如何获取每个评论中每个主题的情绪（pos/neg/neu）？ 从 LDA 获得主题后？

Question

Imagine I have a reviews from Amazon. I will pass those reviews into LDA model to get 4 topics.

Now each review might talk about more than 1 topic. It might contain 2/3/4 topic words. So Now is there any way where I could extract sentiments from those reviews for each topic

This is how I am expecting the function to be

def assign_topics_to_sentences(topics_dict,review):

   return sentences_topic_dict

#topics_dict is a dictionary that contains keys as topic number or name and values contain list like  {'topic1':[(word1,porb)(word2,prob)........],'topic2':[[(word1,porb)(word2,prob)........],.....}

#sentence_topic_dict {'topic1':(sentiment,score),'topic2':(sentiment,score)...}

Actually my main task is to get topic wise sentiment label(positive,negative,neutral for each topic for each review. If a user in a particular review doesn't talk about certain topic then its sentiment will be neutral

       Topic_0          Topic_1               Topic_2       Topic_3         Reviews
0   (positive, 0.4404)  (positive, 0.9612)  (neutral, 0.0)  (neutral, 0.0)  george super friendly easy going apartment gre...
1   (positive, 0.7269)  (positive, 0.9648)  (neutral, 0.0)  (neutral, 0.0)  wonderful love love love location place george...
2   (positive, 0.3182)  (positive, 0.9459)  (neutral, 0.0)  (positive, 0.4215)  brilliant time visit apartment huge great view...
3   (neutral, 0.0)      (positive, 0.9517)  (neutral, 0.0516)(neutral, 0.0)    martin great host although apartment building ...
4   (positive, 0.6705)  (positive, 0.7845)  (neutral, 0.1531)   (positive, 0.34)    would love leaving review apartment unfortunat...
5   (neutral, 0.0)  (positive, 0.946)   (positive, 0.2755)  (neutral, 0.0)  george one colleague meet location not difficu..

I want something like this for all reviews. If we could get a function for 1 review which I mentioned above we could replicate it to entire reviews.

Answer 1

Regardless of the topics, you can have sentiment of each of your sentences using one of the many tools for sentiment analysis, then proceed with assigning each sentence, a topic. Finally, put these two together.