根据列类别从数据框中的行创建文本列表

Question

I have a dataframe of categories and text strings:

category    strings

pets        leash cat dog
pets        cat dog frog
candy       chocolate frog
candy       jelly beans lollipops

I would like 2 lists:

petlist = ['leash', 'cat', 'dog', 'cat', 'dog', 'frog']
candylist = ['chocolate', 'frog', 'jelly', 'beans', 'lollipops']

The following code makes one list of all of the words from the strings column:

all_words = df['strings'].str.cat(sep=' ').split()

How can I split this up into 2 lists based on the category and put the 2 lists in a dictionary?

Here is what I tried:

all_words = {}
for cata in df['category']:
    all_words['wordlist_%s'% cata]=[]
for cata in df['category']:
    df_cata = df.loc[df['category'] == cata]
    all_words['wordlist_%s'% cata].append(df_cata['strings'].str.cat(sep=' ').split())

It has the correct keys but each key gives me the words from the first row of that category over and over. So I've got a dictionary with one list that says leash cat dog leash cat dog and another list that says chocolate frog chocolate frog (so it's clearly starting over in a way that I don't want it to).

Answer 1

You can set the index first then split then group on the index and concatenate all the lists with sum and make a dict out of it.

df.set_index('category').strings.str.split().groupby(level='category').sum().to_dict()

Output

{'candy': ['chocolate', 'frog', 'jelly', 'beans', 'lollipops'],
 'pets': ['leash', 'cat', 'dog', 'cat', 'dog', 'frog']}

Answer 2

this should do it

df.groupby('category').strings.apply(' '.join).str.split()

category
candy    [chocolate, frog, jelly, beans, lollipops]
pets              [leash, cat, dog, cat, dog, frog]
Name: strings, dtype: object

---

extra credit
get unique list

df.groupby('category').strings.apply(' '.join).str.split().apply(np.unique)

category
candy    [beans, chocolate, frog, jelly, lollipops]
pets                        [cat, dog, frog, leash]
Name: strings, dtype: object

---

over achiever
value_counts because I think it's interesting

df.groupby('category').strings.apply(' '.join).str.split(expand=True) \
    .stack().groupby(level=0).apply(pd.value_counts)

 category           
candy     jelly        1
          frog         1
          lollipops    1
          beans        1
          chocolate    1
pets      cat          2
          dog          2
          leash        1
          frog         1
dtype: int64