My current code:
from nltk.tag import pos_tag, map_tag
search_term = 'quaker lemon banana oatmeal'
lst = []
search_term_words = search_term.split()
for w in search_term_words:
if w in flavor_grocer_mapping:
for flavor in flavor_grocer_mapping[w]:
if flavor in search_term:
lst.append((flavor, 'FLAVOR'))
for x in search_term.replace(flavor, '').split():
if x in brand_grocer_mapping:
for brand in brand_grocer_mapping[x]:
if brand in search_term.replace(flavor, '').split():
lst.append((brand, 'BRAND'))
for word, tag in pos_tag(word_tokenize(search_term.replace(flavor, '').replace(brand, '').strip())):
lst.append((word, map_tag('en-ptb', 'universal', tag)))
The result I got:
[('lemon', 'FLAVOR'),
('quaker', 'BRAND'),
('banana', 'NOUN'),
('oatmeal', 'NOUN'),
('banana', 'FLAVOR'),
('quaker', 'BRAND'),
('lemon', 'ADJ'),
('oatmeal', 'NOUN')]
My expected result is:
[('lemon', 'FLAVOR'),
('banana', 'FLAVOR'),
('quaker', 'BRAND'),
('oatmeal', 'NOUN')]
I know the problem is that the splitted words in search terms run recursively in the for loops. How can lookup/map a whole string that contains multiple keys in a dictionary? (For example, lemon and banana are in the search term string and they are the keys in the flavor_grocer_mapping dictionary.)
Well, I tried to recreate flavor_grocer_mapping based on your results, your additional code to lookup is highly unnecessary. Instead, since you have a dictionary that corresponds to search_term_words as keys, do a proper dictionary lookup in a list comprehension.
search_term = 'quaker lemon banana oatmeal'
search_term_words = search_term.split()
#recreating your dictionary
flavor_grocer_mapping = [('lemon', 'FLAVOR'),
('banana', 'FLAVOR'),
('quaker', 'BRAND'),
('oatmeal', 'NOUN')]
flavor_grocer_mapping = {k:v for (k,v) in flavor_grocer_mapping}
#solution
results = [(word,flavor_grocer_mapping[word]) for word in search_term_words]
results
>>[('quaker', 'BRAND'),
('lemon', 'FLAVOR'),
('banana', 'FLAVOR'),
('oatmeal', 'NOUN')]
In the case of your provided example, I suggest your merge all the dictionaries to 1 instead of repeatedly looping over them.
この記事はインターネットから収集されたものであり、転載の際にはソースを示してください。
侵害の場合は、連絡してください[email protected]
コメントを追加