Cannot use the web service

[dragoon]

Hi, how can I run this tool? I'm trying to re-use your tweet segmentation function, but it seems that MS web n-gram service is not working:

Apr 03, 2014 5:02:28 PM com.sun.xml.internal.ws.protocol.soap.MUTube getMisUnderstoodHeaders
INFO: Element not understood={http://www.w3.org/2005/08/addressing}Action
SOAPFaultException MustUnderstand headers:[{http://www.w3.org/2005/08/addressing}Action] are not understood  com.sun.xml.internal.ws.protocol.soap.MUTube.createMUSOAPFaultException (MUTube.java:133)

[tranchis]

Hi,

How have you tried to run it? As far as I remember, I was testing this directly by running the Clojure code on the REPL, with the rest of the WS-related classes in the classpath. The algorithm runs quite ok but it was a bit inefficient, so we decided to explore other options.

Anyway, if you give me a few hours I can prepare the project so that it can be run/tested from the command line.

[dragoon]

Yes, I got this error while running it from REPL, but I'm not sure everything was in the classpath, I thought it is added automatically.

Thanks a lot for help!

[tranchis]

Hi again,

Check the last version of the repo. You can load src/main/clojure/com/github/kemlg/semantweets/core.clj on the REPL and then execute something like this:

(take N (rank-segments (get-phrase ["leo messi is better player than cristiano ronaldo"])))

Where N is the number of top ranked solutions that you want. In the case of the Top 5, the result is:

([([""] ["leo" "messi"] ["is"] ["better" "player" "than"] ["cristiano" "ronaldo"]) 0.6306097560301133] [([""] ["leo" "messi"] ["is"] ["better" "player"] ["than"] ["cristiano" "ronaldo"]) 0.6306010440850567] [([""] ["leo" "messi"] ["is"] ["better"] ["player"] ["than"] ["cristiano" "ronaldo"]) 0.6306007890704534] [([""] ["leo" "messi"] ["is"] ["better"] ["player" "than"] ["cristiano" "ronaldo"]) 0.6306003191714861] [([""] ["leo" "messi"] ["is" "better"] ["player"] ["than"] ["cristiano" "ronaldo"]) 0.6305911620763714])

Which means it sort of works, as Leo Messi and Cristiano Ronaldo have been found as strong candidates of semantic concepts related to Wikipedia.

Let me know if you can successfully run it and what you think of it. Disclaimer: it is truly inefficient :P

[dragoon]

Hi, thanks a lot, at the same time a made a quick hack (in the fork) that uses the REST N-gram service instead of XML, this way you don't need to deal with JAX WS and generate files.

I think it works fine, but I don't see why it's inefficient. The only bottleneck I see is the query to the web service, which is of course slow, but to overcome it you will need a local installation.