AAI_2025_Capstone_Chronicles_Combined

‭ResolveAI‬

‭from small bugs, minor incidents, one-off requests and much more. The RAG component‬ ‭ensures that the LLM’s responses are grounded in relevant data, which helps to reduce‬ ‭hallucinations and improve overall answer quality.‬ ‭Focusing on the classification portion of our project an LSTM is a common approach‬ ‭where examples can easily be found in literature. For example a research group also utilized a‬ ‭bidirectional LSTM as well and compared it across various text classification benchmarks (Zhou‬ ‭et al., 2016). A different group looked at combining both an LSTM and Convolutional Neural‬ ‭Network (CNN) for the same task of text classification‬‭(Zhou et al., 2015)‭.‬ More recent projects‬ ‭can also be found using a similar approach to our study, for example text messages can be‬ ‭classified using a variety of similar approaches, including LSTM, bidirectional LSTM or Gated‬ ‭Recurrent Unit (GRU)‬‭(Nuzulul Khairu Nissa, 2022). In general there has been significant usage‬ ‭of LSTMs and similar neural network architectures in the field of text classification.‬

‭13‬

61