The study of plant-disease relationships is challenging because accurate identification can lead to new medicines, enhancing treatment precision. However, the vast number of publications in this field creates barriers for readers trying to access and interpret this information. Long Short-Term Memory (LSTM) networks, a type of recurrent neural network, are effective at capturing long-term dependencies in text sequences, which is beneficial for analyzing complex plant-disease relationships. LSTM models can retain crucial contextual information during classification. Additionally, Term Frequency-Inverse Document Frequency (TF-IDF) helps highlight key terms by assigning weights based on term frequency, aiding in keyword identification. We hypothesized that combining TF-IDF weighting with an LSTM network would enhance the detection of plant-disease relationships. In our experiment, we used a standard corpus consisting of 1309 plant-disease records from 199 Medline abstracts. We employed an 80%-20% training-testing split and compared the LSTM-TF-IDF model against Logistic Regression and Multi-Layer Perceptron classifiers. The LSTM-TF-IDF model achieved the highest performance, with a macro F1-score of 0.724 and a micro F1-score of 0.931 on the test set. In contrast, a Support Vector Machines classifier, used in previous studies with the same corpus and test scheme, achieved a macro F1-score of 0.567 and a micro F1-score of 0.764. This demonstrates that the LSTM-TF-IDF model significantly outperforms existing methods in classifying plant-disease relationships.