Reference no: EM133905620

Intelligent Systems

Learning Outcome 1: Determine suitable approaches towards the construction of AI systems.

Learning Outcome 2: Determine ethical challenges which are distinctive to AI and issues that may arise with such rapidly developing technologies.

Learning Outcome 3: Apply knowledge based or learning based methods to solve problems in complex environments that attempt to simulate human thought and decision making processes, allowing modern society to make further advancements.

Learning Outcome 4: Communicate clearly and effectively using the technical language of the field and constructively engage with different stakeholders.

Learning Outcome 5: Apply the foundational principles of AI learnt throughout the course and apply it to the different areas of Natural Language Processing, Speech Recognition, Computer Vision and Machine Learning.

Task Summary
In a group (approximately 3 or 4) you should apply the foundational principles of AI to a Natural Language Processing (NLP) or Computer Vision project capable of solving a specific problem. There are two problems defined for this task - and you will need to choose one. The NLP based task is where you can implement a sentiment analysis project. The Computer Vision project will provide you an opportunity to train a model based on sample data that will let you perform a a self-driving simulation without crashing or leaving the road.

If you choose the NLP-based task, your solution should be delivered as a simple website with a text box to enter a sample statement for sentiment analysis and a button to execute the sentiment analysis function. The interface should also present the outcome of executing the sentiment analysis function on the page. Note that you will be training a machine learning model to analyse the sentiments of customers reviews and creating a prediction function to allow for the input text to be subject to sentiment analysis based on your trained model.

With the Computer Vision project, there is also a need to train a model that the simulator will use to run a simulation based on sample data. The model (which is trained on this sample data) will need to be submitted as a deliverable along with a video of a full lap of the car doing a lap in the simulator. Boost your grades with expert-level assignment help today.

Context

This assessment moves further into solving a more realistic program by building a more complex Intelligent System. The Intelligent System can either be an application in Natural Language Processing or Computer Vision - two of the key focus areas in industry and academia.
The project will also give you an opportunity to hone your skills and be able to collaborate with other individuals in a team. Collaboration is common in the workplace, and therefore a skill worth practising. The group work in this assessment will help you to identify the skills you might need to refine and help you to understand how to communicate better with your team mates. There is also a presentation and report deliverable that will help you practise your verbal and written communication skills as these will prove to be vital in the workplace.

Task Instructions

You should work in a group of 3 or 4 people (depending on the numbers in your class) and the tasks of each person should be determined at the beginning of your project. This is important to ensure expectations of individual contributions are set. You are required to use the version control tools Git that can also keep a track of collaboration between members. You also need to deliver a presentation that should be no longer than 15 minutes, and is based on the project you have implemented together. Individually, you must also prepare a report explaining each team member's contribution to the project (250 words). The individual report explains the contribution each person made to the assessment task. Finally, you should include a self-assessment of your perceived percentage contribution to the overall assessment task, and how much each team member contributed. Further instructions and detail are provided below.

To complete this assessment task you must:

Participate in a group project to develop an NLP or a Computer Vision project. Your project can either be the NLP or the Computer Vision application.

In case of the NLP, you need to use the following link to the dataset The dataset is a collection of Amazon product reviews across several categories. Your task will be to train a neural network to perform sentiment analysis to allow it to match with one of the categories in the dataset. Note that the dataset already contains labelled data for positive and negative reviews. You should load all the negative and positive comments, mix and randomise the data, take some percentage of data to train your model, and use the rest for testing your model. Your solution needs to be able to:
Clean the data (punctuation, spelling etc.)

Encode the words in the review
Encode the labels for ‘positive' and ‘negative'
Conduct outlier removal to eliminate really short or wrong reviews.
Pad/truncate remaining data
Split the data into training, validation and test sets
Obtain batches of training data (you may use DataLoaders or generator functions)
Define the network architecture
Define the model class
Instantiate the network
Train your model
Test

Develop a simple web page/create an executable of your solution that will take an input sentence and provide an output of whether the review sentiment was positive or negative.

Run an inference on some test input data - both positive and negative and observe how often the model gets these right.
Repeat training and rearchitect the model if required.

Keep in mind your ethical responsibility as a data science practitioner of the need to be fair and uniform in deriving accurate sentiment from a product review when conducting the above i.e. the dataset may have been split into positive and negative by the owner, however, can you identify any issues in their decision that you've now addressed? Note these in your report.

Deploy the system on a simple website or provide an executable which can be run on the command line.

The interface for the NLP solution should have an input field to insert an input sentence into as well as a button to execute the sentiment analysis function you've implemented. Note that the facilitator will test out a few input statements to verify the accuracy of your model's sentiment analysis capability. The execution of the sentiment analysis should produce an output once an input sentence is entered into the field and the button clicked in the form of "Positive review" or "Negative review" as a text output. If you're confident that you've trained your model sufficiently well on the training data, experiment to see what results you get when you provide it a sample input that is outside the training data.

IF you choose to work on the Computer Vision project, you will work on Udacity's self-driving car simulator project. The download link to the simulator is provided below as is the training data. Your task is to build a machine learning model that is trained on the data provided and when run on the simulator, will hopefully keep the car on the road without running off track.