Wednesday,  4  June  /09:10  –  09:40      

Terminology  as  a  Service  Indra  Samite,  Tilde  

 

TaaS  Workshop  2014  4  June,  Dublin  (Ireland)  

The  research  within  the  project  TaaS  leading  to  these  results  has  received  funding  from  the  European  Union  Seventh  Framework  Programme  (FP7/2007-­‐2013),  grant  agreement  no  296312  

 

 

TAUS  TaaS  Workshop    

Dublin  /  04.06.2014.  

* What  does  a  terminologist  do?  

____________________________________  TAUS  TaaS  Workshop  |  Dublin  |  June  4,  2014  

embracing  innovation  

Terminology  field  

SaaS  Software  as  a  Service  

PaaS  Platform  as  a  Service  

TaaS  Terminology  as  a  Service  

*  Tilde          Latvia  (Coordinator)  

*  TAUS        Netherlands  

*  Kilgray        Hungary  

*  University  of  Cologne      Germany  

*  University  of  Sheffield    UK  

TaaS  Partners  

*  Industry  &  Research  Collaboration  project    *  Supported  by  EU  7th  R&I  Framework  Programme  *  Resulted  in  TaaS  cloud-­‐based  services    *  Accessible  for  free  at  the  online  portal  

 

termunity.com  

TaaS  

*  Ioannis  Iakovidis,  Interverbum  Technology  *  Uwe  Muegge  &  Carl  Yao,  CSOFT  International  *  Luigi  Muzii,  sQuid  *  Maria  Pia  Montoro,  Intrasoft  International,  Terminology  Blogger  at  Wor  lLo  

 

Invited  speakers  

____________________________________  TAUS  TaaS  Workshop  |  Dublin  |  June  4,  2014  

*  Little  survey  for  warming-­‐up  * Panel  of  speakers  * Reflections  from  the  audience  

Discussion  from  11:55  

____________________________________  TAUS  TaaS  Workshop  |  Dublin  |  June  4,  2014  

Welcome  to  the  Cloud!  Terminology  as  a  Service  

 Andrejs  Vasiļjevs,  Indra  Sāmīte  

Tilde  TAUS  TaaS  Workshop  /  Dublin  /  04.06.2014.  

*  Language  technology  developer  *  Translation  and  Terminology  systems  *  >350  000  users  

*  Localization  service  provider  *  Leadership  in  smaller  languages  *  Offices  in  Riga  (Latvia),  Tallinn  (Estonia)  and  Vilnius  

(Lithuania)  *  130  employees  *  Strong  R&D  team  *  5  PhDs  +  and  engineers  and  students,  80+  research  papers  *  Coordinator  of  several  EU  industry-­‐academic  R&D  projects  

About  Tilde  

EuroTermBank  Portal  

Microsoft  Language  Portal  

ECDC  Terminology  Server  

*  Term  identification  in  the  source  text  *  Consulting  online  databases  and  local  files  for  translation  

equivalents    *  Creating  and  maintaining  terminology  glossaries    *  Sharing  term  glossaries  and  involving  others  in  their  polishing  *  Structuring  data  in  the  industry  standard  formats  *  Integrating  term  glossaries  in  CAT  and  other  productivity  tools  *  Keeping  terminology  up  to  date    *  etc.  

Complexity  of  terminology  work  

TaaS  User  Needs  Survey  Results:  Importance  of  terminology  work  

43.5%  

39.9%  

14.8%  1.8%  

Very  important  

Quite  important  

Less  important  

Not  important  

TaaS  User  Needs  Survey:    willingness  to  share  

24.9%  

19.2%  

14.2%  

11.4%  

7.6%  

6.0%  

16.7%  

Yes,  provided  that…  

Joint  contribution  to  the  DB    Access  control  Legal  aspects  External  quality  control  Little  effort  Anonymity  Other  

48.6%  

22.0%  

16.5%  

8.3%   4.6%  

No,  because…  

Legal  restrictions  Poor  quality/Lack  of  time  Own  asset  Risk  of  misunderstanding  

60.5%   39.5%  

*  Simplify  the  process  for  language  workers  to  prepare,  store  and  share  of  task-­‐specific  multilingual  term  glossaries  

*  Provide  instant  access  to  term  translation  equivalents  and  translation  candidates  for  professional  translators  

*  Improve  quality  of  machine  translation  systems  by  dynamic  integration  of  terminology  data  

TaaS  Mission  

cloud-­‐based  platform  that    automates    

the  terminology  work    for  human  and  machine  use  

 

termunity.com  

*  Automatic  extraction  of  monolingual  term  candidates    from  user  uploaded  documents    

