Clean up Dependency Checking and Ensure Full Function Modularity

[xehu]

In the feature_dict, there are many features sharing the same function. For example, "Message Length" and "Message Quantity" are both generated by text_based_features(). This makes dependency checking quite difficult. Furthermore, some of our functions rely on other functions being run first --- e.g., "information exchange" relies on "first person wordcount," which is a lexical feature that is run temporally prior.

We should split the functionality and make sure each feature has its own function, and ensure that functions are fully modular.

[xehu]

When this is complete, we should be able to remove the TODO: check dependencies comment in feature_builder:

        invalid_features = set()
        for feat in custom_features:
            if feat in self.feature_dict:
                self.custom_features.append(feat)
            if feat in self.feature_dict: # TODO: check dependencies

[sundy1994]

Non-modular Functions [with Emily's initial comments]

Below are functions that correspond to more than one features. We need to decide whether to combine them into a single feature or spliting the fuction on a case by case basis. @xehu

text_based_features() --- combine: "Message Length/Quantity"

• Message Length: ["num_words", "num_chars"]
• Message Quantity: ["num_messages"]

-> "Message Length/Quantity": ["num_words", "num_chars", "num_messages"]

other_lexical_features() --- split: These are all different features

• Questions: ["num_question_naive"] -> get_question_naive()
• Conversational Repair: ["NTRI"] -> get_NIRI()
• Word Type-Token Ratio: ["word_TTR"] -> get_word_TTR()
• Proportion of First-Person Pronouns: ["first_pronouns_proportion"] -> get_proportion_first_pronouns()

calculate_word_mimicry() --- combine: These are actually all about mimicry (in addition to the below category. We could either combine all of them ("Mimicry") or split them into "Lexical Mimicry" and "Vector Mimicry", which makes sense as well.

• Function Word Accommodation: ["function_word_accommodation"]
• Content Word Accommodation": ["content_word_accommodation", "content_word_accommodation_per_conv"]

-> Lexical Mimicry:["function_word_accommodation", "content_word_accommodation", "content_word_accommodation_per_conv"]

calculate_vector_word_mimicry() --- see above.

• (BERT) Mimicry: ["mimicry_bert"]
• Moving Mimicry: ["moving_mimicry"]

-> Vector Mimicry: ["mimicry_bert", "moving_mimicry"]

calculate_textblob_sentiment() --- split. Polarity and Subjectivity come from the same package but they don't have the same meaning/interpretation

• TextBlob Subjectivity": ["textblob_subjectivity"] -> get_textblob_subjectivity()
• TextBlob Polarity": ["textblob_polarity"] -> get_textblob_polarity()

[xehu]

@sundy1994 added my initial comments!

[sundy1994]

Added the new feature/function names

[xehu]

Moving this to the BRM R&R Milestone, because it's worth cleaning this up and updating the documentation, but only after we merge in the other changes!