Phrase Indexer

The Phrase Indexer node creates a searchable index from a selected string column that contains multi-word phrases.
Before indexing, each phrase is split into individual tokens based on a user-defined delimiter (default: blank space). Each token is then indexed, enabling efficient approximate string matching over multi-word data.
This node is particularly useful for text fields containing full names, product descriptions, or address lines, where indexing based on words rather than entire strings improves recall and match flexibility.
The generated index can be passed to downstream node Approximate Phrase Index Matcher, enabling rapid retrieval of similar phrases or partial matches from large datasets.
Optionally, an Alias Object can expand the search with synonymous, canonical, or normalized forms (e.g., “NYC” ⇄ “New York City”), improving recall without sacrificing determinism.

Options

Select Column to Index

Selects the string column from the input table that contains the phrases to be indexed. Only string-convertible columns are available.

Delimiter

Specifies the character or pattern used to split each phrase into words before indexing.
By default, phrases are split by a blank space (" ").
Example delimiters: comma (,), semicolon (;), or custom token separators.
Proper delimiter selection helps ensure correct tokenization and optimal match performance.

Representation of Indexed Strings

Determines how indexed values are represented in the output table of downstream matching nodes.
Options:

• Original - Displays strings as they appear in the input column.
• Normalized - Displays transformed versions to improve fuzzy-matching precision.

Aliases

Select the alias set you want to apply to the index.
An Alias Object (created by the Alias Creator node) enables deterministic synonym expansion or canonicalization during indexing.
When an alias set is applied:

• Terms may be expanded or rewritten according to alias rules
• Penalty values influence downstream similarity scoring
• Deterministic synonym handling improves recall while preserving explainability

If no alias set is selected, indexing proceeds without synonym expansion.

Views

This node has no views

Workflows

• Address Normalization & MatchingKNIME Hub
• Product Title Normalization using Phrase Indexer (E-Commerce)KNIME Hub

Links

• No links available

Developers