negation handling in sentiment analysis


m research scholar... I have done my project on Sentiment Analysis using Naive Bayes Classifier in KNIME.

Is there any node for negation handling  of sentiment analysis in KNIME ???

( like "not good", "don't like" --- "This movie is good" or "This movie is not good" . Here, 1st sentence is positive and  2nd sentence is negative but in 2nd sentence, because of word "good", the sentence is considered "Positive" )

Hi RamanSharma,

you can use n-grams features in addition to single word features for classification. For details see:

Cheers, Kilian

Hello Sir...

I have few questions

1. How can i call knime process in eclipse ???

2. Will i do mapreduce coding to call KNIME process in eclipse if i use big data extension in KNIME ???

Hi RamanSharma,

1. you can download the KNIME Analytics Platform as a standalone software. There is no need to call KNIME from eclipse. Additionally there is also a SDK version of KNIME available that allows you to develop your own extensions and start KNIME from eclipse with your extensions. For details please see:

2. this is not possible. You can use the KNIME Big Data Connectors to send SQL via JDBC to Hive or Impala, or use the KNIME Spark Executor to delegate Spark jobs from KNIME to Spark.

Cheers, Kilian

Hi Raman,

Can you please share what sentiment analysis you have done? 


Of course:

Cheers, Kilian

Thank you Sir

Hello Sir,

Please provide me research paper or algorithm of Naive Bayes classifier which is used in Knime Naive Bayes Learner and predictor node

Hi Raman,

that is the regular Naive Bayes algorithm, nothing special here. I am sure you can find a paper about Naive Bayes in Google.

Cheers, Kilian

when I am running Probabilistic Neural Network in sentiment analysis workflow

I find an error in Probabilistic Neural Network predictor ... How to resolve it ?

ERROR PNN Predictor        2:315      Configure failed (IllegalArgumentException): Duplicate column name "P (Document="")" at positions 1141 and 1142.

Please check your feature columns. It seems you are using a column "P (Document)..." as input for the node. Filter this column before hand.

Cheers, Kilian


workflow, "1- gram features and 1- and 2- gram features are used. "1- gram features means unigrams, it is OK.

But how  "1- gram and 2- gram features" (unigrams and bigrams) work together ?

and how to preprocess more than one column in the same workflow ???

In the 1-gram and the 2-gram features are joined together in one data set. This is done by concatenating the bag of words of the 1-grams and 2-grams and then pivoting (creating document vectors).

What exactlI am not quite sure what you mean with your second question " to preprocess more than one column in the same workflow?". Can you specify our question a bit?

Cheers, Kilian