Re: Advice request for tinysvm package description
Hi Justin,
thank you very much for your help.
Il 13/08/2012 12:15, Justin B Rye ha scritto:
> Giulio Paci wrote:
>> I need advice from native English speakers about the following package description:
>>
>> "TinySVM is an implementation of Support Vector Machines (SVMs), for the
>> problem of pattern recognition. Support Vector Machines are supervised
>> learning models with associated learning algorithms and have been proven
>> suitable for a large number of real-world applications, such as text
>> categorization, hand-written character recognition.
>>
>> This package contains development files required to use TinySVM into
>> your own application."
>
> (ITP #668874 also has the short description "SVM trainer and
> classifier toolkit" which is obscure but probably okay, assuming it's
> true that it trains and classifies SVMs...)
>
>> Is this correct English? Can anyone help me to improve this?
>
> It does have a few flaws.
>
> TinySVM is an implementation of Support Vector Machines (SVMs),
>
> This use of the plural seems mildly odd, and the expansion of SVM has
> the drawback that Wikipedia thinks explaining what a "support vector"
> is requires about a page of equations (and terms like "biased
> hyperplane").
I saw this use of the plural quite often, including the Wikipedia page
describing this technology:
http://en.wikipedia.org/wiki/Support_vector_machine
Do you think I should not explain the acronym?
> On the other hand Fermi lab just boils it down to
> "those points in the input space which best define the boundary
> between the classes" which might even be worth trying to fit in here.
>
> for the
> problem of pattern recognition.
>
> That's a bit awkward grammatically. And is it describing TinySVM or
> SVMs in general, and is it saying that they perform pattern
> recognition or that they do something obscure to pattern recognition
> systems?
This sentence is saying that TinySVM has been developed to perform
pattern recognition. This is one of the typical uses of SVM, but TinySVM
has been designed for it (i.e., it may be more straightforward to use it
for pattern recognition rather than for classification). However you are
right that we can drop it.
> One buzzword that turns up a lot is "perceptron"; might that belong
> somewhere in this description too?
I think it would be confusing, as perceptron is a buzzword of "neural
networks"-based classification.
> The next sentence is so much more useful that I would suggest just
> dropping the first one completely... after all, the synopsis has
> already said that TinySVM is (an implementation of) an SVM.
>
>> Support Vector Machines are supervised
>> learning models with associated learning algorithms and have been proven
>> suitable for a large number of real-world applications, such as text
>> categorization, hand-written character recognition.
>
> That either needs an "and" instead of the comma or a third item on the
> list. Google shows people giving the same first two items followed by
> "image classification" and sometimes "biosequence analysis, etc."
>
> Do we have to call it "text categorization", though? Google tells me
> it's also called "topic spotting", which seems a bit clearer.
I think "topic spotting" is more specific than "text categorization":
you can classify documents by topic, by language, by style, ... "topic
spotting" covers just the first one.
Image classification and biosequence analysis may also be included though.
>> This package contains development files required to use TinySVM into
>> your own application."
>
> This has slightly wonky grammar, but more importantly it makes it
> sound as if this is "libtinysvm-dev". Is that true? Does the package
> with this description contain header files and so on, or is it the set
> of binaries with names like svm_learn and svm_classify?
Actually I copied the libtinysvm-dev description. But the tinysvm
package has the same description, without the last sentence.
> So here's an attempt at a revised version:
>
> Description: SVM trainer and classifier toolkit
> SVMs (Support Vector Machines) are supervised learning models with
> associated learning algorithms, and have been proven suitable for a
> large number of real-world applications, such as topic spotting,
> hand-written character recognition, and image classification.
> .
> This package provides tools for developing SVMs with TinySVM.
I think that reverting "topic spotting" into "text categorization" will
fit well for the tinysvm package.
How would you change this description for the libtinysvm-dev package?
Do you think I can use
"SVMs (Support Vector Machines) are supervised learning models with
associated learning algorithms, and have been proven suitable for a
large number of real-world applications, such as text categorization,
hand-written character recognition, and image classification.
This package contains the development files."
Thank you again.
Bests,
Giulio.
Reply to: