How to measure flexibility?

Hello :slight_smile: I am writing a master thesis about service providers’ flexibility and having access to 10.000 chat conversations between a company and its customers. I am planning on using text mining as a method and have a hard time finding a way to measure flexibility. Just wanted to try this forum to see if there is anyone that has worked with this type of measure before and has any tips? (e.g. flexibility dictionaries, articles, tips on ways to measure flexibility).

Hi @Nadine_H -

Can you be more specific about the specific flexibility metric you’re interested in? Maybe reference to a paper or python library that demonstrates what you mean? I thought this topic had come up on this forum in the past, but I’ve searched a little bit and can’t find it…

Hi @ScottF ! I appreciate your interest, thank you for searching around for me. :grinning:
I have a hard time being specific since this seems like a topic that has not been researched much and I have not found any python library. I found this one article that made a communication flexibility scale and compared this to communication adaptability ( http://dx.doi.org/10.1080/10417949409372934). After reading a few different articles I am trying to look into the constructs: the service provider’s argumentativeness, confirmation, relaxation, use of negative words, owning statements, and/or articulation, which are found to be related to perceived flexibility.
If you have any articles related to flexibility in language or any of the constructs mentioned above I would be happy to get some suggestions. :blush:

That sounds really interesting, but unfortunately I don’t have any experience with it myself. Let me tag @julian.bunzel, @victor_palacios, and @dursundelen to see if any of our resident experts have suggestions for you. :slight_smile:

I’ve never seen this topic before, but if this is a new space you can start with what you found: argumentativeness, confirmation, relaxation, use of negative words, owning statements, and articulation.

I would say start with negation words since that is a very simple, common task with clear metrics. Once you have built a pipeline/workflow to examine negative words then you can simply add more complexity with confirmation, relaxation, articulation, etc.

For that, you can go the lexicon-based approach where you have a set of words you know are negation or you can build a model by going through your data and labeling a small sample (say 100 conversations). If you’re feeling brave you can try a machine learning approach or a deep learning approach as well.

Once you’ve gone through 100 samples of your data, you might even discover your own unique way to measure flexibility.

If you need more help with labeling, you might enjoy our active learning tutorial.

1 Like

Thank you @victor_palacios & @ScottF , I am grateful for all inputs! I will look into it! Thank you both :blush:

Just a couple of ideas to push forward the discussion and search for appropriate method:

  1. Analyze replies to see how many strict patterns are used by employees. Exclude NE and evaluate variation of words and combinations of words.
  2. Answer-Reply pair. Check if the employee uses same words and word’s patterns as the user. To trace situations like “I would like to ask if you provide translation services?” - “Thank you very much for your questions. Our company is a globally recognized leader in b2b services and we provide our customers with the highest quality result… etc. including translations”.
  3. Replies - extract the last sentence to see how the act of communication ends. Is it formal? Contain NEs (like “Thank you for asking, John. Get in touch if you need any other assistance” vs “Best regards”)

Customer Services commonly use scripts, so the best result would be to have a script based on actual data :slight_smile: This would be very illustrative and may show the company how flexible/inflexible is it’s CS.