naitian / Condolence-Empathy-Online-Communities

Repository for "Condolence and Empathy in Online Communities", EMNLP 2020
10 stars 1 forks source link

Effective condolences positive and negative samples #1

Closed wangjs9 closed 2 years ago

wangjs9 commented 2 years ago

I am interested in your work after reading you paper. But I have a question that how to obtain the dataset of effectiveness condolence classification (Sec 7 What makes a good condolence?).

Another question is that I find the definition of condolence can be summerized as an expression of sympathy. However, many cited papers in this paper argues the definition of condolence is sympathy with a person in grief. Therefore, I am not sure about the meaning of condolence this paper focuses.

naitian commented 2 years ago

Hi Jessie, thanks for your interest!

To obtain the dataset of effective condolences, we look at the children comment of condolences and filter for appreciative comments (those containing certain high-precision appreciative phrases e.g. "thank you" or "this made my day"). We mark these as "effective". This is a relatively low-fidelity analysis, and could be interesting to explore further.

In our work, we make a distinction between empathetic and sympathetic condolences, both of which exist. Sympathetic condolences are often trite, highly-repeated phrases like "sorry for your loss" which engage with what happened to the condolence seeker. Empathetic condolences demonstrate that the condolence-giver relates emotionally to the condolence-seeker. Again, both types exist, but we make this distinction because some prior works (referenced in the paper) show that different kinds of grief responses have varying levels of helpfulness.

wangjs9 commented 2 years ago

Thank you for your reply!!!

But I think that my questions are not clear enough.

I have the first question because the paper mentions that there are 23,301 and 149,992 positive and negative effective condolence samples. But I cannot find these two numbers nor their sum in the paper/data. Therefore, I wonder where the samples are from? Are they contained in the dataset of condolence_pos? And how can I use the id to get the content? BTW, I am really intertested in the problem of effectiveness of empathetic expression / supportive messages. Therefore, I am really excited that I find some other people have a similar idea.

And another question is that the direct meaning of condolence is to express empathy/sympathy to people who suffer the death of their friends, relatives, pet... But I find that various topics are contained in the datatset, therefore I wonder is the condolence in your paper about empathetic/sypathteic expression to negative emotions because of diverse causes (not only death)?

Thank you again!!!