Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/loredanacirstea/throtrans

graduation thesis project for U.M.F. "Carol Davila" - “Algorithm of translation in the Romanian language for the Terminologia Histologica and APIs for a controlled vocabulary”
https://github.com/loredanacirstea/throtrans

Last synced: about 1 month ago
JSON representation

graduation thesis project for U.M.F. "Carol Davila" - “Algorithm of translation in the Romanian language for the Terminologia Histologica and APIs for a controlled vocabulary”

Awesome Lists containing this project

README 

        
My Graduation Thesis

=========



###_Algorithm of translation in the Romanian language for the Terminologia Histologica and APIs for a controlled vocabulary_



thRoTrans is part of my graduation thesis project for the University of Medicine and Pharmacy "Carol Davila", titled: 

 - En: “Algorithm of translation in the Romanian language for the Terminologia Histologica and APIs for a controlled vocabulary”

 - Ro: "Algoritm de traducere în limba română pentru Terminologia Histologica și utilitare de mentenanță pentru un vocabular controlat"



Field: Semantic Bioinformatics - Semantic Web/Translingual Web in Medicine



Principles and Concepts

-----------------------



 - transparency in the process of creation, collaboration, evaluation:

    - GitHub (code)

    - Google Docs (paper, methods, suggestions, official documents etc.)

    - Google Groups

 - Open Data

 - Semantic Web

 - access to ontology maintenance and data indexing on the targeted subjects directly from R - for researchers 

 - qualitative analysis in R



Code

-------

- for code testing (Code/ontobrowse.r file):

    - browsing functionality:

        - load Code/Code.Rproj in RStudio; it will automatically load workspace data (Code/onto2.RData)

        - type ontobrowse() in the RStudio console and follow the instructions

    -  without loading preexisting data, make sure to: ontoinstall() and ontoinitialize() first

    - the translation algorithm does not have a simple interface for now; the process steps are commented at the end of the Code/ontobrowse.r file, starting with the 1943 line: `newTranslation(sample500);`; process documented in detail in https://drive.google.com/file/d/0B1GtDMuhK2dnZlhIaEJRUkEtdFU/edit?usp=sharing (page 34), unfortunately only available in Romanian at this moment.

   

Short Description

------------------

 - 2 parts:

    - terminology maintenance and browsing, with a simple User Interface included

    - algorithm for creating a computable translation in Romanian, extendable to any other Latin language, exemplified on 500 histological concepts, based on the Levenshtein distance method.

       - the scope here was to obtain both a human and computer readable translation (see computability rules)

 

The thesis also contains:

 - suggestions for improving Terminologia Histologica from a morpho-semantic point of view.

 - issues with Terminologia Histologica as it was published at the time when the thesis was written (final version - August 2014)



Results & Methods Summary

---------------



 - 500 TH concepts 

 - I gathered 1539 terms / 2943 words from 5 Romanian medical sources by OCR and manual search, with at least 3 synonims per concept.

 - Romanian terms were searched by concept, not by words, in order to preserve histological corectness

 - terms were separated and a pool of unique words was created

 - In the first stage, matching Latin words to possible Romanian translations from my word database (by comparing only with Google Translate results), resulted in 66,23% correct matches.

 - The algorithm for matching words is based on the Levenshtein distance method.

 - I made the rest of the matches manually and obtained a pool of unique Romanian words with linguistic similarity to the Latin counterparts.

 - The resulted translation (obtained algorithmically from the above words) had __74.8% correct terms__ (exactly matching terms found in Romanian publications), while Google Translate based translations had an average of 14.5% correct terms. The percentage could have been higher, because Romanian publications do not always use the Latin structure prefered by FIPAT and usually have linking words (eliminated in my computed translation).

 - I have used base translations created with Google Translate: Latin->Romanian, English->Romanian, using both term->term translation and word->word -> recreate terms translations. 

    - En->Ro term->term - 22.2% corectness

    - En->Ro word->word -> recreate term - 13.4% corectness

    - La->Ro term->term - 11.3% corectness

    - La->Ro word->word -> recreate term - 11.1% corectness

 

 - computability rules implemented:

    - no linking words

    - only one term per concept (no synonyms for the terms)

    - only one translation for each Latin word / expression (no synonyms for words / expressions)

    - 1:1 ratio between the Latin words and the new language words for each term

    - words similar to Latin will be preferred

    - terms with fewer words will be preferred.



Links

-------



 - GitHub thesis repo: https://github.com/loredanacirstea/thRoTrans

 - R code used in my thesis: https://github.com/loredanacirstea/thRoTrans/blob/master/Code/ontobrowse.r

 - Video demos of the R package:

    - simple browse function: https://www.youtube.com/watch?v=-wN2op1u7s8

    - semi-automatic translation in a new language - user perspective: https://www.youtube.com/watch?v=-BprxBS7qiY#t=10

 - Summary (English): (in work) https://docs.google.com/document/d/1ze-SDLtB4m6dtVfnq6zqriMVXIoaGmQL2SZO3lCgL8Y/edit?usp=sharing 

 - Editable thesis (Romanian) (with versioning): https://docs.google.com/document/d/19GR2oOqCHePJ5lNXs3runAkDD-INixnhkKej4mqjtwc/edit?usp=sharing 

 - Final thesis (Romanian) (.pdf): https://drive.google.com/file/d/0B1GtDMuhK2dnZlhIaEJRUkEtdFU/edit?usp=sharing 

 - Official documents, data, intermediary work: https://drive.google.com/folderview?id=0B1GtDMuhK2dnZHNpRmhaaElibVU&usp=sharing

 - Transparent collaboration and organisational issues: https://groups.google.com/forum/#!categories/public-witness/



License

=========



Data

----



ODbL : http://opendatacommons.org/licenses/odbl/



DbCL : http://opendatacommons.org/licenses/dbcl/1-0/



Code

----



GPL v3 https://www.gnu.org/copyleft/gpl.html