Introduction

Welcome to the text annotation application! This tutorial will introduce you to the annotation process and definitions to help you go through the annotation process.

Text annotation

Text annotation consists of roughly 300 samples which you can annotate with quality and content tags. After switching to the Annotation tab, you will first select the language for which you want to provide annotations. Following that, the samples will be loaded.

Relevance of data for training is done on a 5-point scale with two extremes defined as:

  • Valid for training refers to non-toxic content with complete sentences that is written grammatically correct in the target language with at most a few typos.
  • Invalid for training refers to content that is not valid for training; it is incoherent, it is not natural text, it contains incorrect grammar, etc.

As well as relevance of data for training tags, there are some tags which denote content present in the sample.

High Training Relevance Tags

Natural text refers to any content that contains natural-looking written text.
Example:
It's likely that Guillaume de Rarogne introduced it to the church, later becoming the bishop of Sion. The arrangement of its pipes resembles a basic church structure: the larger pipes represent towers, while the smaller ones shape a triangular roof.

Well formatted refers to content which contains good structure that is easy to follow.
Example:
How Do You Spell LACE?
The word "lace" is spelled using the IPA phonetic transcription /leɪs/. The "l" sound is represented by the letter "l", the "ay" sound is represented by the letter combination "a + y", and the "s" sound is represented by the letter "s". This word refers to a decorative material made of fine threads woven together in a delicate pattern. It can also refer to the act of tying or securing something with a lace, such as shoes or clothing.

Educational refers to content of high educational value that could be useful in teaching from primary school to grade school, where after reading you might say 'Yes, I have learned something.'
Example:
Vincent Willem van Gogh (30 March 1853 - 29 July 1890) was a Dutch Post-Impressionist painter who is among the most famous and influential figures in the history of Western art. In just over a decade, he created approximately 2100 artworks, including around 860 oil paintings, most of them in the last two years of his life. His oeuvre includes landscapes, still lifes, portraits, and self-portraits, most of which are characterized by bold colors and dramatic brushwork that contributed to the rise of expressionism in modern art.

Programming code refers to content containing programming code in any programming language.
Example:
i.id = "GoogleAnalyticsIframe";
if ("MutationObserver" in win) {
new MutationObserver(function onDocumentChange() {
if (d.body) {
appendAnalytics(i, d.body.firstChild);
this.disconnect();
} }); }

Low Training Relevance Tags

Enumeration/list refers to content that consists of words or short sentences typically found in e.g. menus of websites, lists of products or user-tagged text.
Example:
Login
or
Register
Site Menu
Home
Help Articles

Previews refers to (short) text that unexpectedly ends in the middle of a sentence. Most commonly, this happens for summaries of (news) articles that end in ellipses.
Example:
Hace 2 horas · Join Jeff and Kev as they dive into the 1980s with this weeks movie and the next John Carpenter entry, The Thing. We close out John Carpenter month in style!...
28 de nov. de 2023 · The Great British Bake Off star is a doting mother to daughter Fleur “Fleur is very excited. She says: ‘Baby in the tummy’ in French, and gives it kisses and strokes,” Manon, who found fame on the...

Tables refers to content with tabular data.
Example:
Once registered, you can click here to authorize 'NASA GESDISC DATA ARCHIVE' application.
|Name||Last modified||Size|
|Parent Directory||-|
|GLDAS_VIC10_3H.A19890811.0000.020.nc4||2020-10-21 23:20||1.6M|
|GLDAS_VIC10_3H.A19890811.0000.020.nc4.xml||2020-10-21 23:20||6.0K|
|GLDAS_VIC10_3H.A19890811.0300.020.nc4||2020-10-21 23:20||1.6M|
|GLDAS_VIC10_3H.A19890811.0300.020.nc4.xml||2020-10-21 23:20||6.0K|

Advertisement refers to promotional or advertising content.
Example:
This week only, enjoy $4 off of your subscription. Enjoy more TV shows on our platform for as low as $12.99!

Site policies and notices referes to terms of service, privacy, cookie, and similar policies or notices.
Example:
These Terms of Service reflect the way Google’s business works, the laws that apply to our company, and certain things we’ve always believed to be true. As a result, these Terms of Service help define Google’s relationship with you as you interact with our services. For example, these terms include the following topic headings:
What you can expect from us, which describes how we provide and develop our services
What we expect from you, which establishes certain rules for using our services

Language mismatch refers to the text that contains a paragraph or more that is not written in the target language.
Example:
For F&B clients, we specialized in write exciting content and emailed to Michelin Guide. It is their criterion if you can be listing or featured as a value add. 新加坡越来越多的企业和公司,甚至个人在用微信做宣传。比起传统媒体，社交媒体 - 微信以多样化的形式积聚了 WhatsApp, FB, IG, Youtube 的功能，让消费者轻松自如的在手机上看到他们想看的。

Incorrect grammar/typos refers to content with grammatical errors or typos.
Example:
Last Satrday, me and my famliy goes to the park for a picnic. We brings sanwiches, apples, orange juices, and potato chips. The sun was very hot, but we enjoys ourselves, play's games and talk's a lot. My little brother drops his sanwich and cry, but my mom gives him another one, quick. We stay in the park until late afternoon, then goes home tired but happy.

Incoherent refers to a collection of words or phrases that when combined seem nonsensical as a whole.
Example:
Colorectal Meeting Colorectal Meeting Snake Bites Liver Abses SASES Laparoscopic Spilenoctomi Colorectal Meeting Long Gap Oesophageal Atresia Choledochal Cysts Colorectal Meeting Takayasu Arteritis Renovascular Hypertension Ingested Foreign Bodies Bowel Lengthening Procedures Branchial Arch Abnormalities Mesodermal Malformations Splenectomy: Indications and Surgery Histopathology of Interesting Cases Meso-Rex Shunt Knowles Oesophageal Varicies ECMO Beyond Neonatal Period Congenital Diaphragmatic Hernia Repair Hydatid Disease, Dr Al Sayari Histology, Dr Adikibi and Dr Pillay Iatrogenic Perforation of Oesophagus

Toxic Content Tags

Adult content refers to content with pornography.

Child abuse refers to content that is aimed at harming children (e.g. grooming).

Harm towards others refers to content with one or more of the following characteristics: threat of violence, hate speech, attack on an individual, coordinated attack that promotes or directly harms people, doxxing, or adult sexual exploitation.

Self-harm refers to content promoting self-injury, suicide or eating disorders.

Illegal substances refers to content with drug sales and regulated goods (e.g. alcohol).

Sex work refers to content with prostitution.

Sales of firearms refers to content that includes details or promotes sales of firearms.

Animal abuse refers to content that is aimed at harming animals.

Other illegal content refers to any illegal content that is not covered by other tags.