the actual intent, crucial for accurate sentiment analysis.