28
Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project in Artificial Intelligence – 30ECTS) Proposal Form Title: Learning Named Entities in Maltese from other languages and the Crowd Project Supervisor: Dr Claudia Borg Project Co-supervisor: (if applicable) Main Subject Area/s: Named Entity Recognition, NLP, Machine Learning Brief Project Description inc. References: (word limit approx. 300 words) Maltese is considered a low-resourced language. One of the aspects where we still don’t have any data for Maltese is in Named Entities. Named Entities are used in systems such as information retrieval, and they identify chunks of text that refer to Person, Location, Organisation and other such entities. The aim of this project is two-fold: (i) use machine learning techniques to transfer knowledge about Named Entities from other languages to Maltese and, (ii) to use the crowd to verify the Named Entities identified. There are a number of Machine Learning techniques that can be used (e.g. Cotterell and Duh, 2017). Part of the research phase will be used to identify exactly which machine learning techniques are most appropriate and why. The project will then focus on the application of the machine learning techniques and then the evaluation of Named Entities through the crowd. Crowdsourcing evaluation is a common approach in scenarios where data is not available for an automatic evaluation. Cotterell, Ryan and Duh, Kevin. (2017) Low-Resource Named Entity Recognition with Cross-lingual, Character-Level Neural Conditional Random Fields in the Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 2: Short Papers), Asian Federation of Natural Language Processing

Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Learning Named Entities in Maltese from other languages and the Crowd

Project Supervisor: Dr Claudia Borg

Project Co-supervisor: (if applicable)

Main Subject Area/s:

Named Entity Recognition, NLP, Machine Learning

Brief Project Description inc. References: (word limit approx. 300 words)

Maltese is considered a low-resourced language. One of the aspects where we

still don’t have any data for Maltese is in Named Entities. Named Entities are

used in systems such as information retrieval, and they identify chunks of text

that refer to Person, Location, Organisation and other such entities.

The aim of this project is two-fold: (i) use machine learning techniques to

transfer knowledge about Named Entities from other languages to Maltese and,

(ii) to use the crowd to verify the Named Entities identified. There are a number of Machine Learning techniques that can be used (e.g.

Cotterell and Duh, 2017). Part of the research phase will be used to identify

exactly which machine learning techniques are most appropriate and why. The

project will then focus on the application of the machine learning techniques

and then the evaluation of Named Entities through the crowd. Crowdsourcing

evaluation is a common approach in scenarios where data is not available for

an automatic evaluation.

Cotterell, Ryan and Duh, Kevin. (2017) Low-Resource Named Entity Recognition with Cross-lingual, Character-Level Neural Conditional Random Fields in the Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 2: Short Papers), Asian Federation of Natural Language Processing

Page 2: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Resources Required: n/a

Recommended Prerequisites / Knowledge Required and Supporting 3rd Year Study -units:

ICS2203, LIN3012, ICS3206.

Foreseeable Ethical Issues and How these will be tackled: (if applicable)

No personal data will be collected from the Crowd. They will simply participate

through an online website by accepting or rejecting proposed Named Entities.

Page 3: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: MorphoTest

Project Supervisor: Dr Claudia Borg

Project Co-supervisor: (if applicable)

Main Subject Area/s:

NLP, Machine Learning

Brief Project Description inc. References: (word limit approx. 300 words)

This project is aimed towards creating an educational tool to assist learners of Maltese practice the grammar of the language. The machine learning aspect will focus on two things: (i) clustering types of grammatical components together and (ii) determine which parts of a word represent the grammatical component (e.g. fjura - the ‘a’ at the end indicates that the word is in the feminine and therefore would be followed by sabiħa rather than sabiħ. The interface to create language tests can be done through either a website or a chat service like Telegram, which offers an API to create and develop a number of services. In fact, recently Telegram has also been used by Duolingo, a company offering online language learning. The setup of the architecture will also offer the possibility to test the accuracy of our current resources in morphology and to possibly highlight entries which should be further examined by language teachers. This aspect is often referred to as implicit crowdsourcing. Lange, Herbert and Ljunglöf, Peter. 2018. MULLE: A grammar-based Latin language learning tool to supplement the classroom setting in the Proceedings of the 5th Workshop on Natural Language Processing Techniques for Educational Applications. Association for Computational Linguistics

Resources Required: n/a

Page 4: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Recommended Prerequisites / Knowledge Required and Supporting 3rd Year Study -units:

ICS2203, LIN3012, ICS3206.

