Big Big Things in my Little Little World: July 2015

Now a days, people tend to spend more time on social media platforms such as Facebook, Twitter etc. The interactions in social media leave a trail of huge amount of data. Major portion of this data is in textual form. The art of opinion mining is termed as sentiment analysis. It involves the classification of the text into Positive, Negative and Neutral based on the polarity. It also includes determining the attitude of speaker with respect to the topic. Knowing whether the trending tweets about the product are positive or negative helps in identifying areas of improvement for the company.The usage of social media sites is increasing day by day. Hence huge amount of textual data is generated. This data can be used to analyze the opinion of customers in social media. Thus the reputation of a product in social media can be analyzed and it can also be used to generate offers based on the customer preferences. Here comes the importance of sentiment analysis. It involves the process of extracting opinion of the speaker from plain text. This can be also termed as polarity detection. Based on the opinion of the speaker, the text can be classified as Positive, Negative and neutral. Different tools are used for the analysis of customer sentiment. Sentiment Analysis is done using different tools. The tools can be either open source or commercial. AlchemyAPI, SentiWordNet, Stanford NLP, viralheat API, Sentimatrix and python NLTK are some among them.

Types of Sentiment Analysis

Based on the algorithms used sentiment analysis can be classified into different categories. The classification can be either based on the polarity detection method or the structure of the text analyzed.

1. Classification based on the polarity detection method

Supervised- It is a machine learning technique in which a classifier is trained based on a feature set.
Unsupervised- In the unsupervised method a sentiment lexicon is used to detect the polarity of the given text.
Hybrid- A combination of supervised and unsupervised methods form the hybrid method.

2. Classification based on the structure of the text

Document level- It aims to find the sentiment for the whole document
Sentence level- Here a document is split into sentences and the opinion is analysed for each sentence.
Word level- The opinion mining is done for each word in a sentence.

Steps in Sentiment Analysis

The process of sentiment analysis involves four steps namely:

1. Pre- processing and breaking the text into parts of speech

This involves the following steps:

POS tagging: It is the process of assigning parts of speech such as noun, verb and adjective to each word in a text. It is done based on treebanks. Treebank is a parsed text corpus that annotates syntactic or semantic sentence structure. Penn Treebank is normally used for this purpose. Eg: “This is a sample sentence” will be output as “This/DT is/VBZ a/DT sample/NN sentence/NN” where DT is the determiner, VBZ is the Verb, 3rd person singular present and NN stands for singular noun.

Number	Tag	Description
1.	CC	Coordinating conjunction
2.	CD	Cardinal number
3.	DT	Determiner
4.	EX	Existential there
5.	FW	Foreign word
6.	IN	Preposition or subordinating conjunction
7.	JJ	Adjective
8.	JJR	Adjective, comparative
9.	JJS	Adjective, superlative
10.	LS	List item marker
11.	MD	Modal
12.	NN	Noun, singular or mass
13.	NNS	Noun, plural
14.	NNP	Proper noun, singular
15.	NNPS	Proper noun, plural
16.	PDT	Predeterminer
17.	POS	Possessive ending
18.	PRP	Personal pronoun
19.	PRP$	Possessive pronoun
20.	RB	Adverb
21.	RBR	Adverb, comparative
22.	RBS	Adverb, superlative
23.	RP	Particle
24.	SYM	Symbol
25.	TO	to
26.	UH	Interjection
27.	VB	Verb, base form
28.	VBD	Verb, past tense
29.	VBG	Verb, gerund or present participle
30.	VBN	Verb, past participle
31.	VBP	Verb, non-3rd person singular present
32.	VBZ	Verb, 3rd person singular present
33.	WDT	Wh-determiner
34.	WP	Wh-pronoun
35.	WP$	Possessive wh-pronoun
36.	WRB	Wh-adverb

Table 1: Alphabetical list of part-of-speech tags used in the Penn Treebank Project

Chunking: The process of identifying phrases is called chunking. Eg: “All the humans are cardboard cliches in this film”. This can be divided into two noun chunks such as: “All the humans” and “cardboard cliche in this film”.

