Search Results cluster_word

Overview

APPS.HZ_TRANS_PKG is a foundational PL/SQL package body within the Oracle E-Business Suite Trading Community Architecture (TCA). Its principal business function is to provide the low-level transformation, cleansing, and comparison engine used by Oracle's Data Quality Management (DQM) and Duplicate Prevention infrastructure. The package normalizes and standardizes party names, organization names, addresses, and related attributes so that records can be reliably compared, scored, and matched against one another. In effect, it supplies the deterministic and fuzzy matching primitives on which the Party Match and Deduplication processes depend.

The package is classified as OTHER in the ETRM metadata and is owned by APPS. It is referenced by 43 other packages, confirming its role as a shared utility rather than an isolated application component. The globals declared at the head of the body—threshold lengths, unprintable character sets, and a series of cached "last" comparison strings (for example, the exact and original forms of person, organization, name, address, and state)—indicate a design optimized for repeated transformation of similar values during bulk matching runs, minimizing redundant recomputation.

Key Procedures and Functions

The package exposes 47 documented procedures and functions. Those surfaced in the metadata include:

• EXACT — Performs a direct, literal comparison between two values, establishing the baseline case for matching.
• CLEANSE — Applies general cleansing logic, stripping noise and unprintable characters to produce a canonical value.
• CLUSTER_WORD — Groups or reorders words within a string, enabling order-insensitive comparison of names and descriptions; this is the routine most directly associated with the user's search term.
• RM_SPLCHAR / RM_BLANKS — Remove special characters and redundant blank spaces respectively, feeding cleaner input to downstream comparisons.
• SOUNDX — Generates a phonetic (Soundex-style) key so that names that sound alike match despite spelling differences.
• EXACT_PADDED — Performs exact comparison after padding, used where length normalization is required.
• ACRONYM / REVERSE_NAME — Derive acronym forms and reversed name forms to support alternate match paths.
• EXACT_DATE, EXACT_NUMBER, EXACT_EMAIL, CLEANSED_EMAIL, EXACT_URL, CLEANSED_URL — Type-specific exact and cleansed comparators for dates, numbers, e-mail addresses, and URLs.
• WRPERSON_EXACT, WRPERSON_CLEANSE, WRPERSON_CLUSTER — Word-replacement-aware variants applied to person names, drawing on the word replacement and word list tables.
• WRORG_EXACT, WRORG_CLEANSE — Equivalent word-replacement-aware routines applied to organization names.

Together these routines implement the transformation pipeline: removal of noise, normalization, clustering, phonetic encoding, and final comparison.

Tables Accessed

The package references several tables through APPS synonyms:

• HZ_WORD_LISTS — Supplies the lists of words used during cleansing, word clustering, and noise-word handling.
• HZ_WORD_REPLACEMENTS — Provides substitution pairs that standardize common tokens (abbreviations, alternate spellings) before comparison.
• HZ_DQM_STAGE_LOG — Records staging and processing activity tied to DQM batch operations.
• PLITBLM — A PL/SQL internal table construct used for in-memory manipulation of word collections.

Usage Notes

HZ_TRANS_PKG is not typically invoked directly by end users. It is called programmatically by higher-level TCA and DQM packages, by concurrent programs that run party deduplication and match scoring, and by any custom code that must reproduce Oracle's standard cleansing or matching behavior. Because matching results depend heavily on the content of HZ_WORD_LISTS and HZ_WORD_REPLACEMENTS, administrators should treat those tables as configuration. Behavior is consistent across Oracle EBS 12.1.1 and 12.2.2, where the package remains a core element of the party matching stack.