Use a text annotation tool to find information in text quicker

Photo by PNG Design on Unsplash

Ask anyone dealing with lots of documents (journalists, researchers, analysts, etc.), and you will realize that finding stuff in text has long been a highly time-consuming part of their daily work life. Thousands of pages (hopefully now digitized) need to be read through but how do you find information in this dense mountain of words quickly!

Text annotation tools have been around for a while now in some form or another. But the last years have seen tons of tools that are simply overcomplicated or just focused on highlighting text and adding side notes.

At tagtog we have developed a simple and entirely web-based text annotation tool ready to automate the search of relevant information. The service can be used by anyone for free (larger requirements have pricing plans), you simply need to sign up. Create a project, import your text and start annotating.

So, if now you are thinking what makes this different or better than the basic Ctrl + F on a word doc, then read on -

Annotation goes way beyond just finding a word.

In tagtog you annotate text, and in addition, you can define relations between these pieces of text. For example: if you are looking through an archive of news and you want to find articles of Donald Trump and his views on the environment, you can define this relationship and you will find similar statements that correlate.

All these annotations added to the text are searchable and downloadable!

But the most exciting part is the integration of machine learning (ML) with the tool! This is when tagtog helps you find relevant information in text automatically.

Taking forward the Trump example, but with his Twitter account, let's try a practical example to find tweets that are connected with the environment.

Donald Trump, Laconia Rally, Laconia, Weirs Community Center, 2015. Author: Michael Vadon
  1. You define what you want to annotate: in this case, words representing the environment (e.g. environment, climate change, pollution, fire, etc.).
An entity is just something that exists, something you can annotate. You can use this panel to define what type of information you want to identify in text.

2. You annotate few examples. In this example, each paragraph is a tweet.

3. tagtog picks these annotations up and generate an ML model that annotate automatically these concepts in text. Now you don't need to go through all the tweets, but just those annotated 🎉. Let's give it a try by importing a new set of tweets.

Automatic annotations produced by the ML model. These represent the tweets connected with the environment. Notice that the model does not annotate based on string matching, but it extrapolates rules from the annotations. Notice the manual annotation Keystone XL Pipeline and the automatic annotation Keystone Pipeline

4. You will, of course, have to check the annotations to make corrections and each time you make changes, the ML model learns and becomes more accurate!

So, do you see the possibilities? You can build, train and deploy a custom ML model to automatically find relevant information in literally just 5 minutes — with zero knowledge of the ML or coding or any other scary technical stuff! We have done it all for you. (here’s a link to an earlier blog that takes you into more details).

📄 Official documentation:

At 🍃 we aim to democratize text analytics.

Whether you are an individual or an organization, if text analysis is a part of your work then we strongly recommend you give tagtog a quick spin. It just might be the labeling tool that you are looking for!

👏 👏 👏 if you liked the post and want to share it with others!