*  Automatic  retrieval  of  translation  equivalents    from  different  public  and  industry  terminology  databases  

*  Translation  candidate  acquisition    from  multilingual  web  data  

*  Facilities  for  cleaning-­‐up    by  users  automatically  acquired  terminological  data;  

*  Data  sharing  and  integration  facilities    through  APIs  and  export  tools  

Key  services  of  TaaS  

TaaS  Services  

Term  identification  and  annotation  

*  Support  for  industry  standard  formats  *  Integration  into  CAT  and  productivity  tools  *  API  to  integrate  TaaS  services  into  various  software  applications  

Integration  

TaaS  in  the  service  for  MT  

Online  Terminology  Services

Translation

Training

SMT  System  Training  and  adaptation

Online  Translation  Service

Input  Text  for  Translation

Parallel  corpus

Monolingual  corpus

Bilingual  term  collections

Monolingual  Term  

Extraction

Trained  SMT  Model

Bilingual  Term  

Extraction

Translated  Text

TaaS  Architecture  

Presentation  Layer

Web  Page  UI Public  API

Application  Logic  LayerTerminologycollection  

management

User  management  

Terminologycollection  search

Terminology  collection  creation

Data  Storage  Layer(Shared  Term  Repository)

High-­‐performance  Computing  (HPC)  Cluster

SGE

External  TDBsCAT  tools MT

https

REST

http/https

html

https

REST

https

REST

includ

ed

CPUCPU

includ

ed

 

Shared  Term  Repository

DB

File  Store

Web  Browsers

HPC  frontend

CPU

CPUCPU CPU

CPUCPU CPU

Term  extraction  workflowsFull  collection  

creation  workflow

Monolingual  collection  creation

Translation  candidateextraction

....

Modules

Result  processing

Collection  Importer

Marked  Text  enrichment

Text  tagging  

with  terms

Statistical  DB  acquisition

Statistical  DB  feeding

Bilingual  Term  Extraction  System

Parameter  retriever

Translation  lookupETB  &  STR

IATETAUS  API

Statistical  DBCollection  merger

CPUCPU CPU

Term  extractionTXT  extractor

TWSCKilgray  TermExtractor

Collection  creatorTerm  normalizer

 

Statistical  DB

Research  

Development  

Usage  

 Focus  areas  

*  Term  extraction  *  Collection  of  domain  specific  

multilingual  corpora  *  Max(FTC)    

*  Usability  *  Outreach  *  Sustainability  

*  Quality  *  Performance  *  Scalability  *  Interoperability  

Indra  Sāmīte  Business  Development  Director  

Tilde  

TaaS  in  action  

The  Modern  Translator  

*  Search  for  individual  terms  in  various  sources  *  Identify  term  candidates  in  your  documents  and  extract  them  automatically  *  Automated  Look  up  translation  candidates  in  various  sources  *  Refine  and  approve  terms  and  their  translations  *  Share  your  terminology  with  other  users  *  Collaborate  with  colleagues  &  team  *  Use  your  terminology  in  other  working  environments  

Features  

Simple  Search  

 *  Search  for  terms  in  various  sources  

Simple  Search  

Identify  &  extract  

     

Identify  &  extract  

Automated  Lookup  

Automated  Lookup  

Refine  &  Approve  

Refine  &  Approve  

Refine  &  Approve  

Share  &  Collaborate  

CAT  integrated  

Machine  translation  by  Tilde  

*  In  the  cloud  *  Do  it  yourself  or  Custom  *  CAT  integrated  *  Terminology  ready  *  Vast  data  base  of  resources  for  training  *  Productivity  boosting  

LetsMT  

MT  friendly  

* Trusted  terminology  resources  * Sharing  new  terminology  data  * Reuse  of  terminology  resources  * Quality  improvement  of  MT  systems  * Efficient  work  patterns  * Increase  competitiveness  * Translation  quality  improvement    

Strategic  impact  

* Free  access  to  online  services    * Integrated  with  memoQ  2014  * Integrated  with    OmegaT  (July’14)  

Sign  up  now!  termunity.com  

Thank  You!    

The  research  within  the  project  TaaS  leading  to  these  results  has  received  funding  from    the  European  Union  Seventh  Framework  Programme  (FP7/2007-­‐2013),  grant  agreement  n°  296312  

Contact:  Indra.Samite@tilde.com  

termunity.com    @TermServ  on  Twitter  

Terminology  Services  group  on  LinkedIn