Tamil Dictionary

This program hosts a Tamil dictionary website allowing fast searching in both headwords and definitions. It uses a custom query syntax based on a restricted subset of regular expressions designed specifically for Tamil text. It can be configured to host any Tamil dictionary data, if it is in the correct format.

Implementation

The web server and search engine are implemented in Rust using the Rocket web framework and Handlebars for templating. The pages are written in HTML, CSS, and vanilla JavsScript. The dictionary uses suffix trees to allow for O(n) searches with respect to the length of the query string.

Features

• Word and definition search
• Filter by part of speech
• Suggestions in search bar
• Ranking of results based on similarity to query
• True dictionary sort order (not based on Unicode order)
• Automatic transliteration (maram => மரம்)
• Awareness of grammar rules (கடல்கரை => கடற்கரை)
• Partial support for spoken Tamil (கொளந்த => குழந்தை)
• Definitions contain hints for spoken Tamil pronunciation
• Automatic light/dark theme depending on device settings
• Annotater which adds links to definitions for words (at /annotate)
• Random word search (at /random)

Installation

Clone the repository:

git clone https://github.com/scott2000/tamil-dictionary.git

Dictionary file

Create a file called dictionary.json with the contents of your dictionary represented as a list of objects. In each object, there should be a word field containing the headword (which may optionally contain commas separating multiple words), an optional sub field containing the index of the subword if there are multiple entries for the same word, an optional hint field containing text to place at the start of the definition, a kind field indicating the part of speech, and a list of sections.

The kind field contains an array of parts of speech which may be:

• v - வினைச்சொல் (verb)
• va - வினையடை (adverb)
• vm - வினைமுற்று (finite verb)
• tv - துணை வினை (auxiliary verb)
• p - பெயர்ச்சொல் (noun)
• pa - பெயரடை (adjective)
• sp - சுட்டுப் பெயர்ச்சொல் (pronoun)
• spa - சுட்டுப் பெயரடை (adjective of pronoun)
• vp - வினாப் பெயர்ச்சொல் (question word)
• vpa - வினாப் பெயரடை (adjective of question word)
• i - இடைச்சொல் (grammatical particle)
• ii - இணையிடைச்சொல் (conjunction)
• vi - விளி இடைச்சொல் (vocative)

Each section contains a list of paragraphs, with the first paragraph being treated as a header for the section, and the rest being treated as numbered definitions. A paragraph contains a list of segments, with each segment containing one of txt, ref, sup, bld, wbr and a string. The txt segment renders plain text, ref denotes a reference to another word in the dictionary, sup creates a superscript, bld bolds text, and wbr inserts <wbr> before text. Here is an example object:

{
  "word": "மரம்",
  "sub": 1,
  "formal": false,
  "hint": "(common word)",
  "kind": ["p"],
  "secs": [
    [
      [
        ["txt", "Oblique: "],
        ["bld", "மரத்து"]
      ],
      [
        ["txt", "tree (see also: "],
        ["ref", "ஆலமரம்"],
        ["sup", "1"],
        ["txt", ")"]
      ],
      [
        ["txt", "wood"]
      ]
    ],
    [
      [
        ["txt", "Another section:"]
      ],
      [
        ["txt", "ThisIsAlongWordSoItMay"],
        ["wbr", "BeHelpfulToAddAWordBreak"]
      ]
    ]
  ]
}

Which will be rendered as:

---

மரம்¹ (common word) பெ. Oblique: மரத்து

1. tree (see also: ஆலமரம்¹)
2. wood

Another section:

3. ThisIsAlongWordSoItMayBeHelpfulToAddAWordBreak

---

Verbs file

There is an optional feature located at /annotate which takes some input text, finds the definitions corresponding to each word in the text, and adds links to those words. Since Tamil verbs can be conjugated in many different ways, this feature requires a specification of how to conjugate each verb in the dicitonary. This is done by providing all possible வினையெச்சம் forms of each verb in a file called verbs.json. This file contains an array of objects. Here is an example object:

{
  "word": "வா",
  "sub": 1,
  "ve": ["வர", "வந்து"]
}

If there's a word which shares a conjugation with another verb, you may use hyphens to indicate that only a partial conjugation was given:

{
  "word": "கொண்டுவா",
  "ve": ["-வர", "-வந்து"]
}

Starting the server

Make sure you have installed Rust and Cargo, and run cargo run --release to start the server on localhost:8000. A Rocket.toml file can be used to configure the server further.