[
https://issues.apache.org/jira/browse/OPENNLP-758?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14580590#comment-14580590
]
Anthony Beylerian commented on OPENNLP-758:
-------------------------------------------
By the way, we are referencing to some model files locally like tokenizer,
tagger, etc because we need them for the pre-processing chain.
for example :
TokenizerModel TokenizerModel = new TokenizerModel(new
FileInputStream("models\\en-token.bin"));
tokenizer = new TokenizerME(TokenizerModel);
Where [models] is a local folder in my workspace.
Thought of adding these files in the test folder, but is there a more elegant
way so we don't have to include those on the repo ?
> Unsupervised WSD techniques
> ---------------------------
>
> Key: OPENNLP-758
> URL: https://issues.apache.org/jira/browse/OPENNLP-758
> Project: OpenNLP
> Issue Type: New Feature
> Components: POS Tagger, Sentence Detector, Stemmer
> Reporter: Mondher Bouazizi
> Labels: gsoc, gsoc2015, java, nlp, wsd
> Attachments: opennlp-tools-disambiguator.patch
>
>
> The objective of Word Sense Disambiguation (WSD) is to determine which sense
> of a word is meant in a particular context. Therefore, WSD is a
> classification task, where the classes are the different senses of the
> ambiguous word.
> Different techniques are proposed in the academic literature, which fall
> mainly into two categories: Supervised and Unsupervised.
> For this component, we focus on unsupervised techniques: these methods are
> based on unlabeled data, and do not exploit any manually tagged data.
> The object of this project is to create a WSD solution (for English) that
> implements some unsupervised techniques. For example:
> - Context Clustering
> - Word Clustering
> - Cooccurrence Graphs
> - Overlap of Sense Definitions
> - Selectional Preferences
> - Structural Approaches
> - Etc.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
