.. _politeness_strategies: Politeness Strategies ====================== High-Level Intuition ********************* A collection of conversational markers that indicates the use of politeness. Citation ********* `Danescu-Niculescu-Mizil et al. (2013) `_ `Chang et al. (2020) `_ Implementation Basics ********************** The PolitenessStrategies framework in Convokit identifies linguistic aspects of politeness using an annotated corpus of requests. It evaluates and operationalizes politeness theory components like indirection and deference, with a classifier achieving near-human performance across domains. Each utterance (message) and a Spacy object (to do the parsing) is parsed through the transform_utterance() method of a PolitenessStrategies instance. This method "Extract politeness strategies for raw string inputs (or individual utterances)." It calculates the following politeness strategies: please please_start hashedge indirect_btw hedges factuality deference gratitude apologizing 1st_person_pl 1st_person 1st_person_start 2nd_person 2nd_person_start indirect_greeting direct_question direct_start haspositive hasnegative subjunctive indicative Implementation Notes/Caveats ***************************** NA Interpreting the Feature ************************* List of politeness features returned by function (From cited papers): ====== ============================== ===================== ================== ===================================================== No. Strategy Politeness Score In top quartile Example (Positive = More Polite) ====== ============================== ===================== ================== ===================================================== 1. Gratitude 0.87*** 78%*** I really appreciate that you’ve done them. 2. Deference 0.78*** 70%*** Nice work so far on your rewrite. 3. Greeting 0.43*** 45%*** Hey, I just tried to . . . 4. Positive lexicon 0.12*** 32%*** Wow! / This is a great way to deal. . . 5. Negative lexicon -0.13*** 22%** If you’re going to accuse me . . . 6. Apologizing 0.36*** 53%*** Sorry to bother you . . . 7. Please 0.49*** 57%*** Could you please say more. . . 8. Please start −0.30* 22% Please do not remove warnings . . . 9. Indirect (btw) 0.63*** 58%** By the way, where did you find . . . 10. Direct question −0.27*** 15%*** What is your native language? 11. Direct start −0.43*** 9%*** So can you retrieve it or not? 12. Counterfactual modal 0.47*** 52%*** Could/Would you . . . 13. Indicative modal 0.09 27% Can/Will you . . . 14. 1st person start 0.12*** 29%** I have just put the article . . . 15. 1st person pl. 0.08* 27% Could we find a less complex name . . . 16. 1st person 0.08*** 28%*** It is my view that ... 17. 2nd person 0.05*** 30%*** But what’s the good source you have in mind? 18. 2nd person start −0.30*** 17%** You’ve reverted yourself . . . 19. Hedges 0.14*** 28% I suggest we start with . . . 20. Factuality −0.38*** 13%*** In fact you did link, . . . ====== ============================== ===================== ================== ===================================================== Related Features ***************** :ref:`politeness_receptiveness_markers` contains a similar list of markers related to politeness and receptiveness, computed by the SECR module (Yeomans et al., 2020); this can be though of as a more recent and upgraded version of the original politeness features.