seqtagger/extract_score.py

import re


def extract_score(results):
    res_dico = {}
    split_res = results.detailed_results.split("\n")
    for index, chunk in enumerate(split_res):
        if "F1-score (micro)" in chunk:
            res_dico["F1-score_micro"] = float(re.findall(r"\d+.\d+",chunk)[0])
        if "F1-score (macro)" in chunk:
            res_dico["F1-score_macro"] = float(re.findall(r"\d+.\d+",chunk)[0])
        if "By class" in chunk:
            res_dico["by_class"] = {}
            for res_class in split_res[index+1:]:
                class_splt = res_class.split()
                res_dico["by_class"][class_splt[0]] = {class_splt[1]:float(class_splt[2]), class_splt[4]: float(class_splt[5]), class_splt[7]: float(class_splt[8]),class_splt[10]:float(class_splt[11]), class_splt[13]:float(class_splt[14]), class_splt[16]:float(class_splt[17])}
        log_line = results.log_line.split("\t")
    res_dico["precision"] = float(log_line[0])
    res_dico["recall"] = float(log_line[1])
    return res_dico