Technical Reports

Proposed Draft Unicode Technical Report #XX

Character Foldings

   L2/01-459

NOTE TO UTC:

In response to the review at UTC#88 I have revised the text and present it again as an L2 document for internal review. I will collect all feedback addressed to me and generate a new version for UTC#89.

Summary

Status of this document

This document has been approved by the Unicode Technical Committee for membership-only review as a Proposed Draft Unicode Technical Report. Making this document available for internal review does not imply endorsement by the Unicode Consortium. This is an internal draft document which may be updated, replaced, or superseded by other documents at any time. This is not a stable document; it is inappropriate to cite this document as other than a work in progress.

A list of current Unicode Technical Reports is found on http://www.unicode.org/unicode/reports/. For more information about versions of the Unicode Standard, see http://www.unicode.org/unicode/standard/versions/.

Contents

• 1 Scope
• 2 Rationale
• 3 Definitions
• 4 Specifications
  • 5.1 Folding algorithm
  • 5.1 Specification of foldings
• 5 Notes and Guidelines
• References
• Modifications

1.0 Scope

This technical reports defines a consistent set of folding operations useful for fuzzy searches, among other things. Each of the folding operations specified in this report has well understood properties, and is appropriate in specific contexts. For example some identifiers needs case folding, some do not. Some text searches needs to preserve <super> form (trademark symbol), while others do not. Not all of these folding operations may be appropriate in the same contexts. See the description for some of the more problematic folding or expansion operations.

This report defines a single search term folding specification (STF) which combines canonical normalization with optional folding operations.

e.g. STF = NFx+<case folding>+<font>+<super>+<sup>

This allows implementers to decide which folding option is useful for a particular use.

String foldings for identifier matches are a related to search term foldings, but with the important difference that the set of allowable characters is restricted. Where required, identifier matches are addressed specifically below.

2.0 Rationale

For the purpose of fuzzy text matching, including both defined identifier matching and general text searching it is often necessary to selectively ignore otherwise meaningful distinctions between related characters, for example case, wide vs. narrow, etc. This process can be called search term folding. Depending on the operation, different foldings need to be applied, and possible interactions must be carefully managed.

2.1 Relation to Normalization

Search term folding  is somewhat similar to normalization in that it also reduces several alternative representations into one, but there are important differences. Normalization is often intended for permanent transformation of data,  but search term folding is by nature transient. Where the normalization forms offer two different levels of distinctions they preserve (canonical and compatibility),  the choice of what type of distinction can or even must be ignored for search term folding needs to be more specific and depends on the nature of the operation. One size does not fit all. Some compatibility mappings have the unfortunate side effect of affecting search term stability, preventing their use in general text searches.

Furthermore, normalization and case folding are defined as separate and independent operations, but  case folding often occurs together with other foldings in search term folding. In order to avoid inconsistencies, search term folding needs to address their interaction.

Search term folding is best expressed as applied to text after canonical normalization, however, whether the normalized text is in the composed or decomposed form is of secondary importance in terms of defining the foldings. It will be relevant to distinguish between NFC and NFD whenever pre-folded search terms are to be interchanged. Otherwise, due to the transient nature of search term folding, the distinction is immaterial.

3.0 Definitions

Folding operation – a folding operation removes a distinction between related characters. For example, case folding removes the case distinction, by replacing upper and title case variants of a character with the lower case.

Compatibility mappings – compatibility mappings substitute characters with their compatibility decomposition. Many compatibility mappings are foldings, some are multigraph expansions. 

Multigraph expansion – a multigraph expansion replaces a multigraph, such as e.g. double prime, by its expansion into an equivalent series of single characters, in this case, two single primes. Multigraph expansions are a subset of expanding compatibility mappings. 

4.0 Specifications

4.1 Basic algorithm

The basic algorithm for search term folding can be stated as

a. Apply canonical decomposition
b. Apply optional  folding operations
c. Recurse 1 & 2 until stable (*)
d. Apply composition if necessary

(*) Step (c)  might be removed by by enforcing certain foldings before decomposition.

For identifier folding, it is important to account for prohibited characters. The basic algorithm then becomes:

a. Apply canonical decomposition
b. Apply optional  folding operations (*)
c. Repeat a & b until stable (**)
d. Apply composition if necessary
e. Eliminate or flag forbidden characters

(*) Foldings can disallow certain characters by mapping them to forbidden characters.
(**) Step (c)  might be removed by by enforcing certain foldings before decomposition.

4.2 Specification of folding operations

[Ed note: Files containing tables for Letterform folding and Kana folding are needed]

The following table summarizes the definition of the folding operations. Foldings that are multigraph expansions have been collected at the end of the table.

