Source code?

[AdityaAS]

@xiaoling Hi,

I had the pleasure of reading your TACL '15 paper "Design Challenges for Entity Linking" Will you be making the code for vinculum available anytime soon?

Thanks, Aditya

[AdityaAS]

bump

[Noxeus]

We present a simple yet effective, modular, unsu- pervised system, VINCULUM, for entity linking. We make the implementation open source and publicly available for future research.

As an academic who liked reading your paper, I am interested in the source code as well.

[xiaoling]

First of all, I apologize for the long delay. I clearly failed my commitment on time.

The past couple years have been extremely stressful for me in various aspects of my life, work, family, etc. But they're excuses after all. As I started to gain some free cycles since a few months ago, I've made progress in a release. The main challenge is to release the data together with code (code itself won't do it), including the data for candidate generation and coherence scores. For historic reasons, they were scattered in different servers and in different formats while I was at school. Re-collecting them from them, reconciling them to use the same storage format, and running regressions consumes a big chunk of time.

My current estimate will be end of June, by which I can finish all the tests for the release. Again sorry! But, being late is still better than being absent.

[AdityaAS]

@xiaoling

[XinruZhang]

@xiaoling Hello

Thanks for your works, and I have a question on the computation of coherence.

As you mentioned in the paper, In section 4.5, the way you look for the best candidate is fixing the predictions of other mentions and then calculate the coherence score. I'm confused about how do you fix other mentions? By selecting the best candidate according to the prior probability? Or some other ways?

Lok

[bennokr]

@xiaoling any current status of the code would be welcome, thanks