Spacy detect correctly GPE

I've a set of string where I shall detetect the country its belongs to, referring to detected GPE.

sentences = [
    "I watched TV in germany",
    "Mediaset ITA canale 5",
    "Je viens d'Italie",
    "Ich komme aus Deutschland",
    "He is from the UK",
    "Soy de Inglaterra",
    "Sono del Regno Unito"
]

So my expectation : "I watched TV in germany" => DE "Soy de Inglaterra" => UK etc..

I've tried with this code:

from spacy.pipeline import EntityRuler
nlp = spacy.load('xx_ent_wiki_sm')  # Multilingual model


def detect_country(text):
    doc = nlp(text)
    countries = []
    for ent in doc.ents:
        if ent.label_ == 'GPE':
            countries.append(ent.text)
    return countries


for sentence in sentences:
    countries = detect_country(sentence)
    print(f"Countries detected in '{sentence}': {countries}")

But results are completely empty:

Countries detected in 'I watched TV in germany': []
Countries detected in 'Mediaset ITA canale 5': []
Countries detected in 'Je viens d'Italie': []
Countries detected in 'Ich komme aus Deutschland': []
Countries detected in 'He is from the UK': []
Countries detected in 'Soy de Inglaterra': []
Countries detected in 'Sono del Regno Unito': []

I don't understand if I'm missing something into spacy pipeline and if I'm actually take good choice for multilang model. Thx in advance for any advise