DAWT: Densely Annotated Wikipedia Texts across multiple languages

TL;DR

This paper introduces the DAWT dataset, a large, multilingual, densely annotated Wikipedia text corpus with extensive entity linking and type information, designed to support NLP and IR research.

Contribution

The paper presents a methodology to generate a richly annotated multilingual Wikipedia dataset with increased link density and provides multiple derived datasets for research use.

Findings

Contains 13.6M articles and 5.0B tokens across multiple languages.

Enriches Wikipedia markup to increase entity links by 4.8 times.

Provides derived datasets including mention-entity co-occurrence and entity embeddings.

Abstract

In this work, we open up the DAWT dataset - Densely Annotated Wikipedia Texts across multiple languages. The annotations include labeled text mentions mapping to entities (represented by their Freebase machine ids) as well as the type of the entity. The data set contains total of 13.6M articles, 5.0B tokens, 13.8M mention entity co-occurrences. DAWT contains 4.8 times more anchor text to entity links than originally present in the Wikipedia markup. Moreover, it spans several languages including English, Spanish, Italian, German, French and Arabic. We also present the methodology used to generate the dataset which enriches Wikipedia markup in order to increase number of links. In addition to the main dataset, we open up several derived datasets including mention entity co-occurrence counts and entity embeddings, as well as mappings between Freebase ids and Wikidata item ids. We also discuss two applications of these datasets and hope that opening them up would prove useful for the Natural Language Processing and Information Retrieval communities, as well as facilitate multi-lingual research.

Figures

Tables

Table 1. Table 1: Frequency of Special Characters

Non-English or Math	Frequency	Comments
Ø	1 in 1,000	For Swedish names
$\pi$	1 in 5	Common in math
$	4 in 5	Used in business
${\Psi }_1^2$	1 in 40,000	Unexplained usage

Table 2. Table 2: Some Typical Commands

Command	A Number	Comments
\alignauthor	100	Author alignment
\numberofauthors	200	Author enumeration
\table	300	For tables
\table*	400	For wider tables

Equations

$$\underset{n\to \infty }{\lim }x=0$$

$$\sum _{i=0}^{\infty }x+1$$

$$\sum _{i=0}^{\infty }{x}_i={\int }_0^{\pi +2}f$$

$${\int }_a^{b}f(t)dt=G(b)-G(a).$$

$$\log e^z=z$$

$$\underset{x\to \infty }{\lim }\frac{f(x)}{g(x)}=L.$$

$$l=\underset{x\to c}{\lim }f(x)=\underset{x\to c}{\lim }\left[gx\cdot \frac{f(x)}{g(x)}\right]=\underset{x\to c}{\lim }g(x)\cdot \underset{x\to c}{\lim }\frac{f(x)}{g(x)}=0\cdot L=0,$$

Full text

The developers have tried to include every imaginable sort of “bells and whistles", such as a subtitle, footnotes on title, subtitle and authors, as well as in the text, and every optional component (e.g. Acknowledgments, Additional Authors, Appendices), not to mention examples of equations, theorems, tables and figures.

To make best use of this sample document, run it through LaTeX and BibTeX, and compare this source code with the printed output produced by the dvi file. A compiled PDF version is available on the web page to help you with the ‘look and feel’.

††conference: WWW 2017, April 3–7, 2017, Perth, Australia.††titlenote: (Produces the permission block, and copyright information). For use with SIG-ALTERNATE.CLS. Supported by ACM.††titlenote: A full version of this paper is available as Author’s Guide to Preparing ACM SIG Proceedings Using LaTeX$2_{\epsilon}$ and BibTeX at www.acm.org/eaddress.htm††titlenote: Dr. Trovato insisted his name be first.††titlenote: The secretary disavows any knowledge of this author’s actions.††titlenote: This author is the one who did all the really hard work.

1 Introduction

The proceedings are the records of a conference. ACM seeks to give these conference by-products a uniform, high-quality appearance. To do this, ACM has some rigid requirements for the format of the proceedings documents: there is a specified format (balanced double columns), a specified set of fonts (Arial or Helvetica and Times Roman) in certain specified sizes (for instance, 9 point for body copy), a specified live area (18 $\times$ 23.5 cm [7" $\times$ 9.25"]) centered on the page, specified size of margins (1.9 cm [0.75"]) top, (2.54 cm [1"]) bottom and (1.9 cm [.75"]) left and right; specified column width (8.45 cm [3.33"]) and gutter size (.83 cm [.33"]).

The good news is, with only a handful of manual settings111Two of these, the \numberofauthors and \alignauthor commands, you have already used; another, \balancecolumns, will be used in your very last run of LaTeX to ensure balanced column heights on the last page., the LaTeX document class file handles all of this for you.

