Result Objects¶

All extractor classes return a TextRankResult containing a list of Phrase objects. This page documents both.

TextRankResult¶

Returned by every extract_keywords() call on an extractor class.

Attributes¶

Attribute	Type	Description
`phrases`	`list[Phrase]`	List of extracted phrases, sorted by score in descending order.
`converged`	`bool`	Whether the PageRank iteration converged within `max_iterations`.
`iterations`	`int`	Number of PageRank iterations actually run.
`consensus`	`ConsensusPayload \\| None`	AutoRank-only metadata about the executed variant pool and per-phrase agreement.

Methods¶

Method	Returns	Description
`as_tuples()`	`list[tuple[str, float]]`	Returns phrases as `[(text, score), ...]` tuples. Useful for quick inspection or serialization.
`__len__()`	`int`	Number of phrases. Supports `len(result)`.
`__getitem__(idx)`	`Phrase`	Index into the phrase list. Supports `result[0]`.

Example¶

from rapid_textrank import BaseTextRank

extractor = BaseTextRank(top_n=5, language="en")
result = extractor.extract_keywords("Machine learning is a subset of artificial intelligence.")

# Check convergence
print(f"Converged: {result.converged}")
print(f"Iterations: {result.iterations}")

# Iterate over phrases
for phrase in result.phrases:
    print(f"{phrase.text}: {phrase.score:.4f}")

# Quick tuple output
print(result.as_tuples())
# [('machine learning', 0.2341), ('artificial intelligence', 0.1654), ...]

# Length and indexing
print(len(result))     # 5
print(result[0].text)  # 'machine learning'

Phrase¶

Each item in TextRankResult.phrases is a Phrase object.

Attributes¶

Attribute	Type	Description
`text`	`str`	The phrase text as it appears in the document (e.g., `"machine learning"`).
`lemma`	`str`	Lemmatized form of the phrase (e.g., `"machine learning"`). Useful for deduplication.
`score`	`float`	TextRank score. Higher is more important. Scores are not normalized to any fixed range.
`count`	`int`	Number of times this phrase (or its variants) appears in the text.
`rank`	`int`	1-indexed rank position. The top-scoring phrase has `rank=1`.

Example¶

result = extractor.extract_keywords(text)

for phrase in result.phrases:
    print(f"Text:  {phrase.text}")
    print(f"Lemma: {phrase.lemma}")
    print(f"Score: {phrase.score:.4f}")
    print(f"Count: {phrase.count}")
    print(f"Rank:  {phrase.rank}")
    print()

String representation¶

Phrase supports both repr() and str():

phrase = result.phrases[0]

repr(phrase)  # "Phrase(text='machine learning', score=0.2341, rank=1)"
str(phrase)   # "machine learning"

ConsensusPayload¶

Present only for AutoRank results.

Attribute	Type	Description
`selected_variants`	`list[str]`	Canonical names of the executed member variants.
`selection_reason`	`str`	Human-readable summary of why the pool included focus-driven, semantic-driven, or pre-tokenized members.
`variant_runs`	`list[VariantRun]`	Per-variant convergence summaries.
`phrase_support`	`list[PhraseSupport]`	Per-phrase agreement metadata aligned with `TextRankResult.phrases`.

VariantRun¶

Attribute	Type	Description
`variant`	`str`	Canonical variant name.
`converged`	`bool`	Whether that member extractor converged.
`iterations`	`int`	Iterations used by that member extractor.

PhraseSupport¶

Attribute	Type	Description
`confidence`	`float`	Agreement score in `[0, 1]`, based on the supporting variant weights.
`support_count`	`int`	Number of member variants that supported the phrase.
`supporting_variants`	`list[str]`	Canonical names of the supporting variants.

JSON Result Format¶

When using the JSON interface, results are returned as a JSON string with the same structure:

{
    "phrases": [
        {
            "text": "machine learning",
            "lemma": "machine learning",
            "score": 0.2341,
            "count": 2,
            "rank": 1
        }
    ],
    "converged": true,
    "iterations": 12
}