dictmaker

README

dictmaker

dictmaker is a stand-alone, single-binary server application for building and publishing dictionary websites. Alar (Kannada-English dictionary) is an example in production.

• Generic entry-relation (many)-entry structure for dictionary data in a Postgres database with just two tables (entries, relations)
• Entirely built on top of Postgres full text search using tsvector tokens
• Works with any language. Plug in external tokenizers or use a built in tokenizer supported by Postgres for search
• Possible to have entries and definitions in any number of languages in a single database
• HTTP/JSON REST like APIs
• Themes and templates to publish dictionary websites
• Paginated A-Z (or any alphabet for any language) glossary generation for dictionary words

How it works

dictmaker has no concept of language or semantics. To make a universal dictionary interface possible, it treats all data as unicode strings that can be searched with Postgres DB's fulltext capabilities, where search tokens are generated and stored along with the dictionary entries. There are several built-in fulltext dictionaries and tokenizers that Postgres supports out of the box, mostly European languages (\dFd lists installed dictionaries). For languages that do not have fulltext dictionaries, it is possible to generate search tokens manually using an external algorithm. For instance, to make a phonetic English dictionary, Metaphone tokens can be generated manually and search queries issued against them.

Concepts

• There can be any number of languages defined in the dictionary. eg: 'english', 'malayalam', 'kannada' etc.
• All content, the entry words and their definitions, are stored in the entries table
• Entry-definition many-to-many relationships are stored in the relations table, represented by from_id (entry word) -> to_id (definition word), where both IDs refer to the entries table.

entries table schema

id	SERIAL
guid	TEXT	A custom, unique GUID for every entry, like a UUID or a hash
content	TEXT	Actual language content. Dictionary word or definition entries
initial	TEXT	The first "alphabet" of the content. For English, for the word Apple, the initial is A
weight	INT	An optional numeric value to order search results
tokens	TSVECTOR	Fulltext search tokens. For English, Postgres' built-in tokenizer gives to_tsvector('fully conditioned') = 'condit':2 'fulli':1
lang	TEXT	String representing the language of content. Eg: en, english
types	TEXT[]	Strings describing the types of content. Eg {noun, propernoun}
tags	TEXT[]	Optional tags
phones	TEXT[]	Phonetic (pronunciation) descriptions of the content. Eg: {ap(ə)l, aapl} for Apple
notes	TEXT	Optional text notes

relations table schema

from_id	INT	ID of the head word or the dictionary entry in the entries table
to_id	INT	ID of the definition content in the entry table
types	TEXT[]	Strings describing the types of this relation. Eg {noun, propernoun}
weight	INT	An optional numeric value to order definition results
tags	TEXT[]	Optional tags
notes	TEXT	Optional text notes

Installation

1. Download the latest release release.
2. Run ./dictmaker --new to generate a sample config.toml. Edit the config.
3. Define your dictionary's languages and properties along with other config in config.toml
4. Run ./dictmaker --install to install the Postgres DB schema.
5. Populate the entries and relations tables with your dictionary data. See the "Sample dictionary" section below
6. Run the dictionary server: ./dictmaker

Dictionary query API

# /dictionary/from_lang/to_lang/word
# Optional query params: ?type=noun&type=noun2&tag=a&tag=b ...
curl localhost:8080/api/dictionary/english/english/apple

Sample dictionary

The sample/sample.sql shows how to setup an English-English and English-Italian dictionary. Retaining the English-Italian config in the generated sample config file, execute sample/sample.sql on your database.

Then try:

curl localhost:8080/api/dictionary/english/english/apple
curl localhost:8080/api/dictionary/english/italian/apple

Themes

See the alar-dict/alar.ink repository that powers the Alar dictionary. A theme is a directory with a collection of Go HTML templates. Run a theme by passing ./dictmaker --site=theme_dir.

Tokenizer plugins

For languages that do not have Postgres fulltext dictionaries and tokenizers, dictmaker supports loading compiled Go tokenizer plugins that implement the search.Tokenizer interface. See tokenizers/kannada (and Makefile for compilation help).

Licensed under the AGPL v3 license.