Most Possible Likely Word Generation for Text based Applications using Generative Pretrained Transformer model Comparing to Long Short Term Memory Model
Pilligundla Niharika, S. John Justin Thangaraj · 2022
In contrast to long short term memory techniques, the proposed study intends to produce automatic next word generation for text-based applications while enhancing accuracy using state-of-the-art generative pretrained transformers and recurrent neural networks. Materials and Methods: On the data, which is a text file including a series of words, generative pretrained transformer models and long short term memory are used. Long short term memory model that contrasts cutting edge generative pretrained transformer models for recommendation accuracy of the next word. It has been suggested and created to use LSTM. The sample size was calculated to be 8046 for each group with a G power of 0.8. Results: When compared with a long short term memory model (70.84%) for the same dataset p=0.02(p<0.05), the accuracy in predicting the upcoming word for text editor based Applications utilizing generative pretrained transformers was greatest at 87.98% with the lowest mean error. In generating the next word for text-based Applications, the study shows that generative pretrained transformers are more accurate than long short term memory..Conclusion: The study demonstrates that when recommending the most possible word for text generation based applications, generative pretrained transformers are more accurate than long short term memory.