The goal of this assignment is to focus on Word Embedding and Named Entity Recognition $($NER$).$ You will work with a dataset containing sentences tagged with POS and named entity labels. The assignment is divided into three parts: understanding and pre$-$processing the dataset, implementing a Word$2$Vec model for word embedding, and designing and evaluating a sequential model for NER.

Part I $-$ Understanding the Dataset and Preprocessing $3$ marks

You will become familiar with the dataset and perform the necessary pre$-$processing steps. This includes loading the dataset, explore its structure, and prepare for subsequent modelling tasks.

Loading the Dataset $(1$ mark$)$

Load the provided dataset and display the first few rows.

Exploratory Data Analysis $(1$ mark$)$

Count the number of sentences.

Identify the unique POS tags and named entity tags.

Visualize the distribution of sentence lengths.

Visualize the frequency of different POS tags.

Data Preprocessing $(1$ mark$)$

Tokenize the sentences and map each word to its corresponding POS and named entity tags.

Convert the words and tags into numerical representations suitable for modelling, such as word indices and tag indices.

Part II $-$ Word Embedding using Word$2$Vec $3$ marks

You will implement a Word$2$Vec model to create word embeddings from the dataset and evaluate its effectiveness. This involves training the Word$2$Vec model on the dataset and visualizing the resulting word embeddings.

Word$2$Vec Model Implementation $(1$ mark$)$

Implement a Word$2$Vec model to generate word embeddings from the dataset.

Ensure your model captures the contextual relationships between words effectively.

Word Embedding Visualization $(1$ mark$)$

Visualize the word embeddings using techniques such as t$-$SNE or PCA.

Display how different words and their embeddings relate to each other in a lower$-$dimensional space.

Evaluation $(1$ mark$)$

Evaluate the quality of the word embeddings by examining their effectiveness in capturing semantic relationships.

Present your evaluation results clearly and concisely.

Part III $-$ NER using Sequential Model $4$ marks

You will design and implement a sequential model $($such as RNN$,$ LSTM$,$ or GRU$)$ to perform NER. You will train your model, evaluate its performance, and discuss the results.

Model Design and Training $(2$ marks$)$

Choose a sequential model and justify your choice.

Design the architecture of your model, specifying the input layer, hidden layers, and output layer.

Train the model on the training dataset, ensuring proper handling of sequences and padding.

Model Evaluation $(1$ mark$)$

Evaluate the performance of your trained model on the test dataset using metrics such as precision, recall, and F$1-$score for each named entity tag.

Present the evaluation results clearly and concisely.

Analysis and Discussion $(1$ mark$)$

Discuss the strengths and weaknesses of your sequential model based on the evaluation results. Answer without justification will not be awarded marks.

Suggest potential improvements or alternative approaches to enhance performance. Answer without justification will not be awarded marks.