On November 3, 2016, the Company’s data scientist delivered a "TechNovation Brief - Machine Learning” seminar to United Nations’ officials and staff at the UN Headquarters in New York City. The presentation dealt with two case studies developed especially for ICT: one on global news monitoring and topic retrieval and the other on identifying propaganda disseminated via Twitter. These projects were used to illustrate machine learning’s potential uses in recognizing and categorizing requested information from large data inputs. During the news project’s categorization process, our scientists were happy to find that most of the topics were related to
“In my view, the importance of this project goes well beyond the modeling results by providing an opportunity to explore how data science can impact programmatic needs on use cases related to UN mandates. Presenting these case studies and results at the UN’s first TechNovation briefing, exposes UN staff from different parts of the institution both to machine learning as a powerful tool and to the design thinking that will be useful in implementing this tool. We have greatly enjoyed this opportunity to work with deepsense.io.” Radia Funna, Head of Innovation, Office of Information & Communications Technology.
“Machine learning can be applied to data analysis and automatization, and it can improve global organizations' and companies' everyday processes. We were very happy to share our knowledge and experience with the Office of Information & Communications Technology at the United Nations," added Tomasz Kulakowski, CEO at deepsense.io.
Monitoring and topic retrieval for global news
The project began with collecting United Nations-related news items (raw text) from around the world. The news items were collected through the GDELT project
database which describes important news events and provides a reference to corresponding articles online. After downloading and pre-processing the news our team used the Latent Dirichlet algorithm to summarize the content of the news by a set of topics. The project’s primary focus was on unsupervised learning methods capable of inferring trending topics from a set of documents without additional human input. The results identified a number of topics related to issues such as African development, women's health, current conflicts in the Middle East, nuclear weapons’ proliferation and climate change. Such analyses are helpful in monitoring the carrying out of the Sustainable Development Goals or getting insights into how a given topic is covered by the media.
"Machine learning seems particularly effective in two situations: systems that are unpredictable and constantly changing and systems that are so complex that we cannot fully describe them in one model. The UN is facing both of these: understanding how the world works is infinitely complex, and it changes all the time. That is why I believe machine learning can be of value to us." Lambert Hogenhout, Chief Data Analytics, Innovation and Partnerships at United Nations
Detecting propaganda on Twitter
The second project dealt with monitoring Twitter posts for propaganda aimed at a particular target group. Processing Twitter data is challenging due to the enormous number of Tweets posted every second, their brevity and informal tone. To obtain data about a particular Twitter account, the deepsense.io team combined user-account statistics (number of Tweets and followers) with text analytics methods to process the content published by individual users. This data enabled the team to classify an account to a particular target group. The definition of a target group is arbitrary, for example, it might be a group of users supporting a particular idea/movement.
Inside "TechNovation Brief - Machine Learning” UN Seminar
The deepsense.io projects concluded with an internal UN "TechNovation Brief” seminar held on Nov. 3, 2016. During the meeting, officials from different UN agencies and departments had a chance to learn how powerful machine learning technologies could be used in their work. In addition to presenting the projects’ results, the seminar also included a discussion and a short demo on internal UN machine learning projects (NLP, Entity Recognition, Bots (Global Services Division)), as well as a general review of using machine learning predict domestic violence.
deepsense.io looks forward to continuing its cooperation with the United Nations’ Office of Information and Communications Technology in the future.
About Information and Communications Technology at the United Nations:
The Office of Information and Communications Technology is responsible for defining strategic direction for ICT to the Secretariat. It provides oversight of ICT programmes, budgets and decision-making to ensure alignment with the Secretariat’s overall ICT strategy. The team consists of more than 300 people operating global data centers and coordinating services provided by more than 4,000 people in UN offices worldwide. The Office is headquartered in New York and sits within the Department of Management.
deepsense.io creates deep learning solutions for enterprises. The award-winning team of data scientists and engineers delivers state-of-the-art deep learning, machine learning and Big Data analytics. deepsense.io’s flagship product Seahorse, a scalable data analytics workbench powered by Apache Spark, helps data scientists work more efficiently and intuitively by visually designing, editing and executing Spark applications. Neptune, deepsense.io’s newest technology, is a machine learning platform designed to efficiently manage and monitor data science experiments. The company also runs on-site corporate machine learning and Big Data workshops offering intensive, fully customizable training programs.