Computer science | machine learning | artificial intelligence | bidirectional LSTM | natural language processing | language models | text prediction
The use of machine intelligence to facilitate daily tasks has increased substantially, such as suggestions of what may be written next in text messages, emails, and more. However, sometimes the scope of the context may be too broad and the prediction fails to generate an appropriate term. The main goal of this project is to focus on the professional biomedical context and use bidirectional long short-term memory (bi-LSTM) for text prediction. To achieve this, the model is trained with Wikipedia documents related to this field, which narrows down the ranges of vocabularies fed into it, thus increasing the accuracy and specificity of the output. When given a starter string, our language model, using machine learning algorithms, is successful in predicting the next few words in a sentence with correct usage of biomedical terminology. Extending our research further can allow for the suggestions of nomenclatures in other professional contexts as well.
To learn more about what I did in this project, visit my website at ireneyrzd.github.io/post/asdrp-predict-next-word/
Before using, remember to install the Wikipedia model in order to import and extract sources.
$pip install Wikipedia