Sketchengine TOOL  FOR  T EXT-­‐BA SED  T ERMINOLOGY

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Using  texts  for  term  mining v Building  a  corpus. vChoosing   texts vConverting   into  common   format  (txt) vAnnotation v Croatian:  http://nlp.ffzg.hr/api-­‐for-­‐our-­‐language-­‐technologies/

vAlignment v CAT  tools  (SDL,  memoQ)  or  LF  ALigner,  https://sourceforge.net/p/aligner/wiki/Home/

v Searching  the  corpus. v Concordance  tools:  AntConc (free),   Wordsmith   (€),  ParaConc (free) v Web-­‐based  corpus   workbench:  Sketchengine,   http://www.sketchengine.co.uk

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

What  is  the  Sketchengine? v Very  powerful  corpus  workbench:  https://www.sketchengine.co.uk/ v Provides  access  to  multiple  pre-­‐compiled  corpora  (British  National   Corpus,  hrWaC,  DGT  corpora  and  many  more) v NOT  free,  but  not  expensiveJ (5,99  € per  month) v Allows  the  creation  of  ad  hoc  corpora  from  web  texts v Supports  TMX  import  (for  bilingual  texts!) v Provides  ways  to  extract  terminology  semi-­‐automatically v Online  tutorials:  https://www.sketchengine.co.uk/sketch-­‐engine-­‐ video-­‐tutorials/

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Simple  concordances

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Other  query  types v simple:  searches  for  word  and  its  inflected  forms v lemma:  searches  for  all  words  with  this  lemma v phrase:  for  searching  multiple  words v word:  to  search  for  a  specific  wordform v character:  to  search  for  a  string  of  characters v CQL:  corpus  query  language

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

WordSketches

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Thesaurus   – similar  words

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Keyword  extraction

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Term  queries  in  the  DGT   parallel  corpus v simple  queries:  ribolov,  brancin,  grdobina v lemma  queries:  ribolov -­‐>  ribolova,  ribolovu,  ribolov v parallel  query:  

v querying  using  CQL  syntax:  

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Basic  CQL v Typical  format:  [attribute="value"],  e.g.  [lemma=“riba”] v Specifying  word  class  or  case:  [tag=“N.*”]  (any  noun),  [tag=“A.*”]  (any   adjective) v Regular  expressions:   v .  (dot)   matches  any  single  character v *  (asterisk)  matches  0-­‐100  repetitions v +  (plus)   matches  1-­‐100  repetitions v {n,k}  specifies  exact  range  of  repetitions,   from   n  to  k

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

[lemma=“rad”]

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

[tag=“A.*”][lemma=“riba”]

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

"ulov.*"  []{0,3}   [tag="N.*"]

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Challenges v Search  for  verbs  occurring  before  the  word  “ugovor”  with  up  to  2   words  in  between. v Search  for  words  ending  with  “anje”. v Search  for  defining  contexts  containing  a  noun  in  the  nominative  case   followed  by  “je”  followed  by  an  adjective  and  noun  in  the  nominative   case.

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Looking  for  definitions v Exploit  typical  definition  patterns:   v[X]  is  a  [Y] v [X]  is  defined  as  [Y] v [X]  is  a  kind  of  [Y] v …

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

WebBootCat v Tool  to  create  text  collections  from  web  pages v User  provides  keywords  &  optionally  selects  sites  to  crawl v When  the  corpus  is  compiled  it  can  be  used  for  queries  or  download.

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

TMX  Upload v Allows  you  to  create  corpora  from  your  translation  memories

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Terminology  extraction v Works  for  languages  with  a  predefined  “term  grammar” v Manage  corpus  -­‐>  Keywords  and  terms v Terms  can  be  exported  into  TBX  or  CSV

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

Exercise v Use  the  corpus-­‐derived  information  on  the  following  slides  to  create  a   term  entry  for  “bluetongue”.

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB

INTEGRA   TERMINOLOGY  M ANAGEMENT,  ZAGREB