• The description column identifies the folding.
• The source column identifies the set of characters subject to the folding operation by referencing a set of code points, a set of general categories, or a compatibility mapping tag. All characters matching the source condition are subject to the given folding. Note that this column does not indicate the set of characters with which the source characters are equivalenced by the folding.
• The target column indicates the result of the folding, either by reference to an operation, or, in some cases, by providing the single Unicode character to which a whole set of source characters is folded.
• The data file column indicates which data file carries the character by character information to implement the operation referred to in the target column. 

Notation: 

[Ed. Note: Additional foldings that have been suggested

• Slash folding (eg. 2044 FRACTION SLASH to 002F SOLIDUS)
• Additional look-alike foldings (visually confusable)

would need tables for the latter to add them]

5.0 Notes and Guidelines

5.1 General notes

5.1.1 Case folding

The [CaseFolding] data file provides case folding information.

5.1.2 Diacritic folding

Diacritic folding goes beyond the decomposition and removal of accents, umlauts, cedillas etc. that is provided by the canonical decompositions, but  also includes barred, slashed forms etc, as well as hooks, descenders, etc.

5.1.3 Letter forms

Greek letter forms should be folded for Greek text. They should not be folded for mathematical and scientific usage as doing so would conflate very distinct concepts (e.g. angle (THETA) and temperature (THETA SYMBOL) to give an examples of common usage in physics).

[Ed. note: Letterforms need to have 'final sigma' and 'final Hebrew' added to them. ]

5.1.4 Semantically neutral foldings

Text that is subjected to a standard Unicode rendering process should display the same whether or not the semantically neutral foldings have been applied (in fact, in practice it is far more likely that un-folded data will display poorly on such a system). 

Therefore these foldings are candidates for permanent data transformations.

The following foldings are semantically neutral

• Positional forms folding
• Vertical forms folding

5.1.5 Locale based foldings

Locale based foldings would fold characters that are 'nearly equivalent' in a particular langauage. For example, a locale based folding for Swedish could match

oe - o umlaut o slash - o umlaut ss to sharp s y to u and v to w

etc. In other words, locale based foldings would be different for some user groups using the same script (in this case Latin). A suggestion of how to implement locale based searching based on sorting tables is found in [Collation].

5.1.6 Compatibility decompositions

Compatibility decomposition provides a fixed combination of several foldings and expansions. It is in fact the source of most of the foldings in the table in section 4.0. There are two ways to subdivide compatibility decompositions

1. by compatibility decomposition type (the value in <> in the date file, e.g. <super>)
2. by explicitly limiting the range of source characters

The specifications in section 4.0 uses the first method, whenever the compatibility tag is well defined and meaningful. Where it is too broad, e.g., for the  <compat> tag, foldings are further subdivided by defining specific ranges of source characters.

Using compatibility decomposition is convenient, since existing algorithms for Normalization may provide them. However, the full decomposition includes several foldings that are problematic in and of themselves, and possibly others that happen to not be appropriate for the given purpose. By selectively not applying the decomposition to certain character ranges given in Section 4.0, one can limit the compatibility decomposition to only the desired foldings.

5.2 Problematic foldings or expansions

Some foldings can have unintended consequences, including inadvertent changes in the semantics of the text. In most cases, it is best to be conservative and avoid problematic foldings altogether. There are two general exceptions to this rule. The first is the case of identifier matching. If a folding has a prohibited character as input or as one of the output characters, it will not match any legal identifier. Therefore, for properly restricted inputs, one may safely use fixed combinations of foldings, such as NFKC. The other exception is the case of more extensive string pre-processing, discussed below.

5.2.1 Fraction expansion

Fraction expansion as defined in the compatibility decompositions can lead to a drastic change of the semantics of a string and can lead to term boundary issues for searching. For example:  Expanding the fraction in this string: DIGIT 5 + VULGAR FRACTION ONE QUARTER  turns it into DIGIT 5 + DIGIT 1 + FRACTION SLASH + DIGIT 4. This now will be found by a search for "51". Because of the semantics of FRACTION SLASH the expansion changed the numeric value from "5 and a quarter" into "51 over 4". Fraction expansion is therefore best avoided altogether.

5.2.2 Bullet expansions

If a circled bullet character is simply replaced by its contents, e.g.  CIRCLED DIGIT 5 is replaced by DIGIT 5, the separation from the surrounding text is lost, and the DIGIT 5 could run together with adjacent numbers. For bullet characters using parenthesized or dotted letters or digit, this issue is somewhat mitigated by fact that the bullet itself contains punctuation. Bullet characters are commonly used like footnote marks to refer to other text, in other words, they do not just occur at the beginning of bulleted lines.

