Hi, I asked a similar question before, at which time the error was "Found cost=NaN" but this time it's a negative number. I double-checked and the files are...
... It's an arithmetic precision issue in computing distances. I won't be able to tell you exactly what's going on until I get more info about what you're...
Jack sent me his TestCluster.java, and the problem's in that class's definition of COSINE_DISTANCE. It turns out that TestCluster.java was just a working copy...
All, We are going to start hobby nights at Alias-i again. The goals are: 1) Provide help on how to best encode linguistic problems into NLP (Natural Language...
Any idea on how to compute word-vector matrix ( for a given large document ) using lingpipe. I checkedout the package : com.aliasi.spell and realized that I...
Hi Bob, First of all Thanks a bunch for your detailed response. Really helped a lot. I really dont need to compute the dot-product ( which is why I think the...
... You pretty much need sparse vectors for bags of words if you want to scale at all. For info on how to feed LDA, check out the last section of our...
It seems that we can evaluta the probability to a new document? I would like to know how I can implement that using the class of LatentDirichletAllocation.java...
Did you mean how to get the topciProbability Distribution for a given unseen/new document. Following are the steps to get that : First you need to get an...
Prasenjit's right about getting a point topic estimate for new docs. I'm afraid I didn't implement the probability estimate for new docs. Nor do I have any...
hi there, we use the dictionary chunkers as UIMA components with very large dictionaries. We plan to use a dictionary as singleton in a multithreaded setting...
... Nice. Singleton annotators are our recommended deployment strategy for all of our components. ... Like all the LingPipe components, they're read-write...
... Yes. But it has to be defined at the factory level. We really need a tutorial on this. If you have a filtered tokenizer FilterTokenizer with the obvious ...
Hello, Currently I have a scenario where a list of curse words is put to a trie, and then words entered by users (not text, but individual words) are run...
I'm afraid edit distance won't be sufficient due to the example you mention. This isn't an implementation issue, though it will slow down as you get more and...
Little confused, I understand when using approximate matcher, as the size of the dictionary grows the speed goes down. Question: approximate matcher tries to...
may I add the language is turkish. I have looked at both soundex and metaphone, but i think these are language dependent. I could have a list of clean words,...
... The results won't be equivalent. The string vs. string edit distance matches the whole strings against each other, whereas the approximate chunker looks...
Dear Bob, Thank you for your explanatory post. What I have done was to only check words with more then 6+ inputs, with a max distance of 2. It works fairly...
Thanks to an anonymous bug reporter for submitting these issues with $LINGPIPE/demos/tutorial/posTags First the typo: The read-me.html gives the wrong...
Has anyone successfully compiled LingPipe 3.5.1 using GNU's JDK? We received an error report for some classes that work just fine in the Sun JDKs. Here are...
We just rolled out LingPipe 3.6.0. As usual, find it on our homepage: http://alias-i.com/lingpipe Here are the details: Intermediate Release The latest...
Dear All, my name is Marco, and I'm new in this group and as lingpipe user, so do not hate me for some silly questions :-) My research project requires to...
... We don't distribute any data, but our named entity tutorial points to some sources of data. ELRA and LDC also distribute data, but it's expensive. Most...
Dear Bob, my problem is multilanguage in the sense that I handle documents that can be written in German, French, Spanish and so on, each document is written...
... Nothing free and fast, I'm afraid. We don't have corpora in French or German. Spanish is easy -- you can get it from the CoNLL data. You can get Spanish...
Hello- I'm having a blank. In the context of spell checking, is the edit distance used between the user-entered term and the suggested term, or the reverse? I...
... I should make this clearer in the doc for each operation. ... Right. It's the noisy-channel setup, so it's edits going from the suggested term to the...
Having problems with message search?
Fill out this form to ensure your group is one of the first to be migrated to the new message search system.
