NNTargetedSentiment is a package for targeted sentiment analysis using neural networks based on package LibN3L.
Targeted sentiment analysis aims to detect the sentiment polarity of a given span in a sentence. The span can be one entity or one expression.
Please refer to the following papers for more detailed description:
Adaptive Recursive Neural Network for Target-dependent Twitter Sentiment Classification
Target-dependent Twitter Sentiment Classification with Rich Automatic Features
SemEval-2013 Task 2: Sentiment Analysis in Twitter
- Download LibN3L library and compile it.
- Open CMakeLists.txt and change "../LibN3L/" into the directory of your LibN3L package.
cmake .
make
conll format, sentences are seperated by one empty line, each line contains one word and its labels.
The label is bio format: b-xx denotes the begining word of a span with polarity xx; i-xx denotes the non-starting words of a span with polarity xx; o denotes that it is not a target.
3 o
by o
britney b-positive
spears i-positive
is o
an o
amazing o
song o
paris b-neutral
hilton i-neutral
denies o
racism o
allegations o
- o
digital o
spy o