Foreseeable Ethical Issues and How these will be tackled: (if applicable)

No personal data from participants is required for this project.

Page 5: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Analysis of Aviation Safety and Aviation Accidents

Project Supervisor: Dr Joel Azzopardi

Project Co-supervisor:

(if applicable)

Main Subject Area/s: Information Extraction, Data Mining

Brief Project Description inc.

References:

(word limit approx. 300 words)

In today’s world aviation has become a crucial means of long-distance

transportation. It is not surprising therefore that aviation safety is taken very

seriously and aviation accidents are investigated very thoroughly on a case-by-

case basis.

The Aviation Safety Network (https://aviation-safety.net) provides a few

databases about aircraft accidents with details descriptions of each. Moreover,

aggregated statistics about airline incidents may also be found here:

https://github.com/fivethirtyeight/data/tree/master/airline-safety

Information about the different crashes may also be found on news portals and

on Wikipedia (e.g. https://en.wikipedia.org/wiki/Ethiopian_Airlines_Flight_302)

The aim of the dissertation is to analyse this data, provide aggregations and

visualisations of this data, and identify common aspects behind the different

accidents. Association rule mining can also be performed to identify what

typically goes wrong in the different types of accidents.

Resources Required: Personal Computer, Internet Access

Page 6: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

ICS2205 – Web Intelligence

ARI3216 – Web Data Mining

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

N/A – this dissertation is envisaged to make use of publicly available datasets

Page 7: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: What is the underlying cause behind greenhouse gas emissions?

Project Supervisor: Dr Joel Azzopardi

Project Co-supervisor:

(if applicable) Dr Adam Gauci

Main Subject Area/s: Information Extraction, Data Mining

Brief Project Description inc.

References:

(word limit approx. 300 words)

Climate change is one of the biggest challenges facing modern mankind.

Green house gas (GHG) emissions are causing temperature increase which

may in turn cause certain countries to become uninhabitable. Unfortunately,

environment conservation and greenhouse gas reduction practices may be

perceived to hinder progression and economic growth. Within the local

perspective, Malta recently has been enjoying strong economic growth.

However, it has also registered increased GHG emissions.

Data concerning environmental and climate change is nowadays increasingly

available. For instance datasets published by the World Resources Institute

includes the CAIT - Country Greenhouse Gas Emissions Data that shows GHG

emissions sorted by year, country, emissions type and sector

(http://datasets.wri.org/dataset/cait-country). On the other hand, datasets from

EuroStat (https://ec.europa.eu/eurostat/data/database) show economic

indicator values for the different countries.

This dissertation involves the aggregation of different datasets, and their

automatic analysis to identify the relationships between GHG emissions and

other factors (such as economic growth). Clustering may be applied on the

different datasets to identify ‘similar’ countries and underlying common features

within certain clusters (e.g. high-emitting countries) identified.

Page 8: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Resources Required: Personal Computer, Internet Access

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

ICS2205 – Web Intelligence

ARI3216 – Web Data Mining

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

N/A – this dissertation is envisaged to make use of publicly available datasets

Page 9: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Game Interactions in a CAVE Environment

Project Supervisor: Dr V. Camilleri

Project Co-supervisor:

(if applicable) Prof. V. Briffa (Department of Digital Arts)

Main Subject Area/s: Virtual Reality, CAVE Virtual Environment, Game AI, HCI, Interface Design

Brief Project Description inc.

References:

(word limit approx. 300 words)

Virtual Reality is a term coined to describe an immersive virtual space that uses

senses such as sight, hearing, and touch to evoke in users the illusion of

presence, space and being. The CAVE is a cubic room, served by projection

screens, enabling users to walk around, hold objects and interact in a

controlled environment. This project proposes the creation of a game, that

makes use of captured interactions to provide feedback and generate game

scenes accordingly. The game narrative and interactions are part of the

research that is expected to arise from this project.

