SWIKOweb

Search Field Instructions

The search field allows for any type of simple plain or Regex text and automatically detects whether you are performing a text search or a multi-level Exmaralda search. The more complex your search query, the longer it will take to display the results.

You can control which field you are searching into using the "Search into" dropdown menu, and can control which text is shown using the "output from" dropdown menu. The 'token' level covers most files of the database (across different projects), i.e. any which was simply transcribed. Corrected text and lemma search at the ctok- and lemma-level of the Exmaralda files, so covers all SWIKO files and only a handful of files from other projects that were annotated that far; and Target hypothesis only covers the German SWIKO files which were error annotated.

The display format will control what tool you use and how the results are presented: file list is default and now also includes the concordances embedded in the table; xlsx downloads the results as an Excel table; and frequency distribution / bigram / trigram were options integrated from the previously separate 'tools' menu.

Finally, you can select whether the search is case sensitive or not.

Search for a single word:

Essen
into:tokens       out:tokens    display:file list     case sensitive on

Finds all instances of the exact word form essen in the original text. Displays concordances from original text.


.*essen
Into:tokens       out:tokens       display:file list     case sensitive off

Finds all instances of words ending in essen (essen, but also mittagessen, Lieblingsessen, etc.), with .* as a wild card for any number of characters.


die
Into:lemma        out:tokens        display:file list

Finds all instances of words whose lemma is 'die'. Displays original texts.


h.nd
Into:tokens         out:tokens      display:file list

All instances of words with wildcard at the period (Hund, Hand,...).

Use # to search into any Exmaralda level.
If you specify the level, it does not matter which “input” dropdown menu option you select; it ignores that filter since you have just specified the input level.

These are the abbreviations for the levels that we use:

#tok - original text (with errors)
#ctok - orthographically correct text
#th1 - target hypothesis 1 (German texts only)
#lem - lemma
#coPOS - part-of-speech tags across German, French, and English
#lgPOS - language-specific POS (e.g., STTS for German)
#tag - error and non-target tags
#o - orthography errors
#g - grammar errors

These are the orthography and error tags that we use:

O_cap - capitalization
- subcategories: _noun or _bos (beginning of sentence) or _oth
O_graph - grapheme
- subcategories: _add or _del or _cha or _mul
O_wdbd - word bound errors
- subcategories: _mg or _sp (merge or split)
G_add - word erroneously added
G_del - word missing
G_cha - word changed – subcategories:
- wch: wrong word choice of same part-of-speech
- cha: word had to be modified
- POS: wrong part of speech chosen
G_mov - wrong position
G_wordorder - sentences with at least one position error

For example:

#lgPOS=NN
Into: any		Out: corrected tokens	display: file list

All instances of tokens tagged as nouns.


#o=.*
Into: any		Out: tokens			display: file list

Any type of orthography error.


#g=G_NN_cha	(or alternatively #g=.*NN_cha)
Into: any		Out: tokens			display: file list

Any grammar error where a noun was modified (i.e., inflection errors such as singular vs. plural or wrong case)


#g=G_POS_APPR.*
Into: any		Out: tokens			display: file list

Any grammar error where a preposition was erroneously chosen instead of a word of another part-of-speech.


Use WITH to add search information to the same token:
die WITH #g=.*
Into:lemma        out:tokens        display:file list

Finds all instances of words whose lemma is die, and who have any grammar tag.


#lem=die WITH #g=.*
Into:tokens		out:tok		display:file list

This performs the same query as above (when you specify the level in the search bar, it ignores which “into” dropdown you select)


das WITH #ctok=dass
into token		out:ctok		display: file list

Finds all instances of the word das that was corrected to the word dass, and displays the corrected texts.

Search for chunks (and combine with either specifying a level or WITH function)

nicht einverstanden
Into: corrected tokens	out: tokens	display: file list

Searches for all instances of “nicht einverstanden” in the corrected text.


nicht #lgPOS=ADJ.*
Into: tokens 		out: token 		display: file list

Searches for all instances of the word nicht followed by any adjective (e.g., nicht gut, nicht sauber etc.)


kein.* WITH #g=.*cha #lgPOS=NN
Into: corrected token		out: token	display: file list

Searches for all instances of the corrected word kein and its variants (keine, keiner, keinem,...) which contain a grammatical error; followed by a noun.

Use other tools to display the results (change the display drop-down menu) and specify which texts should be searched through via additional filters:

food
Into:lemma        out:lemma         display:frequency

Displays a frequency count of lemmas of texts containing the word food.


Leave search empty
into: corrected token		out: corrected token		display: bigram
Filter Task > Task Number: SWI02
Filter Language > language of schooling: French		language of text: German

Displays a frequency count of bigrams of orthographically corrected SWI02 DAF texts.