The remainder of this document is concerned with showing, in the context of an “actual” document, the LaTeX commands specifically available for denoting the structure of a proceedings paper, rather than with giving rigorous descriptions or explanations of such commands.

2 The \secitBody of The Paper

Typically, the body of a paper is organized into a hierarchical structure, with numbered or unnumbered headings for sections, subsections, sub-subsections, and even smaller sections. The command \section that precedes this paragraph is part of such a hierarchy.222This is the second footnote. It starts a series of three footnotes that add nothing informational, but just give an idea of how footnotes work and look. It is a wordy one, just so you see how a longish one plays out. LaTeX handles the numbering and placement of these headings for you, when you use the appropriate heading commands around the titles of the headings. If you want a sub-subsection or smaller part to be unnumbered in your output, simply append an asterisk to the command name. Examples of both numbered and unnumbered headings will appear throughout the balance of this sample document.

Because the entire article is contained in the document environment, you can indicate the start of a new paragraph with a blank line in your input file; that is why this sentence forms a separate paragraph.

2.1 Type Changes and \subsecitSpecial Characters

We have already seen several typeface changes in this sample. You can indicate italicized words or phrases in your text with the command \textit; emboldening with the command \textbf and typewriter-style (for instance, for computer code) with \texttt. But remember, you do not have to indicate typestyle changes when such changes are part of the structural elements of your article; for instance, the heading of this subsection will be in a sans serif333A third footnote, here. Let’s make this a rather short one to see how it looks. typeface, but that is handled by the document class file. Take care with the use of444A fourth, and last, footnote. the curly braces in typeface changes; they mark the beginning and end of the text that is to be in the different typeface.

You can use whatever symbols, accented characters, or non-English characters you need anywhere in your document; you can find a complete list of what is available in the LaTeX* User’s Guide[Lamport:LaTeX]*.

2.2 Math Equations

You may want to display math equations in three distinct styles: inline, numbered or non-numbered display. Each of the three are discussed in the next sections.

2.2.1 Inline (In-text) Equations

A formula that appears in the running text is called an inline or in-text formula. It is produced by the math environment, which can be invoked with the usual \begin. . .\end construction or with the short form $. . .$. You can use any of the symbols and structures, from $\alpha$ to $\omega$, available in LaTeX[Lamport:LaTeX]; this section will simply show a few examples of in-text equations in context. Notice how this equation: $\lim_{n\rightarrow\infty}x=0$, set here in in-line math style, looks slightly different when set in display style. (See next section).

2.2.2 Display Equations

A numbered display equation – one set off by vertical space from the text and centered horizontally – is produced by the equation environment. An unnumbered display equation is produced by the displaymath environment.

Again, in either environment, you can use any of the symbols and structures available in LaTeX; this section will just give a couple of examples of display equations in context. First, consider the equation, shown as an inline equation above:

[TABLE]

Notice how it is formatted somewhat differently in the displaymath environment. Now, we’ll enter an unnumbered equation:

[TABLE]

and follow it with another numbered equation:

[TABLE]

just to demonstrate LaTeX’s able handling of numbering.

2.3 Citations

Citations to articles [bowman:reasoning, clark:pct, braams:babel, herlihy:methodology], conference proceedings [clark:pct] or books [salas:calculus, Lamport:LaTeX] listed in the Bibliography section of your article will occur throughout the text of your article. You should use BibTeX to automatically produce this bibliography; you simply need to insert one of several citation commands with a key of the item cited in the proper location in the .tex file [Lamport:LaTeX]. The key is a short reference you invent to uniquely identify each work; in this sample document, the key is the first author’s surname and a word from the title. This identifying key is included with each item in the .bib file for your article.

The details of the construction of the .bib file are beyond the scope of this sample document, but more information can be found in the Author’s Guide, and exhaustive details in the LaTeX* User’s Guide[Lamport:LaTeX]*.

This article shows only the plainest form of the citation command, using \cite. This is what is stipulated in the SIGS style specifications. No other citation format is endorsed or supported.

2.4 Tables

Because tables cannot be split across pages, the best placement for them is typically the top of the page nearest their initial cite. To ensure this proper “floating” placement of tables, use the environment table to enclose the table’s contents and the table caption. The contents of the table itself must go in the tabular environment, to be aligned properly in rows and columns, with the desired horizontal and vertical rules. Again, detailed instructions on tabular material is found in the LaTeX* User’s Guide*.

Immediately following this sentence is the point at which Table 1 is included in the input file; compare the placement of the table here with the table in the printed dvi output of this document.