Resources Required: CAVE Virtual Environment, Unity for Development

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

Computer Vision and Image processing

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

No foreseeable ethical issues

Page 10: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: TAR – on the job Training using AR

Project Supervisor: Dr V. Camilleri

Project Co-supervisor:

(if applicable)

Main Subject Area/s: Augmented Reality, Machine Learning, Professional Development

Brief Project Description inc.

References:

(word limit approx. 300 words)

Augmented Reality (AR) technology is increasingly becoming more popular,

more accessible and more pervasive into the workforce. Training using AR is

also gathering momentum amongst the industry. Several previous studies have

taken into account the possibility of using virtual environments for simulation

and training. However, AR can offer a higher fidelity, just in time approach to

training that other technologies may not offer. What AI can add on to AR is the

possibility of enabling real-world object tagging and predicting the interface that

the user would need given the context of the on-the-job training, leading to a

responsive virtual experience. Using for example the IBM Watson Unity SDK,

this project can add cloud-based AI to the AR application.

Resources Required: AR Headset, Unity

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

Computer Vision and Image processing

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

No foreseeable ethical issues

Page 11: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: A gamified citizen science approach for the collection of data to train deep

learning models

Project Supervisor: Mr. Dylan Seychell

Project Co-supervisor:

(if applicable) Prof. Maria Attard

Main Subject Area/s: Computer Vision and Machine Learning

Brief Project Description inc.

References:

(word limit approx. 300 words)

Deep neural networks have been proved to do very well in image classification.

However, they require a large volume of training data to classify objects

effectively. This approach works well when datasets are available or objects

being classified are general enough to be found in easily labelled datasets. A

scenario such as the classification of Maltese flora [1] or fauna poses several

challenges, such as the limited annotated data available.

In this project, we are proposing a gamified approach to the collection of data

for the training of deep learning models through citizen science. Citizen

Science is research that is conducted with the help of non-professional

scientists [2] to enable a more comprehensive data collection.

The project includes the development of a mobile application that will allow for

the annotation of data through a gamified process while also providing

information to users about the subject matter. The users will also be organised

by level of confidence in the subject matter. The information collected from the

usage of the mobile app will be used to train a model on the cloud that

improves the classification model.

Google’s Firebase ML Kit will be used for the creation and management of the

model, together with its interfacing with the mobile app, also developed on the

same framework. The use of this technology allows project can focus on the

process. ML Kit also provides evaluation tools that will be used to monitor the

performance of the process and the subsequent evaluation of the entire

Page 12: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

concept proved by this project.

[1] Mifsud, S, Malta Wild Plants,

http://www.maltawildplants.com/colourindex.html, accessed June 2019

[2] Gura, T (2013). "Citizen science: amateur experts". Nature. 496 (7444):

259–261

Resources Required: Access to the Google Cloud Platform, available for free within the context of

this project.

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

- ARI2129 - Introduction to Artificial Vision

- ARI3129 - Advanced Computer Vision for AI

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

NA

Page 13: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Presentation of Temporal Information using Spatial Augmented Reality

Project Supervisor: Mr. Dylan Seychell

Project Co-supervisor:

(if applicable) NA

Main Subject Area/s: Computer Vision and Interaction Design

Brief Project Description inc.

References:

(word limit approx. 300 words)

Projection Augmented Reality or Spatial Augmented Reality is the AR method

of projecting a computer graphic onto the real stationery world objects. Such

an approach is limited by the projection range of the projector being used and

the field of view of the supporting camera [1].

Spatial Augmented Reality (SAR) provides a seamless experience for its users

since it does not require a screen or a head mounted display to the user since

everything is projected onto real-world objects.

This first component of the SAR system that reads a physical analogue clock

and processes the time displayed. This result of the first component is then fed

into the second module that would be able to compare the time read from the

clock separate time-related information in a database. This information may

include appointments, traffic or weather information, among others. The

system would then make use of its third component that will project upcoming

information on the space around the physical clock in pre-established graphics.

The projection of supporting information will take into consideration the

dimensions of the clock.