Named Entity Recognition: It involves finding named entities such as names of persons, organizations, locations, expressions of times, quantities, monetary values, percentages, etc. E.g. “Jim bought 300 shares of Acme Corp. in 2006”. This can be parsed as “[Jim]_Person bought 300 shares of [Acme Corp.]_Organization in [2006]_Time”.

2. Subjective/ Objective classification

The different sentences in a document are classified into subjective and objective sentences. Fact based sentences are called objective sentences. E.g. “47% of Americans pay no federal income tax”. Subjective sentences consist of personal opinions, interpretations, points of view etc. E.g. “Spanish is difficult”.

3. Polarity classification

The subjective sentences are classified into positive, negative or neutral based on the opinion of the speaker in the sentence. Calculating the polarity of each sentence is very important to determine the overall sentiment. The different features of SVM classifier is mainly used for this purpose. It includes:

Bag of Words: Here a text is represented as a bag or multiset of words by disregarding the grammar while preserving the multiplicity. E.g. consider the following two text documents:

John likes to watch movies. Mary likes movies too.

John also likes to watch football games.

Based on these two text documents, a dictionary is constructed as:

{

"John": 1,

"likes": 2,

"to": 3,

"watch": 4,

"movies": 5,

"also": 6,

"football": 7,

"games": 8,

"Mary": 9,

"too": 10

}

Which has 10 distinct words. And using the indexes of the dictionary, each document is represented by a 10-entry vector:

[1, 2, 1, 1, 2, 0, 0, 0, 1, 1]

[1, 1, 1, 1, 0, 1, 1, 1, 0, 0]

Here each entry of the vectors refers to count of the corresponding entry in the dictionary. The first vector represents document 1 while the second vector represents document 2. In the first vector the first two entries are "1, 2". The first entry corresponds to the word "John" which is the first word in the dictionary, and its value is "1" because "John" appears in the first document 1 time. Similarly, the second entry corresponds to the word "likes" which is the second word in the dictionary, and its value is "2" because "likes" appears in the first document 2 times.

Negation Handling: Negation plays an important role in polarity analysis. E.g. “This is not a good movie” had the opposite polarity from the sentence “This is a good movie”, although the features of the original model would show that they were of the same polarity. So in order to handle the word “good” in first and second sentences differently, polarity is added to the word. Hence the sentence is interpreted as expressing negative opinion.

4. Sentiment Aggregation

The main task of this step is to aggregate the overall sentiment of the document from the sentences which were tagged positive and negative in polarity classification.

Applications of Sentiment Analysis

Product recommendations: Opinion mining can be used to provide recommendations to customers based on the Word of mouth.
Reputation analysis in social media: The public opinion regarding a product or service can be analyzed by text mining.

Challenges in Sentiment Analysis

o Named Entity Recognition – In some sentences it is difficult to find the topic on which the author speaks. E.g. Is 300 Spartans a group of Greeks or a movie?

o Anaphora Resolution – It is the problem of resolving what a pronoun, or a noun phrase refers to. E.g. "We watched the movie and went to dinner; it was awful." What does "It" refer to?

o Parsing – This deals with finding what is the subject and object of the sentence or which one does the verb and/or adjective actually refer to?

o Sarcasm - If we don't know the author we won’t be having any idea whether 'bad' means bad or good.

o Texts from Social media sites – Ungrammatical sentences, abbreviations, lack of capitals, poor spelling, poor punctuation, poor grammar occurs commonly in social media posts.

o Detecting in depth sentiment/emotion- Positive and negative is a very simple analysis, one of the challenge is how to extract emotions like how much hate there is inside the opinion, how much happiness, how much sadness, etc.

o Finding the object for which the opinion is expressed- For example, if you say "She won him!” this means a positive sentiment for her and a negative sentiment for him, at the same time.

o Analysis of very subjective sentences or paragraphs- Sometimes even for humans it is very hard to agree on the sentiment of this high subjective texts. Imagine for a computer