mordor | Organize learning materials | User Interface library

Try this way,

Welcome to SO, Père Noël. Pacakge {sentimenter}'s get_sentences() breaks the text input into sentences by default, as its name implies. To reconstruct the original text input as the defining key in your final data frame, you need to group and summarize the sentence-based output produced by sentiment(). In this example, I will simply average the sentiment scores, and append sentences by their element_id.

sort is an external command (Application) on Linux systems.
In other words, do not use the short name (sort) but the full cmdlet name Sort-Object:

Something like this should do it:

The block needs to reside inside the block.

From the comments, it seems you are using the very ancient Saxon6 product, which only supports XSLT 1.0. More recent versions (the current is 10.3) implement XSLT 3.0.

When your stylesheet specifies version="3.0" and your XSLT processor only supports 1.0, then it runs in "forwards compatibility mode". In this mode, elements and attributes that weren't defined in the 1.0 specification are ignored. One such element is xsl:mode, so your stylesheet runs as if the xsl:mode declaration were not there. This means that instead of shallow-copy, you get the default "no match" template behaviour, which outputs the text nodes of the source document and ignores the element nodes.

XML must have exactly one root node. Also without knowing the structure in the data I would assume you want to get all notes regardless how deep they are nested. To achieve that use collection($col)//note

You have two … tags in this XML file. The XML specification states that you may only have one root element.

To fix this, remove the:

You are confusing yourself. You have your solution in edges. There isn't a reason to read the data a second time. For example, you can simply output sorted/unique elements of edges, e.g. the modifications to your code are: