Summarizers

[fabcouwer]

Rebased this to master. Do you think it is good to merge, @MBrouns?

[MBrouns]

Don't think so yet, still some cleaning up to do and I still need to write some stuff to the database like the keywords for the document. Will try to finish this tonight

[MBrouns]

I made quite some improvements on the summarizer and I think it's about done. Also created an executable jar that can be called from cakePHP. @fabcouwer could you go over the changes? If you agree it's ready for merging

[bouke-nederstigt]

Once it's merged I'll write some code in cake to call the program. I think cakephp has some nice shell functions that should make this easy. Op 13 mrt. 2014 20:11 schreef "MBrouns" notifications@github.com:

I made quite some improvements on the summarizer and I think it's about done. Also created an executable jar that can be called from cakePHP. @fabcouwer https://github.com/fabcouwer could you go over the changes? If you agree it's ready for merging

Reply to this email directly or view it on GitHubhttps://github.com/yetti4/crowd-summary/pull/5#issuecomment-37574429 .

[bouke-nederstigt]

Just looked at it and it looks pretty good to me (as far as my limited java skills go). Only thing I was wondering about is about preserving the order of the summaries. If all sentences in the summary are ranked 1, how do we know what order to place the sentences when retrieving the summary?

[MBrouns]

That's a good point. Initially we can use the order from the text I think.

I'm currently making a small improvement to the sentence splitter in order to fix the table of contents for example to show up as a single sentence

[bouke-nederstigt]

Wouldn't that mean comparing all the sentences to the entire article every time it's displayed? Might be easier to create an extra db column order to set order when the summary is created? Op 13 mrt. 2014 20:25 schreef "MBrouns" notifications@github.com:

That's a good point. Initially we can use the order from the text I think.

I'm currently making a small improvement to the sentence splitter in order to fix the table of contents for example to show up as a single sentence

Reply to this email directly or view it on GitHubhttps://github.com/yetti4/crowd-summary/pull/5#issuecomment-37575964 .

[MBrouns]

No that wouldn't be necessary. The summarizer stores the sentences in the db in the order they appear in the document which means you can simply use the pk of the sentences table to determine the order

[bouke-nederstigt]

Yeah should work for the initial summary. But it might be problematic when updating the generated summary with extra sentences from the user summaries. Op 13 mrt. 2014 20:29 schreef "MBrouns" notifications@github.com:

No that wouldn't be necessary. The summarizer stores the sentences in the db in the order they appear in the document which means you can simply use the pk of the sentences table to determine the order

Reply to this email directly or view it on GitHubhttps://github.com/yetti4/crowd-summary/pull/5#issuecomment-37576464 .

[MBrouns]

Agreed, but maybe then the sentence order is user specific and should go in the user_sentences table?

My better sentence splitter is complete btw. so for me it's ready for merging

[bouke-nederstigt]

That will only solve the problem for the personal summaries. Not for the one improved by the system right? But let's just leave it for now, we'll tackle the problem once we actually run into it. As far as I'm concerned you can merge.