[CWB] Trying the new dispersion feature

[Rodrigues Gomide, Andressa]

Hi Stefan,

Thanks for the feedback on the dispersion feature. It's still experimental, and various bits don't quite work yet. So it is probably not you getting it wrong.

More work is needed on large datasets especially.

Andrew and I have now added a "busy spinner".

The issue with the clustering of texts is one that we will need to look at in more detail.

Documentation will be done when it's all completed.

We've fixed the "new query " bug.

Thanks!

Andressa


From: cwb-bounces at sslmit.unibo.it <cwb-bounces at sslmit.unibo.it> On Behalf Of Stefan Evert
Sent: 11 June 2019 08:17
To: CWBdev Mailing List <cwb at sslmit.unibo.it>
Subject: [CWB] Trying the new dispersion feature

Hi everybody (and in particular Andressa),

noticing the new Dispersion feature in CQPweb, I tried to play around with it, but I don't quite understand how to use it (and keep getting errors from the Web server if I do something too rash).

My first mistake was not to wait long enough after starting the dispersion analysis (@Andrew: can we get one of your fancy "busy spinners" there, too?).

With a relatively infrequent query term (which I picked to make the analysis faster), I get the following plot:

[cid:image001.png at 01D52146.08C4CC00]

All the texts containing the search term end up clustered to the left instead of their normal position, which makes the plot very hard to read.  Possibly a rendering issue with my browser (Firefox 67)?

If I add a second query, the data points are also clustered to the left, though they don't occur exactly in the same texts (text overlap should be below 20%).

When I tried to add a third query with more hits (ca. 19,000), MySQL ran at 100% CPU for several minutes, but I didn't see any result in the browser (perhaps a timeout?).  Later, the DP and Juilland values do show up in the dispersion table.

A fourth attempt adds points to the plot, but now the dispersion scores (and number of texts) are "undefined".

Is there documentation for the new feature?  I'm clearly not using it correctly.

One small bug: If I select "New query" from the menu at the top right, I get redirected to a completely wrong URL (which is missing the path to CQPweb on the server: in my case I get "http://127.0.0.1/corpus" instead of "http://127.0.0.1/cqpweb/corpus/").

Best,
Stefan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://liste.sslmit.unibo.it/pipermail/cwb/attachments/20190612/1ed567c2/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 247437 bytes
Desc: image001.png
URL: <http://liste.sslmit.unibo.it/pipermail/cwb/attachments/20190612/1ed567c2/attachment-0001.png>