DictionarySentimentAnalyzer

sentiment.DictionarySentimentAnalyzer(
    loader=LexiconLoader(),
    tokenizer=Tokenizer(),
    sentencizer=Sentencizer(),
)

Sentence-level sentiment scoring built on dictionary lookups.

Methods

Name	Description
mixed_messages	Compute Shannon entropy to quantify mixed sentiment signals.
nrc_emotions	Aggregate NRC-style emotion counts for each sentence.
sentence_scores	Return one sentiment score per sentence.
text_scores	Split full text into sentences and score each one.

mixed_messages

sentiment.DictionarySentimentAnalyzer.mixed_messages(
    text,
    *,
    method=_DEFAULT_METHOD,
    drop_neutral=True,
)

Compute Shannon entropy to quantify mixed sentiment signals.

Parameters

Name	Type	Description	Default
text	str | Sequence[str]	Document (or sequence of segments) to analyze.	required
method	str	Lexicon name to use when determining positive/negative tokens.	_DEFAULT_METHOD
drop_neutral	bool	When True (default), neutral words are excluded from the entropy calculation.	True

Returns

Name	Type	Description
	MixedMessageResult	Named result containing entropy (Shannon entropy over positive/negative distribution) and normalized_entropy (entropy divided by token count for length-independent measure).

Examples

>>> analyzer = DictionarySentimentAnalyzer()
>>> result = analyzer.mixed_messages("I love it but I hate it.")
>>> result.entropy  # Higher values indicate more mixing
>>> result.normalized_entropy  # Length-normalized version

nrc_emotions

sentiment.DictionarySentimentAnalyzer.nrc_emotions(
    sentences,
    *,
    language='english',
    lexicon=None,
    categories=None,
)

Aggregate NRC-style emotion counts for each sentence.

Parameters

Name	Type	Description	Default
sentences	Sequence[str]	Sentences to analyze.	required
language	str	Language tag for loading NRC variants.	'english'
lexicon	EmotionLexicon	Specific NRC emotion lexicon to reuse.	None
categories	Iterable[str]	Restrict the output to a subset of emotion labels.	None

Returns

Name	Type	Description
	list[dict[str, float]]	Per-sentence mappings from emotion label to aggregated weight.

Examples

>>> from moodswing import DictionarySentimentAnalyzer
>>> import pandas as pd
>>>
>>> analyzer = DictionarySentimentAnalyzer()
>>> sentences = ["I love sunny days!", "The storm was terrifying."]
>>> emotions = analyzer.nrc_emotions(sentences)
>>>
>>> # Convert to DataFrame for easy viewing
>>> df = pd.DataFrame(emotions)
>>> print(df[['joy', 'fear', 'positive', 'negative']])
>>>
>>> # Analyze only specific emotions
>>> fear_joy = analyzer.nrc_emotions(
...     sentences,
...     categories=['fear', 'joy']
... )
>>> df_fear_joy = pd.DataFrame(fear_joy)

sentence_scores

sentiment.DictionarySentimentAnalyzer.sentence_scores(
    sentences,
    *,
    method=_DEFAULT_METHOD,
    language='english',
    lexicon=None,
)

Return one sentiment score per sentence.

Parameters

Name	Type	Description	Default
sentences	Sequence[str]	Sentences that have already been split from the source text.	required
method	str	Name of the dictionary lexicon to use (syuzhet, bing, etc.).	_DEFAULT_METHOD
language	str	Language tag used when loading multilingual lexicons such as NRC.	'english'
lexicon	SentimentLexicon	Preloaded lexicon. Provide this when you want to reuse the same instance across multiple calls to avoid I/O.	None

Returns

Name	Type	Description
	list[float]	One numeric sentiment score for each input sentence.

text_scores

sentiment.DictionarySentimentAnalyzer.text_scores(
    text,
    *,
    method=_DEFAULT_METHOD,
    language='english',
    lexicon=None,
)

Split full text into sentences and score each one.

Parameters

Name	Type	Description	Default
text	str | Sequence[str]	Either a single string (the full document) or an iterable of pre-separated paragraphs/segments.	required
method	str	Lexicon name passed to :func:sentence_scores.	_DEFAULT_METHOD
language	str	Language tag used when loading the lexicon.	'english'
lexicon	SentimentLexicon	Preloaded lexicon that overrides the automatic loader.	None

Returns

Name	Type	Description
	list[float]	Sentiment score per detected sentence.