Number sentences in paragraphs, as a range

Question

I need to typeset a critical edition of a small philosophical work.

Editorial and typesetting rules require to number sentences at the beginning of each paragraph.

That is easy for 1-sentence paragraphs. But for multi-sentence paragraphs I need to print the range.

How to incrementally number sentences in paragraphs as a range, like the sequential numbers in square brackets shown below?

Example

---

Talk on the Practical Reality

[1] Reality is at once a simple and a complicated word.

[2-3] For an ordinary, unaspiring human being, reality is a far cry. At every moment life offers him frustration.

[4-6] His life is the life of desire. When his desires are not fulfilled, he is frustrated. And when his desires are fulfilled, still he feels no abiding satisfaction; therefore, according to him everything is unreal.

(etc.)

---

MWE

I have a MWE, which only number paragraphs:

\documentclass{book}

\newcounter{para}
\newcommand\mypara{\par[\refstepcounter{para}\thepara] \space}

\begin{document}

\mypara (This should be [1]) Reality is at once a simple and a complicated word.

\mypara (This should be [2-3]) For an ordinary, unaspiring human being, reality is a far cry.  At every moment life offers him frustration.

\mypara (this should be [4-6]. His life is the life of desire.  When his desires are not fulfilled, he is frustrated. And when his desires are fulfilled, still he feels no abiding satisfaction; therefore, according to him everything is unreal.

\end{document}

Answers and packages (but insufficient)

I am aware of these answers:

• Numbering the sentences in a text

• Is it possible to automatically enumerate sentences in LaTeX?

• Also, package scrjura

But they do not seem to help my need, to number the range of sentences in paragraphs.

• This answer may help, but with a quite complex macro:

How to count the number of words in a sentence, the number of sentences in a paragraph, and the number of paragraph?

Question

Is there a package, or a simpler macro, to help incrementally number sentences in paragraphs, as a range?

Thanks in advance.

Answer 1

Here's a re-implementation of yannisl's answer which does not depend on Lua (so should work with any engine), avoids the need to start and end the environment for each paragraph and eliminates the need to explicitly capture the content of the paragraphs using the b specifier. Instead, the code is automatically added to the start and end of paragraphs.

Which is all well and good, but it is not for free. Everything has its price. In this case, the price is dependence on multiple compilations. You need at least two and sometimes three compilations for the numbering to be correct. This is because the numbers are written at the start of the paragraph, when only the first sentence's number is known for this run. To get the number of the last sentence, we record its value at the end of the paragraph and read it back on the next run. So, initially, the final sentences of all paragraphs will be zero and you will see things like '9-0' in the output. That is expected.

An alternative would be to capture the paragraph, but there are some downsides to doing that, so I decided not to complicate things.

\documentclass{article}
% ateb: https://tex.stackexchange.com/a/719068/ addaswyd o ateb yannisl: https://tex.stackexchange.com/a/719024/

\ExplSyntaxOn
\int_new:N \g_gsl_sentences_int
\int_new:N \g_gsl_pars_int
\property_new:nnnn { sinpar }{ now }{ 0 }{ \int_use:N \g_gsl_sentences_int }
\cs_new_protected:Npn \sn #1 \n {
  \int_gincr:N \g_gsl_sentences_int
  #1 ~
}

\NewDocumentEnvironment{segmentation} {}
{
  \hook_gput_code:nnn { para/begin } { gsl }
  {
    \int_gincr:N \g_gsl_pars_int
    \hbox_gset:Nn \g_para_indent_box { [\int_to_arabic:n {\g_gsl_sentences_int} -- \property_ref:ee { gsl@para@\int_to_arabic:n {\g_gsl_pars_int} } {sinpar} ] ~ }
  }
  \hook_gput_code:nnn { para/end } { gsl }
  {
    \property_record:ee { gsl@para@\int_to_arabic:n {\g_gsl_pars_int} } { sinpar } 
  }
}{
  \para_end:
  \hook_gremove_code:nn { para / begin } { gsl }
  \hook_gremove_code:nn { para / end } { gsl }
}

\ExplSyntaxOff
\begin{document}
\parindent0pt
\begin{segmentation}
  \sn Myfirst sentence.\n
  \sn My second sentence.\n
  \sn My third sentence.\n 

  \sn My fourth sentence.\n 
  \sn My fifth sentence.\n
  \sn His life is the life of desire.\n 
  \sn My fifth sentence.\n
  \sn His life is the life of desire.\n 

  \sn This is the next in line!\n
  \sn This is the next in line!\n
\end{segmentation}

Here's some regular text in sentences which aren't marked and in paragraphs which don't get the special treatment.

And here's another.
These sentences aren't marked or counted either.
That's because we finished the \texttt{segmentation} environment and haven't started another.

\begin{segmentation}
  \sn Here we go again.\n
  \sn Counting sentences one-by-one.\n
  
  \sn However, we don't have to add the environment explicitly.\n
  \sn So it is a bit more automatic.\n
\end{segmentation}
\thispagestyle{empty}
\end{document}

Note: the truth is, this was going to depend on Lua because I planned to just add the hooks wrapper around yannisl's code. But I know nothing about Lua and couldn't even figure out how to do that.

