Assignment 1: BMI 550— Applied BioNLP Author: Chase Fensore

• fuzzywuzzy==0.18.0
• nltk==3.8.1
• pandas==2.0.0
• python_Levenshtein==0.22.0
• scikit_learn==1.3.1
• thefuzz==0.20.0

1. Preprocessing: as described in the report, 3 combinations are used.

• P0: lowercasing and word tokenization. TO USE: pass "none" to preprocessing argument of predict_test_CUIs (line 423).
• P1: P0 + punctuation removal. TO USE: pass "pp1" to preprocessing argument of predict_test_CUIs (line 423).
• P2: P1 + Porter stemming. TO USE: pass "pp2" to preprocessing argument of predict_test_CUIs (line 423).

3. Similarity metrics:

• Levenshtein distance: to use, pass "Levenshtein" to sim_metric argument of predict_test_CUIs (line 423).
• Token sort ratio: to use, pass "token_sort_ratio" to sim_metric argument of predict_test_CUIs (line 423).

3. Threshold (for similarity metric):

• Token sort ratio: to adjust between 0-100, change min_pred_thresh argument of predict_test_CUIs (line 423).
• Levenshtein distance: to adjust between 0-1.0, change min_pred_thresh argument of predict_test_CUIs (line 423).

1. conda install requirements.txt
2. To perform automatic symptom detection on Assignment1GoldStandardSet.xlsx:

• Run: python rulebased_system_Fensore.py. (Notes: line 409: must set file to read Assignment1GoldStandardSet.xlsx, line 423: set desired hyperparameters described above, line 424: change output file name, if desired.)
• New output will be stored in: data/result-Assignment1GoldStandardSet.xlsx.
• My best-performing output is currently stored in: results/result-Assignment1GoldStandardSet.xlsx.

4. To perform automatic symptom detection on UnlabeledSet.xlsx:

• Run: python rulebased_system_Fensore.py. (Notes: line 409: must set file to read UnlabeledSet.xlsx, line 423: set desired hyperparameters described above, line 424: change output file name, if desired.)
• New output will be stored in data/result-UnlabeledSet.xlsx
• Existing output is stored in: results/result-UnlabeledSet.xlsx.