FinBERTToneOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for FinBERT-Tone model for fine-grained financial sentiment/tone analysis.

Inheritance: ModelOptions → FinBERTToneOptions<T>

For Beginners

FinBERT-Tone provides nuanced sentiment analysis:

The Key Insight: Standard sentiment models use 3 classes (negative/neutral/positive), but financial communications often have more subtle tones. FinBERT-Tone uses 5 classes to capture gradations like “cautiously optimistic” vs “strongly bullish”.

What Problems Does FinBERT-Tone Solve?

  • Earnings call tone analysis (detecting management confidence levels)
  • Forward-looking statement sentiment classification
  • Analyst report tone assessment
  • Corporate communication sentiment tracking
  • Detecting subtle changes in disclosure tone over time

The 5 Tone Classes:

  1. Very Negative: Strong pessimism, significant concerns
  2. Negative: Pessimistic outlook, concerns mentioned
  3. Neutral: Factual, no clear sentiment
  4. Positive: Optimistic outlook, confidence expressed
  5. Very Positive: Strong optimism, high confidence

Key Benefits:

  • Fine-grained sentiment detection
  • Better captures management tone nuances
  • Useful for sentiment momentum tracking
  • Effective for earnings call analysis

How It Works

FinBERT-Tone is a FinBERT variant specifically focused on capturing fine-grained sentiment and tone in financial communications. It uses 5 classes to capture more nuanced sentiment than standard 3-class models.

Reference: Huang et al., “FinBERT: A Pretrained Language Model for Financial Communications”, 2020.

Constructors

Constructor Summary
FinBERTToneOptions() Initializes a new instance with default FinBERT-Tone configuration.
FinBERTToneOptions(FinBERTToneOptions<T> other) Initializes a new instance by copying from another instance.

Properties

Property Summary
bool UseFinegrainedTone { get; set; } Whether to use fine-grained 5-class tone (true) or standard 3-class sentiment (false).
double DropoutRate { get; set; } Dropout rate (default: 0.1).
int HiddenDimension { get; set; } Hidden dimension (default: 768 for BERT-base).
int IntermediateDimension { get; set; } Intermediate feed-forward dimension (default: 3072).
int MaxSequenceLength { get; set; } Maximum sequence length in tokens (default: 512).
int NumAttentionHeads { get; set; } Number of attention heads (default: 12).
int NumLayers { get; set; } Number of transformer layers (default: 12).
int NumToneClasses { get; set; } Number of tone classes (default: 5 for fine-grained sentiment).
int VocabularySize { get; set; } Vocabulary size (default: 30522 for BERT-base).

Methods

Method Summary
void Validate() Validates the FinBERT-Tone options.