EDIT

This version offers ranges and/or sentence numbers on a per-environment or general basis and provides a key-value interface for customising the appearance of both.

\glsset{<list of key-values>} or an optional argument to the segmentation environment may be used for the customisation. All settings are local.

• sentence numbers=<true|false> determines whether numbers are printed.
  • Default is true; initially false.
• sentence ranges=<true|false> determines whether ranges are printed.
  • Default is true; initially true.
• sentence number format={<code taking one argument>} determines the format of sentence numbers, where #1 is the sentence number.
• sentence range format={<code taking two arguments>} determines the format of ranges, where #1 is the first sentence in the paragraph and #2 is the second.

Note that this will obviously be slightly slower than the less flexible code above.

\documentclass{article}
% ateb: https://tex.stackexchange.com/a/719068/ addaswyd o ateb yannisl: https://tex.stackexchange.com/a/719024/
\usepackage{xcolor}

\ExplSyntaxOn
\int_new:N \g_gsl_sentences_int
\int_new:N \g_gsl_pars_int
\property_new:nnnn { sinpar }{ now }{ 0 }{ \int_use:N \g_gsl_sentences_int }

\cs_new_protected_nopar:Nn \gsl_sentence_no_format:n
{
  \textsubscript{ \textcolor{gray}{ #1 } }
}
\cs_set_protected_nopar:Nn \gsl_sentence_range_format:nn
{
  [ #1 -- #2 ]
}

\keys_define:nn { gsl }
{
  sentence ~ number ~ format .code:n = {
    \cs_set_protected_nopar:Nn \gsl_sentence_no_format:n { #1 }
  },
  sentence ~ numbers .bool_set:N = \g_gsl_sentence_nos_bool,
  sentence ~ numbers .initial:n = false,
  sentence ~ numbers .default:n = true,
  sentence ~ range ~ format .code:n = {
    \cs_set_protected_nopar:Nn \gsl_sentence_range_format:nn { #1 }
  },
  sentence ~ ranges .bool_set:N = \g_gsl_sentence_ranges_bool,
  sentence ~ ranges .initial:n = true,
  sentence ~ ranges .default:n = true,
}

\cs_new_protected:Npn \sn #1 \n {
  \int_gincr:N \g_gsl_sentences_int
  #1 \bool_if:NT \g_gsl_sentence_nos_bool 
  {
    \gsl_sentence_no_format:n { \int_to_arabic:n {\g_gsl_sentences_int} } 
  } ~
}

\NewDocumentEnvironment{segmentation} { o }
{
  \IfValueT { #1 } 
  {
    \keys_set:nn { gsl } { #1 }
  }
  \bool_if:NT \g_gsl_sentence_ranges_bool
  {
    \hook_gput_code:nnn { para/begin } { gsl }
    {
      \int_gincr:N \g_gsl_pars_int
%       \hbox_gset:Nn \g_para_indent_box { [\int_to_arabic:n {\g_gsl_sentences_int} -- \property_ref:ee { gsl@para@\int_to_arabic:n {\g_gsl_pars_int} } {sinpar} ] ~ }
      \hbox_gset:Nn \g_para_indent_box { \gsl_sentence_range_format:nn { \int_to_arabic:n {\g_gsl_sentences_int} } { \property_ref:ee { gsl@para@\int_to_arabic:n {\g_gsl_pars_int} } {sinpar} } ~ }
    }
    \hook_gput_code:nnn { para/end } { gsl }
    {
      \property_record:ee { gsl@para@\int_to_arabic:n {\g_gsl_pars_int} } { sinpar } 
    }
  }
}{
  \para_end:
  \bool_if:NT \g_gsl_sentence_ranges_bool
  {
    \hook_gremove_code:nn { para / begin } { gsl }
    \hook_gremove_code:nn { para / end } { gsl }
  }
}

\NewDocumentCommand \gslset { m }
{
  \keys_set:nn { gsl } { #1 }
}

\ExplSyntaxOff
\pagestyle{empty}
\begin{document}
\begin{segmentation}
  \sn My first sentence.\n
  \sn My second sentence.\n
  \sn My third sentence.\n 

  \sn My fourth sentence.\n 
  \sn My fifth sentence.\n
  \sn His life is the life of desire.\n 
  \sn My fifth sentence.\n
  \sn His life is the life of desire.\n 

  \sn This is the next in line!\n
  \sn This is the next in line!\n
\end{segmentation}

Here's some regular text in sentences which aren't marked and in paragraphs which don't get the special treatment.

