« Back Issue #430
  • 2015-05-05 19:50 Reported by charles.bocage
  • 2016-03-12 09:40 Marked fixed
Status: Marked fixed
Priority: Medium

Weighted Centroid-Based Summarizer

charles.bocage 2015-05-05 19:50

I have another child hood dream I would like to see come true. My childhood dream is to see a centroid-based summarizer be able to apply weights to its term frequencies.

Steps to reproduce: N/A

charles.bocage 2015-05-05 19:52

I have added the my.patch file that contains the code for the additional weighting algorithm as well as some other changes. One thing to note: the additional weight method is written but it is not hooked into the code because I was not able to get the results I wanted yet.

charles.bocage 2015-05-08 00:38

I have combed through the code and changed my camel case variable names to multi-word variable names with an underscore between each word. I also moved the output file variables to constants at the beginning of each summarizer file.

The new patch file my2.patch has been added.

administrator 2015-05-09 13:17

Hey Charles,

Thanks for my2.patch, I've pushed it to the repository. I'll keep this issue open for now, as I will probably do more testing over the next couple days before closing it.

Best, Chris

administrator 2015-05-24 20:41

Hey Charles,

The unit tests in summarizer_test.php seem broken. I fixed one syntax error which was preventing other unit tests from running, but I would like you to revise that file if possible. Here are the changes that I think need to be made:

(1) The code looks like it is based off your dutch stemmer unit tests. So there are weird variable names (they make sense for the dutch stemmer, but not for the summarizer situation). Also, the documentation above additionalWeightTestCase looks like it came from he dutch stemmer, so I deleted it. Could you add back what this function was trying to do.

(2) It would be better not to have an exit in additionalWeightTestCase as it will things when someone tries to run all unit tests.

(3) It would be better to avoid the dynamic require statements in setUp if at all possible, just use a single require_once block at the top of the file, if need be cycling over and require all the summarizers.

(4) I was having problems with it overwriting code in the lib folder, so make sure it if writes anything it is only in the test/test_files folder and it cleans up files when done.

Let me know when you have a new patch ready or if as its summer this might take a while, in which case, I will probably temporarily remove that unit test file.

Best, Chris

charles.bocage 2015-05-26 10:15

I read your note. Strange that the tests are not working for you. Can we setup a time to Skype with each other?

administrator 2015-05-26 10:25

Sure, would Wednesday at 2 work? Right now, the contents of the methods in the test files have been commented out in the file in the repository, so I can do a run all tests without it tripping up on that file.

charles.bocage 2015-05-26 10:30

Yes, Wednesday 5/27 @2pm will work for me. I will look you up on Skype around that time.

charles.bocage 2015-05-30 21:42

I have made the appropriate changes and uploaded the new patch. Again I want to apologize for such a bonehead implementation. I do not know what I was thinking.

administrator 2015-06-01 08:07

Hey Charles,

Thanks for the patch! The unit tests work now and I've pushed the changes to the repository. If you want to work on stuff in the summer, just drop me a line.

Best, Chris

administrator 2016-03-12 09:40

Has been fixed since June 2015.

charles.bocage 2016-03-12 09:40
X