MiscDataNotes

GlobalQuran is a well written eco-system of UI pages, widgets, APIs, search that anyone can use on their site.

This page is just to explore & evaluate new DataSets that may or may not end up being used in GlobalQuran.

Word2Word data:
- Objective: For each quran word, get the english meaning. No compound word meanings.
- Reliability: *** (3/5) almost as good as source
- Demo: [http://qurandev.github.com/ http://qurandev.github.com/] -- see the 4th column
- Challenges: as its machine matched, data may not be 100% accurate. For compound words, most meanings are prefixed with **. These are where the meaning is duplicated for side-by-side words.
- How data generated?:
  - started with the orig data from [http://allahsquran.com/learn/ http://allahsquran.com/learn/]
  - snip out all meanings from js. Ex: [http://allahsquran.com/learn/ayas-s112d7q1f0o4.js http://allahsquran.com/learn/ayas-s112d7q1f0o4.js]
  - Identify all places where theres mismatch in #words in tanzil and the meaning segments.
  - where mismatch, either fixed by hand in few cases. or duplicate the meaning after prepend **.
- Note: Apparently quran.com and corpus.quran.com also had to go thru these word2word data cleanups.

Buck data:
- Objective: Instead of passing Tanzil Quran data as unicode, pass it as ascii. One-to-one mapping of arabic unicode to a ascii, which can be mapped & remapped, without loss of fidelity.
- Reliability: **** (4/5) as good as Tanzil, the source
- Demo: [http://qurandev.github.com/ http://qurandev.github.com/] -- see 3rd column
- Advantages?
  - Buck uses less bandwidth
  - In Javascript, u can search thru entire Buck quran text in 1 shot. intuitive compared to Arabic search
  - Buck to Arabic and Arabic to Buck is a simple js call. Play with live sample here: [http://jsfiddle.net/BrxJP/ http://jsfiddle.net/BrxJP/].
  - You can strip out all vowels from Buck text in few millisecs. Why do this? u can search in javascript, ignoring the taskheel differences (fatha, damma, kasra). They leads to more hits.
  - Regex + buck text can lead to awesome optimizations. All the searches can be run locally. Demo: [http://qurandev.appspot.com/ http://qurandev.appspot.com/]
- How data generated? just one-to-one mapping using: [http://corpus.quran.com/java/buckwalter.jsp http://corpus.quran.com/java/buckwalter.jsp]

Corpus Grammar data:
- Objective: get grammatical data for each word.
- Reliability: ** (2/5) Corpus is a work-in-progress. 40% complete.Will have mistakes!
- Demo: [http://qurandev.github.com/ http://qurandev.github.com/] -- see last column.
  - [http://qurandev.github.com/Grammar.html http://qurandev.github.com/Grammar.html] you can do local live in-browser searches, similar to corpus.quran.com website. can Download & browse locally.
- Advantages? only source which has raw data published.
- Challenges: grammatical data is NOT at word level; its at word-segment level. Each word= n*PREFIX + 1 or 2 STEMS + n*SUFFIX.
- Columns not uniform. i.e. https://raw.github.com/qurandev/qurandev/master/data/quranic-corpus-morphology-0.4.txt
- demo relies on using heavyduty Javacript RegExs to cut out info for word & to parse out grammar info. May want to do 1-time premapping into ur own format, but will have to redo if corpus releases new data file.

Transliteration data based on Corpus:
- Not Done. It shud be simple 1-to-1 mapping using [http://corpus.quran.com/documentation/phonetic.jsp http://corpus.quran.com/documentation/phonetic.jsp]
- Alternate approach: Used a different data file which has tranliteration encoded in html format. using jquery and regex allows me to do search based on transliteration locally. (no server call)
- Demo: [http://qurandev.appspot.com/ http://qurandev.appspot.com/] type in 2:255

Comments / Questions??