And here's another.
These sentences aren't marked or counted either.
That's because we finished the \texttt{segmentation} environment and haven't started another.

\begin{segmentation}
  \sn Here we go again.\n
  \sn Counting sentences one-by-one.\n
  
  \sn However, we don't have to add the environment explicitly.\n
  \sn So it is a bit more automatic.\n
\end{segmentation}

\begin{segmentation}[sentence ranges=false]
  \sn My first sentence.\n
  \sn My second sentence.\n
  \sn My third sentence.\n 

  \sn My fourth sentence.\n 
  \sn My fifth sentence.\n
  \sn His life is the life of desire.\n 
  \sn My fifth sentence.\n
  \sn His life is the life of desire.\n 

  \sn This is the next in line!\n
  \sn This is the next in line!\n
\end{segmentation}

Here's some regular text in sentences which aren't marked and in paragraphs which don't get the special treatment.

And here's another.
These sentences aren't marked or counted either.
That's because we finished the \texttt{segmentation} environment and haven't started another.

\begin{segmentation}[sentence numbers]
  \sn Here we go again.\n
  \sn Counting sentences one-by-one.\n
  
  \sn However, we don't have to add the environment explicitly.\n
  \sn So it is a bit more automatic.\n
\end{segmentation}

\gslset{%
  sentence ranges=true,
  sentence numbers=true,
  sentence range format={\textit{\textcolor{blue}{\small (#1--#2)}}},
  sentence number format={\textsuperscript{\textcolor{red}{#1}}},
}%

\begin{segmentation}
  \sn My first sentence.\n
  \sn My second sentence.\n
  \sn My third sentence.\n 

  \sn My fourth sentence.\n 
  \sn My fifth sentence.\n
  \sn His life is the life of desire.\n 
  \sn My fifth sentence.\n
  \sn His life is the life of desire.\n 

  \sn This is the next in line!\n
  \sn This is the next in line!\n
\end{segmentation}

Here's some regular text in sentences which aren't marked and in paragraphs which don't get the special treatment.

And here's another.
These sentences aren't marked or counted either.
That's because we finished the \texttt{segmentation} environment and haven't started another.

\begin{segmentation}
  \sn Here we go again.\n
  \sn Counting sentences one-by-one.\n
  
  \sn However, we don't have to add the environment explicitly.\n
  \sn So it is a bit more automatic.\n
\end{segmentation}
\end{document}

The blue and red is hideous, but illustrates the idea.

Answer 2

Run this with LuaLaTeX. Sentence segmentation is a major topic in Natural Language Processing, so you will have to mark the beginning of a sentence with \sn and the end with stop . If a sentence ends say with a ; or ! you will still need to add the stop afterwards as it uses a delimited macro. This shouldn't be too cumbersome, and if you generating these through a machine learning program, you can arrange for this to be added automatically. Paragraphs are marked with the environment segmentation.

Tips! You can change the macro delimiter from . to another symbol depending on your use case. David Carlisle or another user might be able to change the macro to capture an eol character automatically thus automating the delimiter. (I dont' remember how to do it). The code will fail for RTL scripts such as Hebrew, Arabic for other scripts, including east Asian scripts, such as Japanese or Chinese as long as you mark the sentences, it shouldn't give you any issues and the reason why I used Lua, as is more versatile when it comes to using non-Latin scripts.

You can change the tex.print(str3) to (str1) or (str2) for differnt styles of marking the sentences.

\documentclass{article}
\usepackage{fontspec}
\usepackage{luacode}

\begin{luacode}
  M=M or {}
  M.data ={}
  M.last =1
  M.initial=1
  M.insertdata=function(x)
    table.insert(M.data,x)
    M.last=M.last + 1
    end
   M.clear=function()
     for k=1,#M.data do
     table.remove(M.data,1)
     end
   end   
\end{luacode}

\long\def\sn#1\n{%
\directlua{
 M.insertdata('#1')
}}

\long\def\printparagraph{{%
\parindent=0pt
%tex.print("\\itshape "..v,"["..(k+initial-1).."]".." ")
\directlua{
  local data = M.data
  local initial = M.initial
  local current_length= table.getn(data)+initial
  tex.print("[".. initial.. "-"..(current_length-1).."]")
  for k,v in pairs(data) do
  local str  = "\\itshape "..v.." "
  local str1 = "\\itshape "..v,"["..(k+initial-1).."]".." "
  local str2 =  "\\itshape "..v,"["..(k+initial-1).."]".." "
  local str3  = "\\itshape "..v.."\\textsubscript{"..(k+initial-1).."}"
  tex.print(str3)
  M.initial=current_length
end   
}}\par}

 
\def\mypara{
 ------------------------------\par
\directlua{
  M.clear()
}}

\NewDocumentEnvironment{segmentation}{+b}
{\mypara #1\printparagraph}
{}
\begin{document}
\parindent0pt
\begin{segmentation}
\sn Myfirst sentence.\n
\sn My second sentence.\n
\sn My third sentence.\n 
\end{segmentation}
\begin{segmentation}
\sn My fourth sentence.\n 
\sn My fifth sentence.\n
\sn His life is the life of desire.\n 
\sn My fifth sentence.\n
\sn His life is the life of desire.\n 
\end{segmentation}
\begin{segmentation}
\sn This is the next in line!\n
\sn This is the next in line!\n
\end{segmentation}
\end{document}