Add types/components/resources related to sentiment analysis

[reckart]

Currently, dkpro-core misses types, components and resources related to sentiment analysis.
Since, there are people working in different projects in this area, I believe we should
have a type system for that, along with some resources and even some components to
be reused by everyone.

Original issue reported on code.google.com by pedrobssantos on 2013-11-11 11:46:13

[reckart]

This should not be a module in DKPro Core. It should better be an independent project
and moved to DKPro Core if the following conditions are true:
- Others want to contribute and reuse
- There exist 3rd-party libraries for sentiment analysis

Original issue reported on code.google.com by nico.erbs on 2013-11-11 14:30:22

[reckart]

I have been thinking about DKPro Sentiment for a while.
Is there a general model beyond dictionary-based methods that the project could model?

Original issue reported on code.google.com by torsten.zesch on 2013-11-11 15:24:20

[reckart]

We couldn't think of any more generic model but we haven't invested too much time. I
think it is the right time to start such a project and invite others with more experience
about sentiment analysis to contribute.

Original issue reported on code.google.com by nico.erbs on 2013-11-11 16:56:40

[reckart]

FTR: there is a demo module in DKPro-TC which does sentiment (polarity) detection.

Original issue reported on code.google.com by daxenberger.j on 2013-11-12 08:21:06

[reckart]

It appears, there will be sentiment analysis in the next CoreNLP version:

http://nlp.stanford.edu/sentiment/code.html

Code for Deeply Moving: Deep Learning for Sentiment Analysis

The original code was written in Matlab. Due to the strong interest in this work we
decided to re-write the entire algorithm in Java for easier and more scalable use without
license restrictions. In this alpha release we focus on the testing portion of the
code that gives the published results. The code already includes the training procedures.
At the end of October we will update this package with further documentation and faster
training code.

This model will be part of the next Stanford CoreNLP release and the below code includes
the entire CoreNLP package. It includes the model source code, as well as the parser
and sentence splitter among other functionality.

Download alpha version: stanford-corenlp-full-2013-10-21.zip (211MB)

You can run this code with our trained model on text files with the following command:

java -cp "*" -mx5g edu.stanford.nlp.sentiment.SentimentPipeline -file foo.txt

Original issue reported on code.google.com by richard.eckart on 2013-11-12 09:13:08

[reckart]

Hi ..I am looking for a trusted and easier way to do text-based sentiment analysis with
Matlab .. and what I expect is that in the result of each row of my matrix (each row
includes a text file)  show me positive or negative or even if there is any professional
tools or code which let me to have Author personality could be very very useful.
thanks in advance

Original issue reported on code.google.com by asana86 on 2014-05-18 12:47:42

[reckart]

@asana86: Please do check out page from Stanford mentioned above. We plan do integrate
these components into DKPro Core at some point, but currently we did not do so. Mind
that the everything below the link in the comment above has been copied from the linked
page for our convenience. We DKPro Core people did *not* create or contribute to this
sentiment analysis code.

Original issue reported on code.google.com by richard.eckart on 2014-05-18 13:21:19

[reckart]

Hi, I've implemented a DKPro annotator for sentiment analysis on sentence level with
5-scores granularity of sentiment (wrapper of Stanford CoreNLP 3.4 sentiment analyzer,
http://www-nlp.stanford.edu/sentiment/code.html). Pros: Out of box functionality, pre-trained
models, good results. Cons: Currently it runs the whole Stanford pipeline (tokenization,
POS, parsing), so it's slow; better seamless integration would require mapping existing
DKPro annotations (POS, parse tree) to Stanford annotations, not sure how demanding
that is. What will be the best way to integrate it to the next release of DKPro core?

Original issue reported on code.google.com by ivan.habernal on 2014-07-17 11:57:49

[reckart]

I believe we have code somewhere to transform DKPro Core annotations into the Stanford
structures. I need to dig around a bit to find those.

I would suggest to add the component as is to DKPro Core GPL trunk into the CoreNLP
module and then we'll look what needs to be done to improve it. I'd also suggest to
open a separate issue for the CoreNLP sentiment integration - this issue is just about
adding annotation types for sentiment analysis which should be independent of a particular
sentiment analysis implementation. We should probably add a new api.sentiment module
for these types.

Original issue reported on code.google.com by richard.eckart on 2014-07-17 12:02:16

[reckart]

Ok, I opened a separate bug for it #424 and submitted two patches (one for dkpro-asl
with typesystem, one for dkpro-gpl with wrapper for Stanford sentiment analyzer).

Original issue reported on code.google.com by ivan.habernal on 2014-07-18 11:59:31