|
Module code and Title
|
DTS406TC Natural Language Processing
|
|
School Title
|
School of Artificial Intelligence and Advanced Computing
|
|
Assignment Title
|
Coursework 1 (CW 1)
|
|
Submission Deadline
|
5 pm China time (UTC+8 Beijing) on March23,2025
|
|
Final Word Count
|
3000
|
|
If you agree to let the university use your work anonymously for teaching and learning purposes, please type “yes” here.
|
|
Policy
The report should be submitted in PDF format. The word count limit of the report is 3,000. All code must be written in Python. It should be well-structured, easy to read, and thoroughly documented. Each function should include comments explaining its purpose and functionality. Ensure that the code runs without errors. All required files should be included in a single zip file. A README file is needed to explain how to run the code and list any dependencies. The document must be submitted through Learning Mall Online to the appropriate drop box.
DTS406TC Natural Language Processing
Coursework 1 (Group Assessment)
Due: 5:00 pm China time (UTC+8 Beijing) on March23,2025
Weight: 40%
Maximum score: 100 marks (80 % group marks + 20 % individual marks)
Groupings: Each group consists of 2-3 students. You are free to select your own team members. Students who do not make a selection will be randomly assigned to a team . Once the teams are confirmed , no changes will be permitted.
Assessed learning outcomes:
A Systematically comprehend the theoretical foundations ofNatural Language Processing B Apply statistical and machine learning techniques to process and analyze natural language data
Overview
Sentiment Analysis is the process of determining whether textual content expresses a positive, neutral, or negative sentiment. With the vast amount of textual data (e.g., tweets, Reddit posts, reviews) generated daily, Sentiment Analysis can automatically identify users' attitudes from User-Generated Content (UGC), assisting companies or organizations in making informed decisions.
1. Literature Review on Sentiment Analysis (20 Marks, Individual Work)
a) Overview of the sentiment analysis and its applications. Please provide three examples of real-life applications of sentiment analysis. (6 Marks)
b) Please list three key challenges in sentiment analysis. (6 Marks)
c) Please elaborate on two traditional methods (e.g., Naive Bayes, SVM) and two deep learning approaches (e.g., BERT, GPT) on the sentiment analysis. Meanwhile, discuss the advantages and disadvantages of each approach. (8 Marks)
d) Each team member should individually complete the literature review on sentiment analysis. This section will be scored individually.
2. Data Collection (12 Marks, Group Work )
Collect two datasets of User-Generated Content (UGC) from platforms like Twitter, Reddit, or Weibo, focusing on sentiment analysis in different scenarios. Each dataset should contain a minimum of 3,000 instances. Preprocess the datasets by performing tasks like stopword removal and tokenization. Finally, a statistical analysis of the two datasets should be conducted (e.g., the word distribution of the corpus). Notice that some UGC data may be downloaded from Kaggle if there are API restrictions preventing direct downloads from social platforms. (6 Marks/dataset x 2=12 Marks)
3. Algorithm Description & Implementation (48 Marks, Group Work)
a) Choose four approaches for the sentiment analysis task on the collected UGC datasets: two using traditional methods and two employing deep learning methods. All four approaches should be applied to each of the UGC datasets. Please provide the pseudo- code and briefly provide the comments for each function of the pseudo-code. (5 Marks/algorithm x 4= 20 Marks)
b) Develop a sentiment analysis system for each approach using Python. The implementation pipeline should include the following components: feature engineering (3 Marks, e.g., converting textual data to embedding space regarding to different approaches), algorithm implementation (3 Marks, with fine-tuning required for the deep learning approach), and metrics computation (1 mark). 7 Marks/algorithm x 4 = 28 Marks)
4. Results Analysis (13 Marks, Group Work)
a) Provide the sentiment analysis results for each approach applied to the two UGC datasets. Select and apply three relevant metrics (e.g., precision, recall, and F 1 score) to assess the performance of the implemented models, with each metric worth 3 Marks. (9 Marks)
b) Explain the reasons behind the model performance for each approach. (1 Mark/algorithm x 4 = 4 Marks)
5. Report Writing (7 Marks, Group Work)
This coursework evaluates your understanding the challenges of the problem and the correctness of the proposed algorithms. It also tests your professional skills in terminology usage, presentation of algorithms and experimental results, as well as the logical manner of the proposal. (7 Marks)
Submission
One of the team members must submit a single zip file. The zip file is named "TeamID_Coursework.zip". It includes: a cover letter with the group member information and the final PDF reports. The final PDF reports includes the individual reports of literature review on the sentiment analysis and the group report. A folder labeled "algorithms" contains all the model implementations, data preprocessing scripts, and evaluation scripts. A folder labeled "data" contains all the datasets and the experimental results in the CSV format.