Firstly, I use FasText for language detection. Then, from the language detection, I want to translate from certain language (in this case, Malay) to English. For the translation part, I use Google Translate API using Python. The problem is, the output return NaN value for other language (in this case, English and Thai). I want to return only the translated text only, which is Malay.

from googletrans import Translator
import pandas as pd
import numpy as np
translator = Translator()


df = pd.DataFrame({
'text': ["how are you", "suka makan nasi ayam", "สวัสด","hai, apa khabar"], 
'lang': ["english", "malay", "thai","malay"]
})
df

Dataframe df:

df1=df[df["lang"] == "malay"]
df['text'] = df1['text'].apply(translator.translate, dest='en').apply(getattr, args=('text',))
df

Generated output:

Desired output:

text                       |    lang
-----------------------------------
how are you                |   english
like to eat chicken rice   |   malay
สวัสด                      |   thai
Hello how are you          |   malay

>Solution :

You need to use a boolean mask:

translate_en = lambda x: translator.translate(x, dest='en').text

m = df['lang'] == 'malay'
df.loc[m, 'text'] = df.loc[m, 'text'].apply(translate_en)
print(df)

# Output
                       text     lang
0               how are you  english
1  like to eat chicken rice    malay
2                     สวัสด     thai
3         Hello how are you    malay

Same with update:

df.update(df.loc[m, 'text'].apply(translate_en))
print(df)

# Output
                       text     lang
0               how are you  english
1  like to eat chicken rice    malay
2                     สวัสด     thai
3         Hello how are you    malay