test_ner.py 文件源码

def train_clf(self, trainfiles):
        # tokens: list of words, labels: list of corresponding labels
        # go document by document because of local context
        final_labels = []
        featmat = []
        for trainfile in trainfiles:
            for tokens, labels in yield_tokens_labels(trainfile):
                final_labels.extend(labels)
                featmat.append(self.make_featmat_rep(tokens))
        featmat = np.vstack(featmat)
        print("training classifier")
        clf = logreg(class_weight='balanced', random_state=1)
        clf.fit(featmat, final_labels)
        self.clf = clf