Analyzing the Performance of Transformer Models - IT Computer Science Assignment Help

Assignment Task

Task 

Comparative Study On Accurately Identifying Paraphrases Using Transformer Based Models 

Background

• Traditional sequential models employed encoder-decoder architecture to create language and translation models.
• Recurrent models often failed in longer sequence modelling since model often forgets contents that are placed farther.
• Convolutional model provided means to parallelize training when compared to Recurrent networks which can process data only sequentially.
• However, both CNN and RNN networks fail to comprehend context for longer sentences and are often cumbersome.
• Transformers(Vaswani et al., 2017) works solely on attention thus avoiding recursions and allows for a greater level of parallelization.

• Figure 1 shows the transformer architecture with point-wise fully connected encoders as seen in the left half of the figure and decoders seen to the right half.
• Self-attention layers introduced in transforms help in parallelization of computation, reduces complexity in each layer as compared to recurrent and convolutional layers.
• It majorly helps with the learning of long-range dependencies and yields better interpretation of language models.
• The conception of transformers has paved way for a large number of related models that aim at improving the performance of transformers over a large variety of NLP tasks.
• This study aims at understanding and analysing the performance of transformer models solely for paraphrase identification.
• The accuracy in identifying paraphrases can yield a better answer to posed question or query on various platforms 

Aim and Objectives
The aim of this research is to provide a comparative study on the performance and accuracy of various state-of-the-art models available for paraphrase identification. Successfully identifying paraphrases can help a host of activities involved in question answering and search engine systems resulting in an improved quality of answers provided to the users.

The research objectives are formulated based on the aim of this study which are as follows:
• To analyse the various state-of-the-art models for paraphrase identification for the Quora Question pairs dataset.
• To suggest the best version of pre-training approaches available for paraphrase identification.
• To compare between the models to accurately identify paraphrases based on model pre-training and fine-tuning approaches.
• To study the impact of size of data on model performance and accuracy

Significance and Scope

Significance
• Paraphrase identification helps machines understand context at par with humans.
• Identification of such paraphrases paves way for better search results on a query, better or a quality answer in case of question answering system and a better way of intent classification for use in chatbots.
• Transformers has provided with plethora of options that are accurately and efficiently perform numerous NLP tasks.
• The expected outcome is a thorough comparative analysis of the said models on its ability to successfully identify paraphrases.

Scope
• The current scope is restricted to the accuracy of the model in identifying paraphrases.
• The current scope disregards the carbon footprint left behind these computationally intensive models.
• As multiple models and pre-training approaches are to be tried and tested for the objectives of this research the environmental impacts of these take a backseat.

Evaluation Metrics
• For the current implementation both F1 score and accuracy of models will be taken into account for assessing the model analysis.
• The best transformer-based model will be a careful consideration between F1 score and model accuracy. 

This IT Computer Science Assignment has been solved by our IT Computer Science Expert at My Uni Papers. Our Assignment Writing Experts are efficient to provide a fresh solution to this question. We are serving more than 10000+ Students in Australia, UK & US by helping them to score HD in their academics. Our Experts are well trained to follow all marking rubrics & referencing Style. Be it a used or new solution, the quality of the work submitted by our assignment experts remains unhampered. 

You may continue to expect the same or even better quality with the used and new assignment solution files respectively. There’s one thing to be noticed that you could choose one between the two and acquire an HD either way. You could choose a new assignment solution file to get yourself an exclusive, plagiarism (with free Turn tin file), expert quality assignment or order an old solution file that was considered worthy of the highest distinction.