[1] M. R. Marner, R. T. Smith, J. A. Walsh and B. H. Thomas, "Spatial User

Interfaces for Large-Scale Projector-Based Augmented Reality," in IEEE

Computer Graphics and Applications, vol. 34, no. 6, pp. 74-82, Nov.-Dec. 2014.

Page 14: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Resources Required:

Pico Projector

Webcam or Raspberry Pi + Camera

(Both to be provided by the project supervisor)

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

- ARI2129 - Introduction to Artificial Vision

- ARI3129 - Advanced Computer Vision for AI

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

NA. The data visualized on the system will be fictitious data.

Page 15: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Implementations of the State Merging Operator in DFA Learning

Project Supervisor: Kristian Guillaumier

Project Co-supervisor:

(if applicable) John Abela

Main Subject Area/s:

Machine Learning, Grammatical Inference, Search Algorithms, Heuristics,

Algorithm Design

Brief Project Description inc.

References:

(word limit approx. 300 words)

Grammatical inference is the task of learning a formal grammar from strings

which belong to a language and strings which do not belong to a language. In

this project we are concerned with the identification of regular languages (as

Deterministic Finite State Automata) from training sets consisting of both

positive and negative examples. Specifically, we are interested in identifying

the smallest DFA which is consistent with the training data. This inference task

has many real-world applications including robotics, data mining, structural

pattern recognition, speech recognition, and bioinformatics.

One of the most studied techniques to perform regular inference is using a

class of methods called State Merging Algorithms which repeatedly merge

states in a DFA together until a final hypothesis is reached. Implementations of

the state merging operation are expensive so it then follows that

experimentation and analysis becomes more restrictive and impractical

especially as the target DFAs we are looking for become larger. The research

hypotheses addressed in this project are:

- Can we design data structures and algorithms to make the merge operator

faster and more efficient?

- How do our data structures and algorithms perform compared to baseline

implementation of the state merging operator?

Page 16: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Resources Required: Material will be provided to the student to support the FYP

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

Data Structures and Algorithms 1 & 2, Machine Learning 1 & 2

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

None

Page 17: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Using Genetic Programming to Evolve Heuristics for DFA Learning

Project Supervisor: Kristian Guillaumier

Project Co-supervisor:

(if applicable) John Abela

Main Subject Area/s:

Machine Learning, Grammatical Inference, Evolutionary Algorithms, Genetic

Programming, Search Heuristics

Brief Project Description inc.

References:

(word limit approx. 300 words)

Grammatical inference is the task of learning a formal grammar from strings

which belong to a language and strings which do not belong to a language. In

this project we are concerned with the identification of regular languages (as

Deterministic Finite State Automata) from training sets consisting of both

positive and negative examples. Specifically, we are interested in identifying

the smallest DFA which is consistent with the training data. This inference task

has many real-world applications including robotics, data mining, structural

pattern recognition, speech recognition, and bioinformatics.

One of the most popular algorithms which deals with the regular inference task

is the Evidence Driven State Merging (EDSM) algorithm which uses a heuristic

to drive a search from an initial hypothesis to a final result. Current research

indicates that while the EDSM heuristic works in certain cases, other heuristics

work better in others. In this project we will be using genetic programming

techniques to evolve a number of heuristics which together perform better than

EDSM by itself. The research hypotheses addressed in this project are:

- Can we design genetic programs which generate heuristics which perform

better than EDSM?

- To which extent do these heuristics perform better or worse?

Resources Required: Material will be provided to the student to support the FYP

Page 18: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

Data Structures and Algorithms 1 & 2, Machine Learning 1 & 2

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

None

Page 19: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Automated Gait Analysis

Project

Supervisor: Alexiei Dingli

Project Co-

supervisor:

(if applicable)

Main Subject

Area/s:

Indicating how it is relevant to the AI degree

Brief Project

Description

inc.

References:

(word limit

approx. 300

words)

Current Movement analysis

Current technology in the analysis of movement, both in research and in the clinical field requires a lot

of manual input, both for the capturing of the data but even more so for the processing and

interpretation of that collected data. In the clinical field it is used primarily to assess complex

movement abnormalities with the intent of quantifying such issues and assist in medical or surgical

intervention. In research, it is mostly used to ‘diagnose’ related or common aspects of disease related

to movement, quantify changes post intervention (ex. drugs or surgery), and to scientifically assess,

quantify and improve interventions that may alter movement.

