On 2018/08/06 12:00 PM, R Smith wrote:
I need to save text files (let say between 1 KB to 20 MB) in a SQLite
DB.
Why not do both?
If it was me, I would write some code to split the text into sentences
(not lines - which is rather easy in English, but might be harder in
some other languages).
//...
I've received two off-line questions as to how I could parse text into
sentences in "English" even, and thought I would reply here since it
might clear up the confusion for others too.
The said questions indicated that the authors probably imagined me
possessing some fancy AI comprehending the language into what
constitutes notional sentences (Subject+Predicate) or such, but I fear
the meaning was much more arbitrary, based on common syntax for written
English - as William Faulkner wrote in a letter to Malcolm Cowley:
*"I am trying to say it all in one sentence, between one Cap and one
period."*
Think of paragraphs in English as large records delimited by 2 or more
Line-break characters (#10+#13 or perhaps only #10 if on a *nix
platform) between texts.
Each paragraph record could be comprised of one or more sentences (in
English) as records delimited by a full-stop+Space or
full-stop+linebreak, or even simply the paragraph end.
By these simple rules, the following can easily parsed into 1 paragraph
with 2 sentences and a second paragraph with 1 sentence (lines here used
as formatting only, actual line-breaks indicated with "<-" marker):
<-
The quick brown fox jumps over the
lazy dog. My grandma said to your
grandma, I'm gonna set your flag
on fire.<-
<-
Next paragraph here...<-
<-
Now a more difficult paragraph would be a the following, all of which
would translate in to 1 single sentence if only the above rules are
catered for:
<-
I have three wishes:<-
- to be outlived by my children<-
- to fly in space once before I die<-
- to see Halley's comet once more<-
<-
That will be a single-sentenced paragraph. It's up to the
end-implementation to gauge whether that would be sufficient a split or
not.
To put this into a DB, I would strip out the line-breaks inside
sentences (perhaps not strip out, but replace with space characters,
much like HTML does) to make them more easily handled as "lines". The
final DB table might then look like this:
ID | fileID | parNo | parLineNo | docLineNo | txtLine
1 | 1 | 1 | 1 | 1 | The quick brown fox
jumps over the lazy dog.
2 | 1 | 1 | 2 | 2 | My grandma said to your
grandma, I'm gonna set your flag on fire.
3 | 1 | 2 | 1 | 3 | Next paragraph here...
4 | 1 | 3 | 1 | 4 | I have three wishes: -
to be outlived by my children - to fly in space once before I die - to
see Halley's comet once more
So yes, not a perfect walk-in-the-park, but easy to do for basic text
parsing.
Stating the obvious: If the intent is to re-construct the file 100%
exact (so it scores the same output for a hashing algorithm) then you
cannot strip out line-breaks and you need to carefully include each and
every character byte-for-byte used to split paragraphs and the like. It
all depends on the implementation requirements.
The above text format should hold for 99.9% of English literature text
that can be had in text files (i.e. no images, tables, etc.). Not so
easy for scientific papers, research material, movie scripts and a few
others.
Sorry for not presenting that great AI solution. :)
Ryan
_______________________________________________
sqlite-users mailing list
sqlite-users@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users