combine nested dictionaries in list of nested dictionaries based on matching key:value pair

Question

I tried to google and find a problem which is very similar to my use case here: combine dictionaries in list of dictionaries based on matching key:value pair. But it seems that it did not correspond 100% to my case as I have list of nested dictionaries. Suppose that I have a list of nested dictionaries (more than 2) but in this case I considered two nested dictionaries to make the example:

my_list = [{'sentence': ['x',
   'ray',
   'diffractometry',
   'has',
   'been',
   'largely',
   'used',
   'thanks',
   'to',
   ],
  'mentions': [{'mention': [27, 28],
    'positives': [26278, 27735, 21063],
    'negatives': [],
    'entity': 27735}]},
 {'sentence': ['x',
   'ray',
   'diffractometry',
   'has',
   'been',
   'largely',
   'used',
   'thanks',
   'to',
   ],
  'mentions': [{'mention': [13, 14],
    'positives': [7654],
    'negatives': [],
    'entity': 7654}]}]

How can I merge these two dictionaries based on the matching of key(sentence) and value(list of all tokens) So that I can get the desired result as below:

my_new_list = [
{'sentence': ['x',
   'ray',
   'diffractometry',
   'has',
   'been',
   'largely',
   'used',
   'thanks',
   'to',
   ],
  'mentions': [
    {'mention': [27, 28],
    'positives': [26278, 27735, 21063],
    'negatives': [],
    'entity': 27735
    },
   {'mention': [13, 14],
    'positives': [7654],
    'negatives': [],
    'entity': 7654
     }
   ]
}
]

How to merge the list of key "mentions" when matching the key(sentence):value(list of all tokens)? In my actual list, there will be a lot of dictionaries with the same style.

Many thanks for your help.

You should make your example smaller. That will help you and us :) — log0
– log0, Commented Aug 18, 2021 at 9:41
@log0-- Thank for the advice. I reduce some tokens from the key "sentence". — Erwin
– Erwin, Commented Aug 18, 2021 at 9:44

Amith Lakkakula · Accepted Answer · 2021-08-18 10:01:48Z

1

my_dict = {}
for row in my_list:
    key = ' '.join(row['sentence']) # use sentence as key
    if key in my_dict:
        my_dict[key]['mentions'].extend(row['mentions'])
    else:
        my_dict[key] = row
        
my_list = list(my_dict.values())

edited Aug 18, 2021 at 10:01

answered Aug 18, 2021 at 9:55

Amith Lakkakula

5163 silver badges8 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

ira Over a year ago

What if the keys are dynamic? lets say key 'mentions' can be a different name based on the list of dicts passed by the user

log0 · Accepted Answer · 2021-08-18 10:19:56Z

0

From what I understand you want to group information by "sentence".

You can do this by iterating on your array and fill a dictionary of list indexed by sentence.

Something like:

from collections import defaultdict
sentences = defaultdict(list)
for element in my_list:
   key = tuple(element["sentence"])
   sentences[key].append(element)

this gives you

 { sentence1: [element1, element2], sentence2: [element3] }

From there should be able to easily construct the structure you want.

edit removed reference to specific fields

edited Aug 18, 2021 at 10:19

answered Aug 18, 2021 at 9:50

log0

11k4 gold badges32 silver badges68 bronze badges

6 Comments

Erwin Over a year ago

Can you purpose a solution with the codes? As I tried with the linked I search but i could not solve the problem.

log0 Over a year ago

Converted the sentence to a tuple to make it hashable (usable as a key in a dictionary)

Erwin Over a year ago

The solution did not work when you have more than two lists in it.

log0 Over a year ago

@Erwinwin What do you mean ? can you write down the structure

Erwin Over a year ago

I mean this solution did not work when you have more than two nested dictionaries in that list.

|

Collectives™ on Stack Overflow

combine nested dictionaries in list of nested dictionaries based on matching key:value pair

2 Answers 2

1 Comment

6 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

6 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related