Services

Data-capture

We help you to get and organize information from public data or different websites with scrapping.

Data-cleaning

We structure databases with information from multiple databases in multiple formats. Organization of information and standardization of variables.

Visualization apps

We create public data visualization applications so that your users can know and explore databases. We use the latest technologies in data visualization to communicate information.

Algorithms

We implement artificial intelligence algorithms to facilitate your work with data, from predictive algorithms to pattern recognition.

Web specials

We develop interactive web specials based on data. The specials have different visual components to guide your readers. See examples of our specials.

About us

Datasketch is a digital platform of investigative and data journalism. Our portal allows journalists, data scientists, social scientists and citizens in general to learn and consult on data visualizations, tools, software and in-depth research on various short-term issues. We have free data tools and different projects to bridge the gap between data and citizenship that facilitates the democratization of knowledge and a critical review of social realities based on information contrasts.

Our team

Juan Pablo Marín

Electronic engineer with a master's degree in computational statistics. Expert in data science with applications in multiple areas such as economics, hydrology and journalism.

Camila Achuri

Statistics and expert in R programming language. She has developed various applications of data visualization in mobility and open data subjects.

Juliana Galvis

Politologist and candidate for a Master in Digital Humanities. She is currently leading the development of the Who Is database, as well as supporting journalistic research and the creation of databases.

David Daza

Bachelor of Electronics. Expert in development of applications and websites with emphasis on data journalism and content management of multiple databases.

Verónica Toro

Anthropologist and researcher. Responsible for the management and organization of the data-community in Colombia and Latin America and provide support in journalistic investigations and the creation of databases.

Andrea Cervera

Journalist responsible for writing articles, provide investigative support and community manager.

Mariana Villamizar

Systems engineer and designer. Expert in user experience, data visualization and graphic communication. Feminist.

Contact

If Maluma painted with his lyrics...

February 16, 2018

If Maluma painted with his lyrics, what would be the color palette of his songs? Would they be happy or sad pictures? What color predominates in his rhythms?

If Maluma painted with his lyrics, what would be the color palette of his songs? Would they be happy or sad pictures? What color predominates in his rhythms? These were some of the questions we raised in Datasketch on an afternoon of music and beers. And although they initially seemed impossible to solve, we set ourselves the task of working with data to find the answers.

 

To do so, we chose the lyrics of the five songs that have had the most views on YouTube of the. From there we did our own algorithm to discover how Maluma paints.

 

Algorithms in general allow you to translate and transform data from one source into another, such as converting letters into numbers or numbers into symbols ... let's say they are the magic wand of this trick.

Then, we developed an algorithm that allowed us an unusual transformation but with creativity and some computational tools allowed us to convert words into colors. For this we start converting words into images and then color images.

 

Convert words to images: A Google image search of any word brings us with a high the most popular images that represent the searched word. This search is limited by the region, like all information that Google filters us.

 

Convert images to colors: Once we see the images, we must find the representative color. We do this by sort of mixing all the colors in the image. This is called "clustering", a technique  used by many companies like Pinterest, Twitter and Google. As users we see it when the app shows us a blur single tone image while loading the final image. Has this happened to you?

 

Thus, we discover the color that represents each of the words in the most popular lyrics of the artist. We could see that Maluma, the painter, likes pastel colors that are in the range of greens and purples.

 

This seems like a shocking once you analyze the artist’s image of a rude man who always wears dark colors… but the words in his songs seem to prefer much more feminine tones.

 

Of course the algorithms we use do not analyze contexts or understand the meaning of what is being sung ... they don’t know if the meaning of the word is violent or sexual. If so, very the color palette would be different.

 

The result? Watch this video.

 

Datasketch

Data team