Sentiment Analysis in Go

Sentiment Analysis is the process of identifying and classifying assessments which are present in form of text, that represents the inclination of writer’s sensibility towards a particular entity.

Sentiment analysis is particularly useful for monitoring of reviews, tweets or feedback.

Here, we are analyzing sentiments out of IMDB Movie reviews.

This dataset is available at: https://www.kaggle.com/lakshmi25npathi/imdb-dataset-of-50k-movie-reviews

Dataset contains 50k movie reviews along with the sentiment mentioned, but we are going to just use the movie review and not the provided sentiment in the file.

We are going to use the go package : https://github.com/cdipaolo/sentiment

This is just a started code where we are going to read reviews from a CSV file and print the sentiment of the movie review.

First, we have to install the package by running

go get -u github.com/cdipaolo/sentiment

Create a main.go file in your project directory and you can use the following snippets.

Importing the packages

package main
import ( 
  "encoding/csv" 
  "fmt" 
  "io" 
  "log" 
  "os"  
  "github.com/cdipaolo/sentiment"
)

Reading the CSV file that contains the movie reviews

func main() { 
  // Open the file 
  csvfile, err := os.Open("data/IMDBDataset.csv") 
  if err != nil {  
    log.Fatalln("Couldn't open the csv file", err) 
  } 
  defer csvfile.Close() 
  // Parse the file 
  r := csv.NewReader(csvfile)

Create a sentiment Analysis Model

• This sentiment analysis model uses Naive Bayes classification model for prediction.

// Creating Sentiment Analysis Model 
model, err := sentiment.Restore() 
if err != nil {  
  panic(err) 
} 
var analysis *sentiment.Analysis

Iterating through records, we read each record and determine its sentiment.

// Iterate through the records 
for {  
  // Read each record from csv  
  record, err := r.Read()  
  if err == io.EOF {   
     break  
  }  
  if err != nil {
     log.Fatal(err)  
  }  
  analysis = model.SentimentAnalysis(record[0], sentiment.English)  
  var sentiment string  
  if analysis.Score == 1 {
     sentiment = "positive"  
  } else {   
     sentiment = "negative"  
  }  
  fmt.Printf("Review: %s \n and Sentiment:%s\n", record[0], sentiment)  
}

Model by default returns 0 or 1 for the sentiment(negative=0, positive=1). Here, we just put up a conditional statement to print out the actual sentiment of the review.

Full Code