5.2.3 Spacing accents substitution

Spacing accents are mapped by compatibility decomposition to SPACE + non-spacing accent. This inappropriately introduces a space character into the term, as well as introducing non-spacing marks where none were in the data before. This former is especially problematic where the matching operation is bounded by spaces.

5.2.4 Math folding

The set of compatibility decompositions include the folding of letterlike mathematical symbols to their nearest ASCII or Hebrew equivalent. In particular the Hebrew characters used as letterlike symbols do not have RIGHT TO LEFT directionality and the set of such letters in mathematical usage is sufficiently restricted that such folding makes little sense, except in pure 'looks like' style searches.

5.2.5 Various "cluster" expansions

Unicode contains many clusters, e.g. square symbols, some of the letterlike characters that are made up of several characters. 'Decomposing' these may or may not be the right thing for search equivalence. Parenthesized characters and numbers would probably be immune to the term boundaries issues raised earlier, but the story is less clear for others.

[Ed Note: add section on Jamos - get text for note from Mark]

5.2.6 Preserving semantics

To a limited extent, the problems surrounding bullet expansion can be mitigated by inserting a THIN SPACE around the expansion to set the expanded text off from the surrounding text. This must be applied after any space folding, as otherwise the THIN SPACE will become SPACE and might be considered a search term delimiter.

References

[AccentFolding]

Data file <ftp://ftp.unicode.org/Public/UNIDATA/AccentFolding.txt>

[CaseFolding]

Data file <ftp://ftp.unicode.org/Public/UNIDATA/CaseFolding.txt>

[Case Mapping] 

Mark Davis, Unicode Technical Report #21: Case Mapping, <http://www.unicode.org/unicode/reports/tr21>

[Character Mapping Tables]

Mark Davis, Unicode Technical Report #22: Character Mapping Tables, <http://www.unicode.org/unicode/reports/tr22>

[Collation]

Mark Davis, Unicode Technical Report #10: Collation, <http://www.unicode.org/unicode/reports/tr10>

[CompatibilityTags]

[EastAsianWidth]

Data file <ftp://ftp.unicode.org/Public/UNIDATA/EastAsianWidth.txt>

[East Asian Width] 

Asmus Freytag, Unicode Standard Annex #11, East Asian Width, <http://www.unicode.org/unicode/reports/tr11>

[KanaFolding]

Data file <ftp://ftp.unicode.org/Public/UNIDATA/KanaFolding.txt>

[SpecialCasing]

Data file <ftp://ftp.unicode.org/Public/UNIDATA/SpecialCasing.txt>

[Unicode]

The Unicode Standard, Version 3.0, Addison Wesley Longman, 2000.

[UnicodeCharacterDatabase]

Readme file, <ftp://ftp.unicode.org/Public/UNIDATA/UnicodeCharacterDatabase.html>

[UnicodeData]

Data file <ftp://ftp.unicode.org/Public/UNIDATA/UnicodeData.txt>

[UnicodeData-Format]

Readme file <ftp://ftp.unicode.org/Public/UNIDATA/UnicodeData.html>

[Ed. note: These may not be needed. Remove unneeded ones before publication]

[Bidirectional Algorithm]

Mark Davis, Unicode Standard Annex #9: The Bidirectional Algorithm, <http://www.unicode.org/unicode/reports/tr9>

[LineBreak]

Data file <ftp://ftp.unicode.org/Public/UNIDATA/LineBreak.txt>

[Line Breaking]

Asmus Freytag, Unicode Standard Annex #14: Line Breaking Properties, <http://www.unicode.org/unicode/reports/tr14>

[NamesList]

Data file <ftp://ftp.unicode.org/Public/UNIDATA/NamesList.txt>

[NamesList-Format]

Readme file <ftp://ftp.unicode.org/Public/UNIDATA/NamesList.html>

Changes from previous drafts

Improved the notes and guidelines section, the algorithm description, and many details based on UTC feedback. Rationalized editorial notes throughout.

Copyright © 2000-2002 Unicode, Inc. All Rights Reserved. The Unicode Consortium makes no expressed or implied warranty of any kind, and assumes no liability for errors or omissions. No liability is assumed for incidental and consequential damages in connection with or arising out of the use of the information or programs contained or accompanying this technical report.

Unicode and the Unicode logo are trademarks of Unicode, Inc., and are registered in some jurisdictions.