loading wai labels

load_cbt_annotations.py

import pandas as pd

def load_cbt_annotations(path):
    import json
    with open(path) as f:
        x = json.load(f)
    # make session numeric
    df = pd.DataFrame()
    for k, v in x.items():
        while v:
            session, vv = v.popitem()
            session = int(session[0])
            df_v = pd.DataFrame()
            for annotation in vv:
                a_id = annotation.pop('annotator_id')
                df_v = df_v.append(pd.DataFrame(annotation, index=pd.MultiIndex.from_tuples([(session, a_id, k)],
                names=['session', 'annotator', 'speaker'])))
            df = df.apply(pd.to_numeric)
            df = df.append(df_v)
    return df
  

goal = ['wai_4', 'wai_6', 'wai_10', 'wai_11']
task = ['wai_1', 'wai_2', 'wai_8', 'wai_12']
bond = ['wai_3', 'wai_5', 'wai_7', 'wai_9']
categories = ['goal', 'task', 'bond']
wais = [f'wai_{i}' for i in range(1, 13)

filter_annotators = True
df_label = load_cbt_annotations('transkriptions_coadapt/CBT_WAI.json')
df_label = df_label[wais].astype(float).dropna()
df_label = df_label.reset_index()
if filter_annotators:
    df_label = df_label[df_label.annotator.isin([43, 52])]
df_agg_labels = df_label.groupby(['speaker', 'session']).mean()
df_agg_labels = df_agg_labels.reset_index()[['speaker', 'session', *wais]]
df_agg_labels['id'] = df_agg_labels[['speaker', 'session']].apply(lambda x: '_'.join(x.astype(str)), axis=1)
df_agg_labels['goal'] = df_agg_labels[goal].apply(lambda x: round(sum(x) / len(goal)), axis=1)
df_agg_labels['bond'] = df_agg_labels[bond].apply(lambda x: round(sum(x) / len(bond)), axis=1)
df_agg_labels['task'] = df_agg_labels[task].apply(lambda x: round(sum(x) / len(task)), axis=1)