Informatics Institute of Technology better known as IIT organized the Technical Forum on Big Data Analytics. The event was held on the 10th of May 2016 at BMICH premises. Things started off with registration followed by a welcome speech delivered by the CEO of IIT, Dr. Sampath Kannangara. Immediately afterward was a speech by Professor John Harper.
The main speeches on the topic of Big Data Analytics were initialized by Professor Frans Coenen from the Department of Computer Science, University of Liverpool. His speech was on the topic of “Knowledge Discovery in Image Data: A Practitioners View”. Through his speech, he managed to point out how two-dimensional sets of images are compared in order to identify age-related macular degeneration. This is currently being done by experts at hospitals as computers are yet to be involved in the process as it is still a work in progress.
The next instance Professor Coenen shared with the audience is the observation made by Google in the UK, where census data could be gathered by taking a look at the content of the Google satellite images. This worked and was conducted in rural areas of Ethiopia with high levels of accuracy.
The next instance which was shared was the data from a company which is trying to prevent Osteoporosis using software. This is achieved through analyzing the texture of images. Texture analysis of images can be used in brain scans to test for certain conditions such as epilepsy to a certain level. He also mentioned of the usage of this type of analysis to distinguish between musicians and non-musicians.
He also spoke of a machine that pushes out shapes using a process known as asymmetric incremental sheet metal forming, which was part of a large-scale European project to build parts required for aircraft and space research. The task here was to identify the amount of distortion created by the machinery which pushes out the shapes. Among the techniques which can be utilized are statistical technique, local binary patterns, and KNN. From this statistical technique is the simplest and the most widely used technique. It was made clear in the presentation that whichever representation used doesn’t have to be interpretable by humans but rather be used by machines for analysis in the form of binary data.
After Professor Frans’s speech, it was time for Dr. Nirmalie Wiratunga to deliver the speech titled “Sentiment-driven Knowledge Discovery from Social Media”. She started her talk by explaining what sentiment analysis is, going on to describe it as the computational study of opinions expressed in textual form. Here the interest lies in finding out whether the opinions contain subjective content and can it be differentiated from objective content. If it is subjective then it has to be identified as positive or negative. This type of analysis is useful for various types of industrial applications.
The answer to the question of why this sort of analysis might be useful would be because there is an exponential growth of content generated on social media over the past couple of years. Statistically, that is almost 350,000 tweets in a minute, on YouTube, there are almost 3 million videos watched every minute.
According to Dr. Nirmalie, the involvement of natural language processing, information retrieval, and machine learning, interests people to do research on this area. Main methods which can be used to do sentiment analysis was introduced to the audience as lexicon-based approach and machine learning approach. Out of which she used the lexically based approach for her presentation.
What the lexical approach uses for this purpose is a previously defined set of data which is fed. In this case, it is the English dictionary which acts as the data which is fed for the sentiment analysis. The nonlexical approach was also mentioned to discuss the use of additional letters, emoticons or punctuation to further show expressions. Emoticons glossary will be used to label positive or negative tweets. This will largely benefit in many areas but in the cases, she pointed out it is mainly suggesting to buyers which items to purchase based on the reviews posted by those who have purchased those items before.
It was then time for Professor Christina Jayne to present to the audience about her speech titled “Deep Learning and Its Applications”. Labeled ‘data and unsupervised machine learning aids in the task of classification’ according to the Professor. Raw data is used for abstract representation and to keep learning from past results.
This is related to the area of Neural Networks which basically is replicated from none other than our own brains. Deep learning has spread into speech and text recognition areas. Google’s acquisition of the AI Company ‘Deep Mind’ with the aim of creating general purpose AI, was also mentioned by her to show the growth in the subject of deep learning over the past couple of years. Systems can be trained faster with the involvement of neural networks. Other applications of deep learning include speech recognition, optical image recognition, and mold data.
Dr. Srinath Perera was the final speaker for the evening, his topic was “Machine Learning in the Real World” and he shared with the audience real world cases which he has come across. The first is from transportation in London where machine learning can be used to give real time results. He showed the importance of using machine learning in instances such as identifying handwritten text, where he did not fail to mention the attempts made by IBM to achieve voice and character recognition by understanding how language works. One such instance where machine learning was used by him and his team is to determine the winner of the US election based on tweets. Another one being the analyzing of airport activity.
After the Q&A session, the forum was concluded. Overall it was a success considering the amount of people who were keen on Big Data Analytics filling up the entire hall. There was definitely a lot of things which we were able to learn out of it.