hallo,

I need to extract a word and few text that
precedes and follows it (about 30 + 30 chars)
from a long textual document.

Like the description that Google returns when
it has found a given word.

In example from:

"Sylvia Brunner, a marine mammals researcher at the museum in Fairbanks,
identified the decomposing carcass and oversaw its recovery on Wednesday.
The "bloated, black thing on the beach" was about 12 feet from the river's
edge, she said."

I have to find the 'carcass' word
and finally return with:
"identified the decomposing carcass and oversaw its recovery on"

---

Which is the *fast* method in VbNet?

In VB6 I would have used
InStr (with Binary option because faster)
to find the position of the word,
then Mid to extract the preceding text,
then Mid to extract the following text,
then build up my phrase in this way:  text1 & word & text2 .

Any suggestion in VBNet ?
New methods, StringBuilder, Regular Expression... or what else?

--------


Thanks
(examples are obviously very appreciated  ;-)  )

teo,
It sounds like you have the methods identified, you simply want someone else
to test them for you. Why not test them yourself, as you probably already
have the situation (program) and data to test them with.

I would probably use a regular expression, as regex feels like the "correct"
solution (not necessarily the fastest method). The trick is going to be
ensuring that it is an efficient expression and not a poorly performing
one... For example using a lazy compare instead of a greedy compare on the
30 before & after...

If I have time later I will see what RegEx I can come up with...

Show quoteHide quote

Use the same method as you would in VB6. Use the IndexOf method to find
the string and the Substring method to extract the part of the text.

However, I don't see the reason for getting the preceding text and
following text, just to put them together, when the string that you want
already exists in the text.

teo wrote:
Show quoteHide quote

On Sat, 17 Jun 2006 23:12:22 +0200, Göran Andersson <gu***@guffa.com>
wrote:

The fact is that I'm building a searching engine, and I need to
format the searched word as Bold,
so I'm compelled to have two chunk of text ,
so I can format my final string as this:
plain Text1 + Bold word + plain Text2.

Because I have to extract the integral text from a column of a DB
(then extract only a part of it, as described above),
do you know if SQL syntax is able to perform such extraction?

Or I'm compelled to extract the string using the VB methods
after having stored the integral text in a DataReader?

Rather then search for the text each time, have you considered, "indexing"
each document.

Then when you need to do a search, you simply check the index, the index
would return where in the text the word was found.

Show quoteHide quote

On Sat, 17 Jun 2006 22:42:07 -0500, "Jay B. Harlow [MVP - Outlook]"
<Jay_Harlow_***@tsbradley.net> wrote:

I know that there is such an option,
but I didn't think about it about a solution
because it seemed to me that it would have required
a lot of job to do firstly ;
also going to search a given word among the resulting huge list of
indexed words,
I think it would require a lot of time, maybe the same time than
it would require searching for the given every time.
I'm only guessing about, I've no benchmark....

Maybe I'm going to implement such a solution
when I've finished this method I've started to develop now.

----------

Another question:

the RegExp sample we discuss above
returns 30 + 30 , regardless how the 30 on the left start.
I'll try to explain what I mean.

I'd like to have the chunk of text on the left
starting where the sentence containing the given word starts,
(so to have the very first letter capitalized),
like the way Google displays the results,
that is, if you search  'Lewinsky'
Goggle returns with:

To maintain the *Lewinsky* Story's original feel, we will leave much
of this ... These were gifts the president had originally given
to Ms. Lewinsky himself. ...



In this way,
the 'T' letter is at     # -16 position,
I renounce to the preceding 14 chars
and decide to start straight at  # - 16
and decide to increase the chunk of text on the
right to 44  ( = 30 + 14).

If the "T" isn't within the 30 chars on left,
no problem, I accept the old 30+30 solution.

Is it possible?

-----

Basically,
we need to trace of the   .   (= dot)  char
that signals to us that a sentence (within to 30 left) is going to start.

If a single RegExp doesn't workt,
we could maybe go with doubling the first RegExp (60 +60)
and then
with a second RegExp find the  dot   char
and then simply extract the following 60 chars chunk of
text on the right.

What about this?

Teo,
Here is a regex:

        Dim input As String = "Sylvia Brunner, a marine mammals researcher
at the museum in Fairbanks, identified the decomposing carcass and oversaw
its recovery on Wednesday. The ""bloated, black thing on the beach"" was
about 12 feet from the river's edge, she said."
        Dim pattern As String = ".{1,30}?carcass.{1,30}"

        Dim match As Match = Regex.Match(input, pattern,
RegexOptions.Multiline)

        If match.Success Then
            Debug.WriteLine(match.Value)
        End If


Show quoteHide quote

Thanks;
I didn't want you to do the work for me indeed,
I only liked to know the name of the functions
it is advisable to use for this case...




Show quoteHide quote

Learning Programming
FileSystemWatcher raises Changed Twice....
Memory Leaks
Overrides and mybase ??
Refer to an Instance of a form
Newbie Question
Timer1_Tick is never fired
Date time format
News content into an XML file
Can we seperate the Bussiness Logic and User Interface?