At Gulf Air, an internally developed big data solution is helping the airline understand how passengers feel about its products and services, as well as enhancing more customer loyalty by delivering offers and promotions. The solution — called Arabic Sentiment Analysis — is built on open source technologies and runs in the airline’s private cloud. This cloud is using Red Hat products to achieve reliability, availability, and scalability in a virtualised environment that encompasses more than 200 servers and more than 100 different enterprise applications.
Sentiment analysis for social media
Gulf Air has developed a sentiment analysis solution based on big data technologies that is capable of addressing social media posts in both Arabic and English languages. Called Arabic Sentiment Analysis, the solution was developed entirely in-house by Gulf Air.
“As an airline, it is crucial to know what our customers are thinking. We want to know who is satisfied with Gulf Air’s service, who is not, if not then why, and what Gulf Air can do to improve its offering. Analysing social media can help us gain more insight into potential customer demands for everything from on-board services to new destinations, information which can be incorporated into decision-making to enhance the airline’s offering,” said Dr. Jassim Haji, Director of Information Technology, Gulf Air.
Arabic Sentiment Analysis is based on Cloudera’s distribution of the open source Hadoop big data framework, running across a cluster of servers based in Gulf Air’s private cloud IT environment. This, in turn, uses a number of Red Hat products, including Red Hat Enterprise Linux, Red Hat JBoss Enterprise Application Platform, and Red Hat Storage. The private cloud encompasses 200 servers running more than 100 core applications and holds more than 50 terabytes of data. The scalability achieved using virtualisation meant there was plenty of capacity to host the big data solution.
The private cloud is based on a dedicated, state-of-the-art “green” data centre and has been in place since 2012. During its planning, Gulf Air’s IT team made the strategic decision to migrate all the airline’s UNIX platforms to Red Hat Enterprise Linux. Red Hat’s technologies fully integrate with all other major infrastructure components that Gulf Air uses, from companies such as EMC, Cisco, and F5 Networks.
Today, Arabic Sentiment Analysis sits on this private cloud infrastructure and processes hundreds of thousands of social media posts, providing sentiment results in minutes that indicate what passengers are talking about and how they feel about particular routes, products, or services. The solution also uses social media posts as the basis of wider analysis of the state of the market and actions taken by Gulf Air competitors.
Increased competitive advantage
Arabic Sentiment Analysis lets Gulf Air provide passengers with offers and promotions on a timely basis, often in response to competitor actions. Real-time reports, for example, alert Gulf Air marketing executives when a competing airline launches a particular promotion or offer involving a destination that Gulf Air flies to.
“The airline industry is very price-sensitive, and the majority of travellers look for the cheapest fares available. During peak travel times like summer, quick action is crucial, because any delay in responding to the competition could mean the loss of passengers and revenue. Such real-time reports have helped Gulf Air to respond very quickly to its competitors,” said Dr. Haji. Using Arabic Sentiment Analysis has meant improved time to market for new offers and promotions, as well as improved sales cycles.
Big data made reliable
Gulf Air has been able to tackle a reliability issue inherent to many big data implementations. Their big data solution is based in its Red Hat private cloud, it was easy for IT staff to create “ghost images” of the master server, which can be deployed in minutes if the original master server fails.
Faster response times
Social media monitoring and analysis, when performed manually, can be slow and resource-intensive, but with Arabic Sentiment Analysis, it is “automatic, quick and efficient,” says Dr. Haji. New posts and changes in social media trends are periodically captured and stored locally, and then transformed into reports that make it easy for Gulf Air employees to act on their findings. This gives them a good, up-to-date view of any changes in sentiment, shifts in demand, or emerging market trends, enabling them to respond quickly.
The journey ahead
Many big data projects come with a hefty bill for the companies concerned. At Gulf Air, however, Arabic Sentiment Analysis has been delivered very cost-effectively, thanks in part to the company’s enthusiasm for open source solutions.
“The big data solution deployed is an open source one, which meant no special license fees to pay, while the infrastructure where it runs is based on an existing open source-based private cloud environment and virtual servers, which meant no hardware investment was required. Open source technologies have many other advantages, including wide community support and quick and frequent fixes and new releases,” said Dr. Haji.