This is normally referred to as Clinical Gait (or Motion) Analysis, abbreviated as CGA.

Coincidentally, Computer Graphic Animation also abbreviated as CGA, normally uses the same

apparatus, method of capturing and processing data.

Current technology is heading towards, ‘biometric’ analysis of human movement for security

purposes, collation of data of movement disorders for the implementation of specific movement deficit

profiling, On the fly functional stimulation of limbs using captured data from non-affected limbs, etc.

Capturing

In the present, data is collected at anywhere between 100Hz and 2000Hz in the form of coordinates in

a known volume. Commonly, this can be done in one of two ways;

Using RadioFrequency emitters placed on various parts of the body, whose position is

Page 20: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

detected precisely by a receiver which is able to calculate the position of each emitter in

space. Then using an array of accelerometers and gyroscopes also work out the

orientation. This works on the concept of a localised GPS system. These have the benefit

of ease of use and lack marker screening, but are affected by other radiofrequency and

are less accurate.

Using Retroreflective markers on specific parts of the body, whose position is detected by

a series of synced IR cameras. Using triangulation methods, each marker corresponding

to a body location is worked out on the fly by a dedicated system, which results in very

accurate detection of quantified movement between all different moveable body

structures, after all the data is collated. This method is more precise, but to the detriment

that marker detection may be impeded by body structures during movement.

In both cases, data is exported in the form of integers/values that correspond to the

location in a specified volume in the three planes of motion (X,Y,Z coordinates).

Processing

The data captured in raw format requires processing to produce plots of all joints in all planes of

motion, that are used to produce inferences during interpretation. The processing involves detecting

and naming each marker according to the position it is placed, filling any gaps in data, removing any

artefacts, detecting specified events in the walk, filtering and smoothening the data, using specific

biomechanical models to calculate the angles accordingly, and finally producing a graphical output

for each walk. This process is usually done by a specifically trained biomechanical engineer. Locally

this is being done by 2 kinesiologists

Plotted graphs are also superimposed over data taken from normal individuals as guidance. Normally this data is plotted in a way so that the clinician can then proceed to interpret the data.

Page 21: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project
Page 22: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Interpretation

In clinical scenario, the plotted data, together with video-graphical imaging, pictures and physical examination are used to come up with a movement profile of that particular subject. The discrepancies in the plots are scoured and compared to normative data can be suggestive of patterns of erratic movement (like spasticity or tonicity issues), tightened structures (like ligaments and muscles), non-functional movement units, abnormal growth (like rotated bones and joints), or abnormal timing of activity. This is extremely time consuming and may require 20 to 40 hours per patient

In the research scenario, interpretation of data is slightly different in that since a study usually investigates one particular condition, this normally requires collation, averaging and statistical analysis of similar data, where the researcher knows already where the focal point is.

There may be an area for development of novel systems of analysis. Considering that normal subjects, although slightly different, walk in much the same way, any abnormal pattern should easily be detected when it falls out of the normative barriers.

Generally speaking, when 12 or more normal subjects’ walking data are processed and ‘averaged’, the resultant averaged graphs and standard deviation bands are enough to encompass most disease free walks. In other terms, even though everyone walks differently, and has his/her own specific walk, the pattern of events, once normalised to time, is grossly the same. When disease or other medical conditions cause some form of impairment, the highly efficient, explicit pattern of walking is immediately disrupted and such disruptions may easily be detected and flagged by an ‘automated’ system.

The scope of this thesis is to detecting, flagging and quantifying such impairments the data provided.

Resources

Required: Anonymized data will be provided by Mater Dei.

Page 23: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Recommended

Prerequisites /

Knowledge

Required and

Supporting 3rd

Year Study -

units:

Machine Learning and Statistics

Foreseeable

Ethical Issues

and How these

will be tackled:

(if applicable)

No since all the data is anonymized.

Page 24: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Interactive Chatbot

Project Supervisor: Alexiei Dingli

Project Co-supervisor:

(if applicable)

Main Subject Area/s: Intelligent Interfaces, Conversational Agents

Brief Project Description inc.

