Internship available: Online Social Networks and Privacy-Preserving Machine Learning Applications

HOR-INTERN-2018-01

To apply, please complete the Internships Application Form.

Closing date

Applications must be made before the 20^th December and are assessed on an ongoing basis.

Proposed dates

The Internship will run for up to 6 months. Anticipated interview and start date January 2019. Please indicate your availability on the application form, where requested.

Overview

Online social networks (OSNs) have become key platforms for social interactions. Interactions that leave a personalised digital footprint, which can either be private or public. In an OSN, private data exchanged in is directly available for processing from the platform itself, while publicly shared data is available for processing by everyone. We argue that privacy preservation should be the key aspect of data processing, especially when dealing with human generated content.

We aim to analyse and implement privacy aware machine learning methods based on natural language. Our goal is to introduce a layer of privacy when dealing with human generated inputs, and perform a multiclass classification using the latest machine learning methods.

The subjective nature of text is best represented by a range of possible classes expressed through multi-class and multi-label models. In Machine Learning, multi class refers to the output and the multiple classes to be predicted, while multi label refers to the range of properties to be predicted, such as type of item and its colour.

We aim to develop a model which combines both types of methods mentioned for text analysis purposes. The model will be fed (input) a range of text properties on a unigram basis and will predict (output) a range of properties for groups of unigrams.

The model will find usage in human generated text applications, regardless of the type of the underlying text properties needed. Example applications are sentiment and medical transcript analysis. However, as a pilot study we will be focusing in news articles and the emotion conveyed through them.

We are looking for someone to develop the required tools on the fields of sentiment analysis and multiclass/multilabel machine learning. You will be working within Horizon Research group, closely collaborating with our researchers.

Applicants must outline their area of interest, experience and skills.

Who should apply?

Ideal applicants should be studying for a postgraduate degree in computer science or related discipline and have experience in Python, Natural Language Processing and Machine Learning, with an avid interest in conducting research in these topics.

Required skills

• Programming experience in Python (nltk, scipy, etc.)
• Background in Machine Learning or related area
• Ability to work independently as well as part of a team

Desirable skills

• Experience of working with Crowdsourcing and Natural Language Processing
• Experience with academic writing

Eligibility and financial aspects

This is a full-time internship for up to 6 months. For postgraduate students who receive a stipend from their home university during the internship, a bursary of £300 per week will be available. For postgraduate students who suspend their stipend, a casual wage of £350 per week will be available, and this may be subject to tax deductions depending on the successful candidate’s circumstances.

In general, students from The University of Nottingham are able to apply on the understanding they suspend their stipend, this is due to the nature of the funding source. For overseas students a Visa must be in place, covering the duration of the internship.

Internships will be based at Jubilee Campus, The University of Nottingham (NG7 2TU), and may not be undertaken remotely. 

Informal enquiries

Informal enquiries may be made to Giannis Haralabopoulos however applications should be made via the Internships Application form – applications to this email address will not be accepted.