-
Notifications
You must be signed in to change notification settings - Fork 4
LtgOslo_LaTeX
People, even within our homogeneous research group, differ in their aesthetic judgment, specifically when it comes to typography. When writing collaboratively, it can be beneficial to rely on a set of shared assumptions and conventions, to help synchronize LaTeX sources from multiple authors. Following are some suggestions for best practices that at least some LTG members have found useful in the past.
For maximum compatibility across platforms, use Unix-style newlines and the UTF-8 character encoding. The following may serve as a good template header for LaTeX (where the specifics of the document class and default font size are subject to adjustments, naturally):
% -*- coding: utf-8; -*-
\documentclass[11pt]{article}
\usepackage[T1]{fontenc}
\usepackage[utf8]{inputenc}
\usepackage{apacite}
Limit lines to at most 79 characters. In emacs, use auto-fill mode and set fill-column to 79, in vim set textwidth to 79.
When working on a shared set of LaTeX sources, avoid unnecessary, purely cosmetic edits. For example, to follow the above recommendation it may be tempting to re-fill entire paragraphs, i.e. after inserting a word somewhere make the editor reaarange line breaks (which have no significance to LaTeX in running text). Such re-filling will typically lead to a seeminly large revision in SVN, as it effectively changes around a large number of lines in the source file (for no logical change). One technique (advocated by at least some LTG folks) to mitigate this issue can be starting each sentence on a line of its own in the LaTeX sources, e.g.
Wikipedia is met with great interest by researchers in (computational)
linguistics.
It provides a massive and relatively high-quality collection of text
and (predominantly unstructured) encyclopedic knowledge.
Following are a few more general LaTeX recommendations (with more to come, surely):
-
For emphasis in running text, use the \emph{} macro rather than the deprecated \em, e.g. We define \emph{coverage} as the ratio of ...; the benefit of \emph{} over \em is that it automatically inserts what is know as italic correction, i.e. what would otherwise have to be typeset as {\em coverage\/}.
-
Similarly, use the macros \textit{}, \textsc{}, and \textbf{} instead of the deprecated \it etc. macros.
-
Avoid the use of bold face in running text; bold can be suitable in the keys of itemized definitions or table headings, but it should not be used in running text for emphasis.
-
Avoid simple double quotes ("); use LaTeX-style opening and closing sequences of back and single quotes, respectively, instead.
There is a wide variety of macro packages for linguistic data, and it is important that we make a consistent selection throughout the book, both for aesthetic and technical reasons.
Feature structures (including f-structures), should be typeset using the avm.sty package by Chris Manning; the package provides reasonably extensive on-line documentation. By default, all \avmoptions{} are disabled; thus, in case you wanted to use active or centered mode, say, you should place a command like
\avmoptions{active,centered}
at the start of your own file chapter.tex. Note that active and centered are enabled by default in the derived package avm+.sty by Walt Detmar Meurers, so in case you are used to this package, make sure to add the right \avmoptions{} to the preamble of you chapter source file. Please note that your feature structures should not contain font selection instructions, as these are determined globally for the book as a whole; accordingly, please always typeset feature names in all uppercase letters (unlike the examples given in the documentation for this package, which assumes that the font used for features is set to small caps). For in-text occurences of AVM elements, the LOGON book setup augments the AVM package by commands \avmsort{}, \avmfeature{}, and \avmvalue{}, which each take one argument and typeset it using the appropriate font. For example:
In \begin{avm}\[ \asort{foo} BAR & baz \]\end{avm}, the type of the structure is
\avmsort{foo}, and it has one feature \avmfeature{BAR} with value \avmvalue{baz}.
For the construction of trees, there are several options, and your choice may vary according to parameters like the complexity of individual nodes and your requirements on fine-grained control over node placement, labeling of arcs, et al. Our default suggestion for tree drawing is the QTree package, which strikes a good balance in ease of use and control options; QTree comes with good documentation, including many examples. Note that QTree is an extension of the earlier QobiTree package, and if you prefer the original (somewhat more verbose) syntax of QobiTree, that is available too. For TikZ users, there is a drop-in replacement for qtree, tikz-qtree, that draws trees using TikZ and supports some additional TikZ features.
