如何将列从数据帧传递到NLTK Python中的wordnet.synsets（）

Question

I have a dataframe in which one the columns contains english words. I want to pass each of the elements in that columns through NLTKs synsets() function. My issue is that synsets() only takes in the a single word at a time.

eg wordnet.synsets('father')

Now if I have dataframe like:

dc = {'A':[0,9,4,5],'B':['father','mother','kid','sister']}
df = pd.DataFrame(dc)
df
   A       B
0  0  father
1  9  mother
2  4     kid
3  5  sister

I want to pass column B though synsets() function and have another column that contains its output. I want to do this without iterating through the dataframe.

How do I do that?

Answer 1

You could use the apply method:

In [4]: df['C'] = df['B'].apply(wordnet.synsets)

In [5]: df
Out[5]: 
   A       B                                                  C
0  0  father  [Synset('father.n.01'), Synset('forefather.n.0...
1  9  mother  [Synset('mother.n.01'), Synset('mother.n.02'),...
2  4     kid  [Synset('child.n.01'), Synset('kid.n.02'), Syn...
3  5  sister  [Synset('sister.n.01'), Synset('sister.n.02'),...

However, having a column of lists is usually not a very useful data structure. It might be better to put each synonym in its own column. You can do that by making the callback function return a pd.Series :

In [29]: df.join(df['B'].apply(lambda word: pd.Series([w.name for w in wordnet.synsets(word)])))
Out[29]: 
   A       B            0                1            2                   3  \
0  0  father  father.n.01  forefather.n.01  father.n.03  church_father.n.01   
1  9  mother  mother.n.01      mother.n.02  mother.n.03         mother.n.04   
2  4     kid   child.n.01         kid.n.02     kyd.n.01          child.n.02   
3  5  sister  sister.n.01      sister.n.02  sister.n.03           baby.n.05   

             4                     5             6         7           8  
0  father.n.05           father.n.06  founder.n.02  don.n.03  beget.v.01  
1  mother.n.05           mother.v.01    beget.v.01       NaN         NaN  
2     kid.n.05  pull_the_leg_of.v.01      kid.v.02       NaN         NaN  
3          NaN                   NaN           NaN       NaN         NaN  

(I've chosen to display just the name attribute of each Synset ; you could of course use

df.join(df['B'].apply(lambda word: pd.Series(wordnet.synsets(word))))

if you want the Synset objects themselves.)