It depends completely on your problem and research question. I think
thinking about the appropriate granularity is important and one way is
to start broad and see what you get then narrow downs. ANother is to do
a pass first to explore the level where there is interesting meaning
with respect to the research question.
Ive done it on lesson plans and lessons. Then I started just with the
course steps of the lesson plan just looking at whether they were waves
overall. That was useful. In some things I realised that I was missing
some interesting detail of the way I intended there and was enacting
waves within waves, so I went more fine grained down to smaller steps. I
think that is a useful approach.
Ive used LCT autonomy (same issue applies) to analyse some of my own
writing (a chapter in a magic book) and to start I have just looked at
the section level. That gives me an overview (that is interesting) but
again it was clear I was missing some interesting detail when I did it
so next Im going to go down deeper and finer grain. I probably wont go
down to sentence level just some rough interesting version of paragraph
level of sentences but more likely starting at paragraphs or groups of
sentences that change in length.
You have a similar thing in text analysis (eg doing thematic analysis)
and natural language processing generally. DOing thematic analysis on
transcripts Ive done it at a paragraph level initially, but really what
you need to do is do a pass over it first to decide what is the
interesting unit of analysis. In some places where the paragraphs have
detail Ive switched to sentences for those parts. Go too fine at the
outset and you miss the interesting overall meaning. DOnt go fine enough
and likewise depending on the data, but you can be pragmatic and decide
as you analyse each chunk.
In NLP different research looks at different levels too and some goes
down even below word (when it is spoken) to utterances include breaks in
words, ums and ers and pauses but thats only when they are interested in
what they mean and contribute. It is the same here - I imagine you would
only go down to word level if your interest was really about precise
words chosen not courser ones. If the thing you are looking at changes
in gravity / semantics every sentence then sentence level may be
appropriate. If multiple sentences about one thing are all using similar
level terminology and context then no need even to go down to sentences.
That is why doing a pass to get a feel for the data and make the
decision may be the way forward.
Paul
On 06/04/2024 21:44, jory Good wrote:
You don't often get email from jory90@xxxxxxxxx. Learn why this is important <https://aka.ms/LearnAboutSenderIdentification>
Hello everyone
I'm a new member and forgive me for sending this question. I've a simple question about semantic gravity and density in writing. Should we focus during analysis on sentences OR words/technical vocabulary? This is not clear to me, and I wish I could find simple examples to start with.
Many thanks in advance,
Jory
