Application of Deep Learning For Sentiment Analysis

- Deep learning is a type of artificial intelligence that employs neural networks, a multi-layered structure of algorithms. Deep learning is an accumulation of artificial intelligence statistics based on artificial neural networks for the teaching of functional hierarchies. In sentiment analysis, deep learning is also applied. This paper begins with an overview of deep learning before moving on to a detailed examination of its present uses in sentiment analysis.


INTRODUCTION
The process of detecting positive or negative sentiment in text is known as sentiment analysis. Clients are expressing their ideas and sentiments more freely than ever before, and emotion recognition is quickly to become an indispensable tool for tracking and understanding that opinion. Brands can learn what makes customers happy or unhappy by automatically analyzing customer feedback, such as opinions in survey responses and social media conversations. This enables them to tailor goods and services to satisfy their users' requirements.

II. TYPES OF SENTIMENT ANALYSIS
Models of sentiment analysis concentrate on polarity (neutral, positive, and negative), as well as emotions and feelings (sad, angry, glad, etc.), urgency (not urgent, urgent), and even intents (not interested, interested).
1. Precision in polarity is critical in business. The following are the polarity categories: Sentiment analysis, which is very pleasant, helps to determine feelings such as joy, annoyance, rage, and melancholy. 2. "The battery life of this camera is very short," for example, an aspect-based classifier would be possible to ascertain that the statement communicates a negative judgment regarding the attribute standby time. 3. Preprocessing and resources are required for multilingual sentiment analysis. The majority of such tools are available on the internet; others must be generated, and you must be able to compute in order to use it.

III. BENEFITS OF SENTIMENT ANALYSIS INCLUDE
1. Sorting data from tens of thousands of tweets, customer support conversations, or surveys? There is simply too much corporate data to manually process. Sentiment analysis aids firms in efficiently and cost-effectively processing large amounts of data.
2. Authentic Method can identify key concerns in real-time, such as whether a media platforms PR disaster is developing. Is a disgruntled consumer about to leave? Sentiment analysis methods can assist you in quickly identifying these types of circumstances so that you can take appropriate action.

IV. DEEP LEARNING
Deep learning is a type of artificial intelligence that is intentionally programmed. In order to execute feature extraction and transformation, it employs a number of nonlinear processing units. Each of the subsequent layers uses the output from the previous layer as input.

A. Sentiment Analysis Based of Customer Reviews
Deep Learning based Sentiment Analysis (DL-SA) is implemented to achieve framework for sentiment analysis.
Sentiment analysis results in identifying reviews as positive or negative. the solution is based on supervised learning method which needs training data [1]. 5. In the sentiment analytics module, the admin can analyze the sentiment based on products from positive sentiment words, products from negative sentiment words, products from neutral sentiment words and View Products Rating based on sentiment words.

B. A Deep Learning Classification Approach for Short Messages Sentiment Analysis
Peoples can communicate with each other through social media applications like whatsapp, facebook, twitter etc. social media apps get social media data from the applications and check what sentences are positive and negative sentiment using sentiment analysis. Deep learning methods like deep neural networks for using the Hindi tweets dataset and classifying them positive or negative sentiment polarity from twitter accounts. 2. The dataset is being used for text cleaning to get rid of the following things as follow [2][10]s:  Those words which have same linguistic are combined together. This is stemming or lemmatization.  All the text has been converted into lower case so that "the" and "The" will be consider as same word.  Check all digits, stop word like etc, special symbol like & has been removing it.  In preprocessing of text cleaning: non characters from each sentence has been Cleaned,  After this the sentences has been tokenized. 3. Using Machine Learning Algorithms: After cleaning the data, the following algorithm has been used. 5. We found that deep neural network model [8] stood out with a greater accuracy and the machine learning model scans the negative sentences and gives a score of 0.34 while the machine learning algorithms give a score of 0.54 [2].

C. Sentiment Analysis and Contrastive Experiments of Long News Texts
There are three methods are tested separately and the experimental results will be compared [3].
1. Dictionary-based Method: It is based on grammatical rules. Dictionary-based approaches are as follows: 2. Read the text data and segment the text.
3. Find the emotional words in the sentences and record the emotional values and positions.
4. Find the degree words before the emotional words. Set weights for degree words and multiply them by emotional values.
5. Look up the negative words before the emotional words and find out all the negative words. If then number is odd, the emotional value is multiplied by -1, and if it is even, the emotional value is multiplied by 1. Minimizing semantic granularity and using the two segment text.
Feature extraction, including feature selection and feature extraction. The steps of feature selection are as follows: Calculate the information quantity of each word in the whole corpus by adding positive chi-square statistics to negative chi-square statistics of a word.
According to the amount of information, author rank the words in reverse order and select them with the highest information quantity as the characteristics.
10. Based on Deep Learning: Deep learning is suitable for word processing and semantic understanding because of the flexibility of deep learning structure. The underlying word embedding technology can avoid the processing difficulties caused by uneven text length. Using deep learning abstract features, you can avoid the work of a lot of manual extraction of features. Deep learning has local feature abstraction and memory functions that can simulate the connection between words and words [3].

D. Sentiment Analysis Using SVM and Deep Neural Network
The use of Support vector machines, embedding deep neural networks which are suitable for the high dimensional data analysis. This will act as the generalized automated review system for various industries Such as E-Commerce platform, YouTube comments, movie review rating etc. with more high precision and has potential to replace orthodox star-based rating System. Support Vector Machine: Support vector machines (SVMs) are more of the rugged method for both classification and regression work, and have very good generalization performance. In DNN SVM joined model for the estimation analysis. The CBOW show is utilized here to take in word installing from a huge gathering from the crude content. After that DNN demonstrate is acquainted with build disseminated sentence portrayals for the more extensive info information. At last, the appropriated sentence include portrayals are being utilized as the highlights for SVM [9] classifier preparing them by learning the likelihood circulation over marks [4].

E. Sentiment Analysis of Movie Reviews Using Heterogeneous Features
System identified sentiment orientation from review text documents using a hybrid approach. The hybrid approach means a combination of Machine learning and Lexicon-based (knowledge-based) approach [5].  In this paper, the study is presented on application of deep learning or machine learning for sentiment analysis. These techniques are applied in different application like twitter, Facebook, short message, images and movies data.