Sketch Engine KB - How to search for direct speech in a corpus?

If the corpus is annotated for direct speech, the searching is easy using the relevant structure and metadata. Most corpora in Sketch Engine are not, however, annotated like this. Therefore, you have to make use of clever use of tokens, words and attributes to identify direct speech.

The concordance ADVANCED tab should be used with the CQL option.

Useful tips

If you are new to CQL, start here: https://www.youtube.com/playlist?list=PLPVgp0Xm6hWBLZzD7haCVbhC9pSfWZk1S or here: https://www.sketchengine.eu/documentation/corpus-querying/
Use the CQL builder to avoid problems with query syntax
Punctuation must be escaped https://www.sketchengine.eu/my_keywords/escaping/

Examples

Here are some simple examples and more complex ones too.

Any text between quotes that is at least 5 tokens long

[lemma="\""] []{5,} [lemma="\""]

It will, however, find any two quotes, even quotes not belonging to the same direct speech such the quote before 'Go' and the quote at the very end:

She said, "Go home," and then she said , "And do not come back!"

To avoid this, the tokens in between must not be a quote:

[word="\""] [word!="\""]{5,} [word="\""]

To find direct speech that is inside one sentence (does not extend across several sentences):

[word="\""] [word!="\""]{5,} [word="\""] within <s/>

To find sentences containing direct speech, use:

<s/> containing [word="\""] [word!="\""]{5,} [word="\""]

You may also apply a filter on the result to only results that include a reporting verb (said, asked, replied, etc.) before or after the result. You can include the reporting verb into the CQL directly too, for example:

[word="said|replied|asked"] [word="\,"] [word="\""] [word!="\""]{5,} [word="\""] within <s/>