A fairly traditional tree drawing package that we do not want to include is tree-dvips.sty by Emma Pease (of CSLI). Its main limitations are the relatively crude node placement and exclusive support for PostScript generation only. However, we provide two closely related packages, viz. LingTrees by Avery Andrews and the allmighty PSTricks by Timothy van Zandt. Both packages include comprehensive documentation (available from the links above); in many respects, LingTrees is a front-end to PSTricks, aiming to make partially automate the generation of complex tree descriptions by virtue of a specialized description language and pre-processor script (named trees.py and included in our CVS tree). Furthermore, LingTrees provides a legacy layer of compatibility macros for the most common commands of tree-dvips.sty, so in some cases you might not even have to adapt existing tree code (much).
The LOGON book setup includes a new set of macros for typesetting MRSs. As with many of the other macros we provide, the main goal here is to separate the structural description of an MRS (or parts of an MRS) from decisions about typography or aesthetics. Please try to avoid all direct format control in MRSs (and AVMs, trees, et al.), e.g. do not select a specific font, font attributes, or size. The MRS macros essentially provide one command per MRS component type, e.g. (working from the smaller to the larger) for a single variable, one role--value pair, one EP, and so on. At the top-level, there is a choice as to whether a full MRS should be typeset in standard, running-text mode (which is only feasible for quite short expressions) or in multi-line, aligned mode. Consider the following sample (see below for details on the example environment):
\rolesfalse
\propertiestrue
\begin{exe}
\ex\label{ex:generation:athlete}
\sblock{\sh{0}}{%
\slep{0}{proposition\_m}{\srv{ARG0}{e}{0}{}, \srv{MARG}{h}{1}{}},
\slep{2}{\_run\_v}{%
\srv{ARG0}{e}{0}{\svp{TENSE}{past}}, \srv{ARG1}{x}{0}{}},\\
\sep{\srule}{\_the\_q}{%
\srole{ARG0}{\svar{x}{0}{}}, \srv{RSTR}{h}{4}{}, \srv{BODY}{h}{5}{}},
\slep{6}{\_athlete\_n}{\srv{ARG0}{x}{0}{}},\\
\slep{6}{\_young\_a}{\srole{ARG0}{\srule}, \srv{ARG1}{x}{0}{}},
\slep{6}{\_polish\_a}{\srole{ARG0}{\srule}, \srv{ARG1}{x}{0}{}}%
}%
{\sqeq{1}{2}, \sqeq{4}{6}}
\end{exe}
This example deliberately uses several variants of the macros: the general command to typeset one EP, for example, is \sep{}{}{}, where the three arguments are the handle, predicate, and roles, respectively. The variant command \slep{}{}{} is an abbreviation for convenience: instead of an arbitrary first argument, it expects as its first argument the variable identifier of the handle, i.e. \slep{0}{foo}{...} translates into \sep{\sh{0}}{foo}{...}. Likewise, \sh{0} is an abbreviation of \svar{h}{0}{}, i.e. a variable of type h, with identifier 0, and an empty set of variable properties. Please see the style file `mrs.sty` for more examples and the full inventory of MRS macros. Two boolean toggles control whether role labels (ARG0 et al.) and variable properties (e.g. TENSE) are included in the output. The defaults are \rolesfalse and \propertiestrue, but it is possible to change these at any point (note that such changes remain in effect until a later declaration, though the above defaults are restored upon entry to each chapter). For increased flexibility, please always include role labels in the LaTeX code for MRSs, even if you believe you never want to activate display of such labels in your chapter.
For numbered linguistic examples and other in-text numbered and displayed material, we provide the fairly standard package gb4e. Among other things, it provides nested examples, flexible referencing to examples or sub-examples, and aligned glosses. For a simple example, including an in-text reference, consider the following
Examples \refx{ex:erg:bridge} and \refx{ex:erg:shelter} below are taken from
the development corpus.
\begin{exe}
\ex\label{ex:erg:bridge}
We asked him about the classic bridge.
\ex\label{ex:erg:shelter}
He recommended that the club build a shelter.
\end{exe}
To provide an aligned gloss, and optionally a translation, there are some additional commands, e.g.
\begin{exe}
\ex\label{ex:erg:vei}
\gll
Veien mot Bergen er kort.\\
Road\sub{def} towards Bergen is short.\\
\trans The road towards Bergen is short.
\end{exe}
Note that the commands \refx{} and \sub{} are LOGON-specific extensions to the gb4e package.
The package gb4e turns the \automath feature on making symbols _ (underline) and ^ (caret) active. This means you can get sub- and super-scripting outside maths mode, e.g. write foo_i instead of foo$_i$. However this can cause errors such as ! TeX capacity exceeded, sorry [parameter stack size=10000]. You can turn off the \automath feature by putting \noautomath in your preamble.
