5.1 Displaying aligned sentences

• CWB can encode information about sentence-level alignment between parallel corpora in its index. For each pair of source and target corpus, only a single alignment may be defined; the name of the corresponding alignment attribute (a-attribute) is a lowercase version of the CWB name of the target corpus.

• For example, the English component of the EuroParl corpus

  > EUROPARL-EN;

  is aligned to the French (EUROPARL-FR) and German (EUROPARL-DE) components inter alia.

• The available alignment attributes are listed as “Aligned Corpora:” in the output of show cd;.

• One or more alignments can be displayed in the KWIC output produced by cat. For example, in order to find out how the idiom take the biscuit can be expressed in Franch and German, we activate the corresponding a-attributes:

  > show +europarl-fr +europarl-de;
> [lemma="take"] "the" "biscuit";

  the target languages are always printed in the same order as in the show cd; output; in this example, the German translation will be shown first, followed by the French translation.

• It is recommended to set the KWIC context for the source language to a full sentence

  > set Context 1 s;

  However, translations are always displayed as complete alignment beads, which can be confusing if multiple sentences in the source language are translated into a single target sentence (a 2:1 bead) or divided in a different way in the target language (a 2:2 bead). For example, the single hit of the query

  > "price-tag";

  is translated 1:1 into French, but combined with the previous sentence in the German translation (resulting in a 2:1 bead).

• An unofficial feature allows setting the KWIC context to an a-attribute, ensuring that a complete alignment bead (for the selected target corpus) is displayed.

  > set Context europarl-de;
> cat;

  Keep in mind that this has been implemented as a special case: a-attributes cannot be used as context specifiers elsewhere (e.g. in a within clause).

• You will find that some sentences have no translation into the target language, e.g.

  > "cats" cut 6;

  Notice that the alignment KWIC context shows only the matching string without surrounding words in this case.

• In order to exclude matches outside alignment beads (i.e. without a translation), you can add a trivial alignment constraint to the query (see Sec. 5.2). The example below shows that out of 49 occurrences of cats, only 46 have a translation into French:

  > AllCats = "cats";
> size AllCats;
> GoodCats = "cats" :EUROPARL-FR [];
> size GoodCats;