To set a wider table, which takes up the whole width of the page’s live area, use the environment table* to enclose the table’s contents and the table caption. As with a single-column table, this wide table will “float" to a location deemed more desirable. Immediately following this sentence is the point at which Table 2 is included in the input file; again, it is instructive to compare the placement of the table here with the table in the printed dvi output of this document.

2.5 Figures

Like tables, figures cannot be split across pages; the best placement for them is typically the top or the bottom of the page nearest their initial cite. To ensure this proper “floating” placement of figures, use the environment figure to enclose the figure and its caption.

This sample document contains examples of .eps files to be displayable with LaTeX. If you work with pdfLaTeX, use files in the .pdf format. Note that most modern TeX system will convert .eps to .pdf for you on the fly. More details on each of these is found in the Author’s Guide.

As was the case with tables, you may want a figure that spans two columns. To do this, and still to ensure proper “floating” placement of tables, use the environment figure* to enclose the figure and its caption. and don’t forget to end the environment with figure*, not figure!

2.6 Theorem-like Constructs

Other common constructs that may occur in your article are the forms for logical constructs like theorems, axioms, corollaries and proofs. There are two forms, one produced by the command \newtheorem and the other by the command \newdef; perhaps the clearest and easiest way to distinguish them is to compare the two in the output of this sample document:

This uses the theorem environment, created by the\newtheorem command:

Theorem 1

Let $f$ be continuous on $[a,b]$. If $G$ is an antiderivative for $f$ on $[a,b]$, then

[TABLE]

The other uses the definition environment, created by the \newdef command:

Definition 1**.**

If $z$ is irrational, then by $e^{z}$ we mean the unique number which has logarithm $z$:

[TABLE]

Two lists of constructs that use one of these forms is given in the Author’s Guidelines.

There is one other similar construct environment, which is already set up for you; i.e. you must not use a \newdef command to create it: the proof environment. Here is a example of its use:

Proof.

Suppose on the contrary there exists a real number $L$ such that

[TABLE]

Then

[TABLE]

which contradicts our assumption that $l\neq 0$. ∎

Complete rules about using these environments and using the two different creation commands are in the Author’s Guide; please consult it for more detailed instructions. If you need to use another construct, not listed therein, which you want to have the same formatting as the Theorem or the Definition[salas:calculus] shown above, use the \newtheorem or the \newdef command, respectively, to create it.

A \secitCaveat for the TeX Expert

Because you have just been given permission to use the \newdef command to create a new form, you might think you can use TeX’s \def to create a new command: Please refrain from doing this! Remember that your LaTeX source code is primarily intended to create camera-ready copy, but may be converted to other forms – e.g. HTML. If you inadvertently omit some or all of the \defs recompilation will be, to say the least, problematic.

3 Conclusions

This paragraph will end the body of this sample document. Remember that you might still have Acknowledgments or Appendices; brief samples of these follow. There is still the Bibliography to deal with; and we will make a disclaimer about that here: with the exception of the reference to the LaTeX book, the citations in this paper are to articles which have nothing to do with the present subject and are used as examples only.

4 Acknowledgments

This section is optional; it is a location for you to acknowledge grants, funding, editing assistance and what have you. In the present case, for example, the authors would like to thank Gerald Murray of ACM for his help in codifying this Author’s Guide and the .cls and .tex files that it describes.

Appendix A Headings in Appendices

The rules about hierarchical headings discussed above for the body of the article are different in the appendices. In the appendix environment, the command section is used to indicate the start of each Appendix, with alphabetic order designation (i.e. the first is A, the second B, etc.) and a title (if you include one). So, if you need hierarchical structure within an Appendix, start with subsection as the highest level. Here is an outline of the body of this document in Appendix-appropriate form:

A.1 Introduction

A.2 The Body of the Paper

A.2.1 Type Changes and Special Characters

A.2.2 Math Equations

Inline (In-text) Equations

Display Equations

A.2.3 Citations

A.2.4 Tables

A.2.5 Figures

A.2.6 Theorem-like Constructs

A Caveat for the TeX Expert

A.3 Conclusions

A.4 Acknowledgments

A.5 Additional Authors

This section is inserted by LaTeX; you do not insert it. You just add the names and information in the \additionalauthors command at the start of the document.

A.6 References

Generated by bibtex from your  .bib file. Run latex, then bibtex, then latex twice (to resolve references) to create the  .bbl file. Insert that  .bbl file into the .tex source file and comment out the command \thebibliography.

Appendix B More Help for the Hardy

The sig-alternate.cls file itself is chock-full of succinct and helpful comments. If you consider yourself a moderately experienced to expert user of LaTeX, you may find reading it useful but please remember not to change it.