This project is read-only.

program running ???

Jul 24, 2013 at 12:13 PM
hi, i am student. 3 questions.

1.which module to use for training and testing?
2.I want to change the optimization method used in the program in which the program module to be able to focus on it.
3.version in me. http://sourceforge.net/projects/crfsharp/files/?source=navbar
where can i find this version of the operating instructions.
Jul 25, 2013 at 12:31 AM
Hi Metbes,
  1. In command line, "CRFSharpConsole.exe -encode" is for training and "CRFSharpConsole.exe -decode" is for testing. For API, "bool Learn(...)" is for training and "int Segment(...)" is for testing.
  2. The optimization method is "LBFGS.optimize(..."). As the name shows, it's based on LBFGS. If you have any good idea about the optimization method improvement, please feel free to let me know.
  3. Please visit the home page of the project: CRFSharp and the downloads section. They show you detailed information about how to use CRFSharp in both command line and API, and how to use the demo packages.
I hope above information can help you, if you have any concerned, please feel free to let me know.
Jul 25, 2013 at 4:10 PM
Thank you for your answer first.
does not encode or decode the value as a parameter in me crfsharp_console version.
As far as I understand it in me testcrfencoder release training, the testing phase is carried out crfsharp_console.

I want this to interfere with the program's code.

Did you need to download the latest version.

I wrote an earlier version sorumda do it hard?
Jul 26, 2013 at 12:27 AM
To use the latest features, please visit source code section and sync the latest code from code base.
To use CRFSharp's API into your program, please visit homepage and see Use CRFSharp API in your project section. It introduces how to use APIs in your program in details.
Jul 27, 2013 at 7:55 AM
I have no idea about the optimization method. better understand the system before he Could it be better after investigating an optimization method.

decode step:


CRFSharpConsole.exe -decode ner.model nertest.txt nertestresult.txt


==>Where can I find "nertest.txt" file named.
Jul 28, 2013 at 11:46 PM
Please visit [Download] section and download demo packages. They all contain "nertest.txt" used for model testing.
Jul 29, 2013 at 11:29 AM
thanks.

I have some questions if you have time.
1.
S ==> ?
B_ORGANIZATION ==> ?
M_ORGANIZATION ==> ?
E_ORGANIZATION ==> ?
S_ORGANIZATION ==> ?

2.Do you publish it as an article that you wrote this program?

3.Is it possible to try different datasets on this program?
4.Where can I get a different DATASETS on this topic.


Thank you in advance for your response.
Jul 31, 2013 at 2:02 AM
1.
In demo package, we designed tag names as "POS_TYPE". POS means the position of the character in the term, TYPE means the named entity type of the character.
For POS, in the demo, it supports four types as follows:
S : the term has only one single character
B: the begin character of the term
M: one of the middle characters in the term
E: the end character of the term

For TYPE, the demo package contains many types as follows:
ORGANIZATION : the name of one organization
LOCATION : the name of one location
PERSON: the name of one person

2.
No article published

3.
Of course, please see CRFSharpConsole.exe introduce in home page. You can specify different dataset by parameter in command line

4.
For published data set, you can download it in [DOWNLOAD] section.
Jul 31, 2013 at 10:08 PM
Thank you very much. The information you provide is very helpful. What is the difference between L1 and L2?
Aug 2, 2013 at 6:02 AM
The weight matrix generated by L1 is usually more sparse than that generated by L2.
Aug 10, 2013 at 10:03 PM
basemodel.cs
    string[] BOS = { "_B-1", "_B-2", "_B-3", "_B-4" };
    string[] EOS = { "_B+1", "_B+2", "_B+3", "_B+4" };

what these are use for ???
Aug 11, 2013 at 1:55 AM
metbes wrote:
basemodel.cs
    string[] BOS = { "_B-1", "_B-2", "_B-3", "_B-4" };
    string[] EOS = { "_B+1", "_B+2", "_B+3", "_B+4" };

what these are use for ???
For generating feature set.
Aug 12, 2013 at 12:09 PM
I understand that the property value generated encrypted with md5
"8987446084019999210"
modelwritter.cs

foreach (long v in feature_cache_[i])
            {
                if (old2new.TryGetValue(v, out rstValue) == true)
                {
                    newfs.Add(rstValue);
                }
            }
sample: 8987446084019999210 -> 306

how're turning
Aug 12, 2013 at 2:19 PM
thread_num ==> What is different is 1 or 2
Aug 13, 2013 at 7:27 AM
metbes wrote:
I understand that the property value generated encrypted with md5
"8987446084019999210"
modelwritter.cs

foreach (long v in feature_cache_[i])
            {
                if (old2new.TryGetValue(v, out rstValue) == true)
                {
                    newfs.Add(rstValue);
                }
            }
sample: 8987446084019999210 -> 306

how're turning
It's used for updating feature ids after feature set is shrinked.
Aug 13, 2013 at 7:27 AM
metbes wrote:
thread_num ==> What is different is 1 or 2
The parameter is used to specify the number of threads used for model encoding.