I rebuilt LingPipe 2.2.1 on the web site so that it would work with Java 1.4. I hope this clears up major/minor version problems. - Bob Carpenter Alias-i...
235
Sanjay Singh
sspal
Mar 29, 2006 9:50 pm
Hi, I have a question regarding the sentiment api. The LMClassifier does a good job in getting the best catgeory among "pos" and "neg". How can I get a ...
236
carp@...
colloquialdo...
Mar 29, 2006 10:40 pm
... You'll have to get that through the API -- the tutorial doesn't cover it (yet, anyway). The classification API is set up to run through implementations of...
237
npeterson808080
Apr 1, 2006 3:45 am
I'm trying to write a java program that will label named entities of any given string, but I can't quite figure out how to do it using LingPipe. Exactly what...
238
carp@...
colloquialdo...
Apr 3, 2006 5:32 pm
... That's about what our demo does. But we don't have dates turned on in the demo models, and if we get this case right, "McDonald's" will be of type...
239
todd0053
Apr 4, 2006 3:46 am
I want to train the Named Entity Detector to extract entities that I specify. In the command demo it says this can be done using WordFreak. If I did the...
240
npeterson808080
Apr 4, 2006 9:58 pm
Hi. Yes, I looked through the demo and saw that method, but what I get confused by is that in that method I need an already defined NEChunker. But how do I...
241
carp@...
colloquialdo...
Apr 5, 2006 5:08 pm
... We ship a couple of them. One's trained on English news and the other on genomics data from MEDLINE. For instance, here's the code to create one in the...
242
reckb
Apr 5, 2006 10:55 pm
... Do you have WordFreak working from the included command? ... Have the annotated files in a directory and run the training commmand as illustrated in the...
243
jude.yang
Apr 10, 2006 11:22 am
Hi, I have used EN_GENOMICS.model to identify biomedical terms in some medline abstracs. I need the term class information in the tagged abstracts for our...
244
carp@...
colloquialdo...
Apr 10, 2006 6:32 pm
... Training data is GENIA 3.02 (the most recent release as of today). ... Yes. The types are defined on their ontology page: ...
245
npeterson808080
Apr 10, 2006 7:53 pm
Hi. In one of the recent updates of LingPipe (starting with 2.2.0, I think), has there been a change in how LingPipe marks sentence boundaries when there is an...
246
carp@...
colloquialdo...
Apr 10, 2006 8:31 pm
... Yes. But it's easy to recreate the old behavior. ... But it's possible to change this based on the parameters in the models. The long answer is the...
247
seth_a_farrington
seth_a_farri...
Apr 11, 2006 6:51 pm
Bob- ... I think that I might be accidentally training with the same data multiple times. Can you explain a little more about how to use the "cross-entropy...
248
carp@...
colloquialdo...
Apr 11, 2006 7:06 pm
... The number is a log (base 2) probability. Probabilities are numbers between 0 and 1. For instance log2(1) = 0 log2(1/2) = -1 log2(1/4) = -2 log2(1/8) = -3...
249
zhblu
Apr 11, 2006 9:24 pm
I tried to parse medline with lingpipe 2.2.1. When I used MedlineCitation.keywordLists, it always returned a list with 0 length. Anyone can help? Thanks...
250
carp@...
colloquialdo...
Apr 11, 2006 9:48 pm
... Are you sure you're looking at citations that have keyword lists? Most of them don't seem to have keywords. If you know you have keyword lists that...
251
maikuraki_standup
maikuraki_st...
Apr 13, 2006 8:09 pm
I tried using the code below (trained on the EN_NEWS model) to do some named entity annotation, and I got a few errors I was wondering how to fix. Namely, in...
252
Bob Carpenter
colloquialdo...
Apr 13, 2006 10:01 pm
... I wish fixing this was so easy. The models are statistical and reflect the distributional aspects of their training data perhaps a bit too closely. The...
257
reckb
Apr 18, 2006 6:01 pm
Apologies for the new spate of spam. New memberships now require approval and posts are moderated as well. Once the banned member info gets propogated across...
258
Ravikumar.Kondadadi@...
kondadadi
Apr 18, 2006 7:01 pm
Hi I'm trying to use the class com.aliasi.dict.ApproxDictionaryChunker to do some fuzzy matching. I created a trie-dictionary with all my dictionary terms and...
259
reckb
Apr 18, 2006 7:36 pm
... Your mother board is facing south--it will never work that way because of some bad experiences I had as a youth.... But on a more serious note, you just...
260
seth_a_farrington
seth_a_farri...
Apr 22, 2006 9:40 pm
I'm still struggling with working with large language models. My DynamicLMClassifier trained with an ngram of 8 over 100,000 documents of about 1K each results...
261
Bob Carpenter
colloquialdo...
Apr 23, 2006 4:26 pm
... Pruning should make a very big difference in size, even pruning to min count of 2. If that isn't making your models much smaller, could you send the code...
262
julioxgodoy
Apr 23, 2006 4:28 pm
hi! i am working in my thesis resolving anaphoras related to geo-references in spanish texts. I would like to know if it's possible to use Lingpipe to find...
263
seth_a_farrington
seth_a_farri...
Apr 23, 2006 8:43 pm
Bob- ... My pruning code is pretty simple: private static void prune (DynamicLMClassifier classifier, int threshold, long trainedBytes) { if (threshold < 2) ...
264
reckb
Apr 25, 2006 12:14 am
... I assume that you are going to be developing your own anaphora resolution algorithms--the one in LingPipe is based on my thesis, heuristic and going to be...
265
Bob Carpenter
colloquialdo...
Apr 25, 2006 3:30 pm
... The algorithm's javadoc-ed, but it assumes first-best named-entity input and only performs coreference within a single document -- not across documents. ...
266
Bob Carpenter
colloquialdo...
Apr 25, 2006 3:30 pm
... The FixedWeightEditDistance() no-arg constructor sets match weight to zero (0), and all other weights to infinity. I'm guessing you don't have upper-case...
267
Anton Jones
gyrocyclist69
Apr 25, 2006 6:32 pm
Hi, I'm comparing the performance of various coreference resolvers on the MUC6/7 corpuses. At the moment I'm looking at OpenNLP, JavaRAP, and LingPipe. So far...