References:

(word limit approx. 300 words)

The idea behind the Interactive Chatbot is to make use of an existent famous

portrait (such as the Mona Lisa) and make her talk. The student needs to

create a sophisticated chat bot capable of engaging in a meaningful

conversation with the user. Students are encouraged to use existent

frameworks such as ChatScript, the open-source Natural Language scripting

language and engine running bots, which successfully managed to win the

Loebner Prize.

The system should also make use of the camera on the mobile device in order

to identify people and make the painting interact with the people and the

environment in front of it.

The final deliverable in this case is a painting running from a mobile device and

capable of conversing with the users. The topics discussed should be related to

the environment in which it is placed and the painting i.e. the context in which it

was painted, the artist, the subject of the painting, etc.

Resources Required: Mobile Device

Page 25: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

Intelligent Interfaces, NLP

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

No

Page 26: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: AI in Education Resource

Project Supervisor: Prof Matthew Montebello

Project Co-Supervisor: Dr Vanessa Camilleri

Main Subject Area/s:

AI in Education;

Augmented Reality;

Internet Technologies.

Brief Project Description inc.

References:

(word limit approx. 300 words)

Teaching AI to young students is a challenge that few have ventured due to the

difficulty of scoping down such complex issues to the level of kids. Through the

use of Online resources as well as Augmented Reality this task can be

rendered easier while at the same time ensuring that real-time simple machine

learning can be achieved for students to try out and experience as part of a

learning programme.

Resources Required: Smart phones are available at the AI department.

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

Knowledge acquired through the ARI2131 – AI in Education 2nd

year study-unit

would be an asset.

Interesting link: http://cognimates.me/home/

Yet another: https://machinelearningforkids.co.uk

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

No third parties will be involved except for testing purposes.

Page 27: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Drone Intelligence

Project Supervisor: Prof Matthew Montebello

Project Co-supervisor: Dr Conrad Attard

Main Subject Area/s:

Drones or unmanned aerial vehicles are ever so common with numerous

intelligent applications that can provide services and assistance to human in a

variety of ways.

Brief Project Description inc.

References:

(word limit approx. 300 words)

Three or more different projects can potentially be considered within this area

as students can propose their own intelligent application to research & develop

for their FYP. The three proposals are:

i. Image-based destination finder: Given an image within a specific

distance the drone can locate through image recognition and move

close to it;

ii. Swarm formations & tasks: Artistic designs, word formations or even

symbols formed in the sky at a specific location as instructed by the

user from a central application;

iii. Personal butler: One or more drones identify a person and hover within

a specified distance providing images, security, and gesture

recognition to perform specific tasks.

Resources Required: Programmable DJI drones are provided by the AI department but students are

encouraged to use their own devices as well.

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

Knowledge acquired through the GAPT about Drones would be an asset.

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

If the images involve third parties then ethical clearance will be required.

Page 28: Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 … · 2019. 10. 11. · Faculty of ICT B.Sc. I.T. (Hons.) in Artificial Intelligence ICT3909 (Final Year Project

Faculty of ICT

B.Sc. I.T. (Hons.) in Artificial

Intelligence

ICT3909 (Final Year Project in

Artificial Intelligence – 30ECTS)

Proposal Form

Title: Student Agents

Project Supervisor: Prof Matthew Montebello

Project Co-Supervisor: Dr Vanessa Camilleri

Main Subject Area/s:

Intelligent Agent Systems;

AI in Education;

Virtual Reality.

Brief Project Description inc.

References:

(word limit approx. 300 words)

Student agents can simulate different behavior within a virtual classroom for a

student teacher to experience as well as take remedial action to maintain

control of the class. Through the use of VR technologies it would be possible to

create such an instructional environment and log the human teacher in training

actions to gain insights and harmless experience of typical disruptive students

within a classroom.

Resources Required: VR kits are available at the AI department and will be investing in new kits.

Recommended Prerequisites /

Knowledge Required and

Supporting 3rd

Year Study -

units:

Knowledge acquired through the ARI2131 – AI in Education 2nd

year study-unit

would be an asset.

Foreseeable Ethical Issues and

How these will be tackled:

(if applicable)

No third parties will be involved except for testing purposes.