The gb4e package is incompatible with the beamer class. A more modern package providing the same features as gb4e, but also compatible with beamer, is expex.
The default support for creating tables in LaTeX is not optimal and much better results can be produced using the booktabs package of Simon Fear. The documentation that comes with it also serves as an excellent style guide for tables. The following is an example of a simple table using booktabs. (Note how the table is structured around the commands \toprule, \midrule and \bottomrule, which will insert appropriate vertical spacing. Should you need additional spacing use \addlinespace.)
\begin{table}[t!]
\begin{center}
\begin{tabular}{@{}lcc@{}}
\toprule
&\multicolumn{2}{c}{\textbf{Judgement}}\\
\cmidrule{2-3}
\textbf{Element} &\textbf{Yay}&\textbf{Nay}\\
\midrule
Vertical rules & & X \\
Double rules & & X \\
Horizontal rules & X & \\
Single rules & X & \\
\bottomrule
\end{tabular}
\caption{This is a simple table.}
\label{tb:mytable}
\end{center}
\end{table}
The package ctable is extends booktabs to a simple but flexible command \ctable, and abbreviates the rule commands to \FL, \ML, \LL (first, middle, last). The package has useful options for things like notes in the table and landscape tables as well. Using ctable, the table above becomes:
\ctable[botcap,
caption={This is a simple table},
label=tb:mytable,
pos=t!]{@{}lcc@{}}{}{
\FL
&\multicolumn{2}{c}{\textbf{Judgement}}\NN
\cmidrule{2-3}
\textbf{Element} &\textbf{Yay}&\textbf{Nay}\ML
Vertical rules & & X \NN
Double rules & & X \NN
Horizontal rules & X & \NN
Single rules & X & \LL
}
In our line of work, tables of numbers are quite frequent. In particular, tables of decimal numbers are very common. Ideally, this kind of table should have the numbers aligned on the decimal point with either the widest number centered or right-aligned (according to CMS16, anyways). There are various hacks to get this right, but the best solution is to use the siunitx package. This package supports both ways of formatting columns. For a right-aligned column of numbers, automatically rounded to three significant digits, with a maximum of 1 digit before the period and a maximum of 3 after, use the following:
% In the preamble:
\usepackage{siunitx}
\sisetup{table-number-alignment=right,
table-text-alignment=right,
detect-mode=true}
% And then in the document:
\ctable{lS[table-figures-integer=1,table-figures-decimal=3,round-mode=figures,round-precision=3]}{}{
\FL
Header 1 & {Header 2} \ML
Label & 1.23456 \NN
Label & 12.3678
\LL
}
Note the braces around the header of the numerical column. These are required to prevent siunitx from parsing the text as a number. This will result in a typeset value of "1.23" in the first row and "12.4" in the second. Since the S column specifier is rather verbose, I recommend wrapping it in a custom column type, like so:
\newcolumntype{R}[2]{S[table-figures-integer=#1,table-figures-decimal=#2,round-mode=figures,round-precision=3]}
This lets the example above use the column spec lR{1}{3} instead of the overlong S[...] spec. For centered numbers, the table-* options to \sisetup and the S specifier can be omitted. For chapter and verse on this (and many other neat features), see the siunitx documentation.
A well-known tribulation in academia is the dreaded page limit. Sometimes, you just have too much to say and not quite enough space on the page to say it. The first thing to look for is paragraphs whose last lines are very short. Reword them so that you can squeeze them into one line less.
If you still need to squeeze a bit, there are some slightly more naughty things that can be done:
-
The amount of space between the title and author info and the start of the text: \titlebox3.5cm
-
Spacing after a floating figure: \setlength\textfloatsep{2.5mm}
-
If you have enumerations or itemizes, use the enumitem package remove the spacing between items: \setlist{nolistsep}
The exact spacings above can be tweaked to taste; and remember to not make it too blatant that you're squeezing.
Warning! This last option is truly the nuclear option. It is more than a little shady, and if you're caught it can be grounds for outright rejection of your paper. Here be dragons, caveat emptor, cave canem, etc, etc, etc.
Finally, if you just can't get rid of those last few lines, you can reduce the amount of space between lines: \renewcommand\baselinestretch{0.95}. Note that this parameter is very sensitive, and can make your paper look very funny indeed. Use with care.
Home | Forum | Discussions | Events