text dictionary rules

Text Dictionary Rules (.NET Port)

This page is the ZemberekDotNet port of Java wiki page Text Dictionary Rules.

It explains dictionary-line syntax used by the Turkish lexicon loader in this repository.

Related implementation:

• TurkishDictionaryLoader
• TurkishDictionaryLoader tests

1. Dictionary Line Format

General format:

word [P:PRIMARY_POS,SECONDARY_POS ; A:ATTRIBUTE1,ATTRIBUTE2 ; Pr:PRONUNCIATION ; Roots:r1-r2 ; Ref:itemId ; Index:n]

Minimal examples:

kalem
okumak
Ankara

2. POS Rules

Metadata key for POS is P.

Examples:

ekşi [P:Adj]
ve [P:Conj]
bu [P:Pron,Demons]
Tdk [P:Abbrv]
… [P:Punc]

Inference rules (same intent as Java docs):

• If a lemma ends with mak or mek, loader infers Verb.
• Otherwise loader usually infers Noun.
• Capitalized lemma is treated as proper noun unless overridden.

Examples:

elma           // inferred noun
okumak         // inferred verb
çomak [P:Noun] // noun ending with -mak requires override

3. Attribute Rules

Metadata key for attributes is A.

Common examples:

bulut [A:NoVoicing]
turp [A:Voicing]
saat [A:InverseHarmony,NoVoicing]
hat [A:Doubling]
ağız [A:LastVowelDrop]
kavurmak [A:LastVowelDrop]

Frequently used attributes in dictionary lines:

• Voicing / NoVoicing
• InverseHarmony
• Doubling
• LastVowelDrop
• CompoundP3sg
• Aorist_A / Aorist_I (verbs)
• NoQuote

4. Compound Entries

For compound forms with embedded possessive behavior, use CompoundP3sg and Roots.

aşevi [A:CompoundP3sg; Roots:aş-ev]
atkuyruğu [A:CompoundP3sg; Roots:at-kuyruk]

5. Pronunciation Rules

Metadata key for pronunciation is Pr.

Use this especially for abbreviations and foreign names where orthography does not provide correct Turkish phonetics.

Google [Pr:gugıl]
A101 [P:Abbrv; Pr:ayüzbir]

Multiple pronunciation workaround (using Ref and Index):

VST [P:Noun,Abbrv; Pr:viesti]
VST [P:Noun,Abbrv; Pr:vesete; Ref:VST_Noun_Abbrv; Index:2]

6. Metadata Keys Supported by Loader

Based on current loader implementation:

• P: POS
• A: Attributes
• Pr: Pronunciation
• Roots: Compound roots
• Ref: Reference item
• Index: Index for alternative entries
• S: Suffix metadata (reserved/special use)

7. C# Usage

Load from inline lines

using ZemberekDotNet.Morphology.Lexicon;
using ZemberekDotNet.Morphology.Lexicon.TR;

RootLexicon lexicon = TurkishDictionaryLoader.Load(
    "elma",
    "okumak",
    "çomak [P:Noun]",
    "Google [Pr:gugıl]",
    "atkuyruğu [A:CompoundP3sg; Roots:at-kuyruk]"
);

Load one item

DictionaryItem item = TurkishDictionaryLoader.LoadFromString("turp [A:Voicing]");

Load from file

RootLexicon lexicon = TurkishDictionaryLoader.Load("Resources/my-dictionary.txt");

8. Verification

Use existing tests as canonical behavior references:

• POS inference and noun/verb behavior
• voicing and attribute handling
• pronunciation + reference/index behavior
• compound root handling

See TurkishDictionaryLoader tests.

9. Differences and Notes

• This page is a .NET-targeted port, so examples and method names follow C# APIs.
• Behavior is aligned with current loader implementation and tests in this repo.
• If behavior differs from Java wiki wording, implementation and tests in this repository are authoritative.