What classification to choose?

Question

I have a huge amount of screeching data, and I have to categorize reviews into 8 different categories.
Categories

Cleanliness Customer Service Parking Billing Food Pricing Food Quality Waiting time Unspecified 

Reviews contain several categories, so I used a multi-stage classification. But I am embarrassed how I can deal with the positive / negative. An example review may be positive for food quality, but negative for customer service. Ex- food taste was very good but staff behaviour was very bad. so review contains positive food quality but negative Customer service food taste was very good but staff behaviour was very bad. so review contains positive food quality but negative Customer service How can I handle this case? Do I have to analyze moods before classification? Please help me

Answer 1

I think your data is very similar to restaurant reviews. It contains about 100 reviews with a different number of aspects in each ( additional information ). Thus, you can use Aspect-based mood analysis as follows:

1-aspect selection

Extract aspect aspects from reviews.

2 aspect polarity detection

For a given set of aspect aspects within a sentence, determine whether the polarity of each aspect of the aspect is positive, negative.

3- Define category aspect

Given the predefined set of aspect categories (for example, food quality, customer service), identify the aspect categories discussed in this proposal.

4 - Determine the polarity

Given a set of predefined aspect categories (for example, food quality, customer service), determine the polarity (positive, negative) for each category aspect.

For more details on such a project, see.

Hope this helps you.

Answer 2

Yes, you need a mood analysis. Why don’t you create tokens of your data, and find the words you need from the sentence, now the approach most suitable for you is to find related words along with your feelings. that is, the food was good, but the cleanliness was impractical.

In this case, you have [food, good, clean, and not, appropriate] now food links with the following term and cleanliness, "does not fit" the following conditions

again, you can classify either two classes, that is, 1.0 for good and bad .. or you can add classes based on your case. Then you will have the data as such:

 -------------------- FEATURE | VAL -------------------- Cleanliness 0 Customer -1 Service -1 Parking -1 Billing -1 Food Pricing -1 Food Quality 1 Waiting time -1 Unspecified -1 

I gave this as an example, where -1,1,0 are not considered, both good and bad, respectively. You can add more categories as 0,1,2 bad honest goods. Maybe I do not answer it very well, but this is what I relate to.

Note. You must understand that a model cannot be ideal, because what machine learning is, you must be wrong. Your model cannot give an ideal classification; it must be incorrect for certain resources, which it will study over time and improve.

Answer 3

There are many ways to classify multiple labels.

The simplest one will have a model for each class, and if the review reaches a certain threshold value for this label, you apply this label to the overview.

This will apply to classes on its own, but it seems like a good solution to your problem.