Computational Approaches to Language Research Workshop

This meeting provides an opportunity for researchers across several disciplines to meet and compare approaches to computational research with language.

It will include three informal presentations and refreshments.

Natural Language Processing and Big Data Analytics for Oil and Gas Supply Chains

Arda Gezdur

Accurate oil price forecasts are essential for better supply chain management in the petroleum industry. High fluctuation in oil prices significantly impacts the profitability of oil companies, nations, and even regions on a global scale. The risks associated with oil price variations can be alleviated through improvements in forecasting capabilities.  This work-in-progress applies natural language processing (NLP) techniques on qualitative data from daily oil industry reports, Twitter and Google Trends. The data are trained via machine learning algorithms to improve the accuracy of the oil price forecasts using multivariate regression analysis. NLP techniques are employed to provide a solution for a refinery in Europe, applying design science research methodology which aims to iteratively develop solutions to problems in practice. The study aims to develop a decision support tool to assist companies in the petroleum industry with their production planning and supply chain management decisions.

Arda is a PhD candidate working with Dr Jyoti Bhattacharjya and Prof. Rico Merket in the Sydney Business School’s Institute of Transport and Logistics Studies

Analysing Preventative Mobile Health Messages and their Responses

Di Lu

About 90% of heart attack worldwide can be accounted for by modifiable risk factors. This is why the Westmead Applied Research Centre has successfully trialled using regular text messages to help patients reduce their cardiovascular risk. After several such trials, Professor Clara Chow’s team wanted to better understand the messages they sent, which encourage participants to choose a healthy lifestyle or to seek support, and the responses from participants. Sydney Informatics Hub supported them in grouping together messages, providing better measures of message stylistics, classifying participant responses into categories, and understanding which messages received responses. This analysis lays groundwork for more personalised and more scalable approaches to messaging for preventative health.

Di Lu is a Data Science Software Engineer at the Sydney Informatics Hub.

Quantitative SAS user meets qualitative text data!

Jacques Raubenheimer

Coming from biostatistics, Jacques found himself working with textual data when social media was an interesting way to study drug users and usage, and when supposedly quantitative data turned out to be textual and messy. He will share some of his early explorations in grappling with textual data and is very keen to hear about what he could do differently.

Dr Jacques Raubenheimer is a senior research fellow in biostatistics in Pharmacology at the School of Medical Sciences