During the decision-making process, everybody wants to hear feedback or review about the product. But, manually going through each review is very time-consuming and difficult to get a proper conclusion regarding the product which implies the need for sentiment analysis. Sentiment analysis is an automatic method used to determine whether the review about a subject is positive or negative. In sentiment analysis, for automatic review mining, there are lots of words whose different senses depend on the context they are used. This is the issue in sentiment analysis which is called word sense disambiguation. Most of the existing sentiment analysis techniques determine the polarity without any word sense disambiguation. Few methods have been proposed to achieve this as they are not able to properly disambiguate the context in which the words are used. Here is discussed a feature level sentiment analysis method, which produces a summary of opinions about different product features. A word sense disambiguation method is deployed which accurately locates the sense of a word in a sentence while determining the polarity.
Before taking some decisions or buying some products, people always want to hear others’ experiences, comments, and/or remarks (review) regarding the product which will help for decision making. Anybody can get a number of reviews and thoughts about the product that can be found in different web portals. But still, it is very difficult to read each review thoroughly and manually which is more time consuming and makes it confusing to make decisions because of a lot of reviews. So the need of an automatic way to analyze those reviews for a proper decision-making process is needed, which is called Sentiment analysis or Opinion is mining. Sentiment analysis refers to the process of determining opinions or emotions expressed in the text whether it is expressed positive, negative, or neutral. Sentiment analysis is a wide research area at the intersection of different domains: natural language processing, computational linguistics, and text mining. It arouses a lot of interest and numerous standalone applications are developed to identify sentiment and opinions, e.g., in product reviews, news, Twitter, and blogs. Sentiment analysis can be determined at different levels such as document, sentence, clause, and feature level. Those methods have their own way of analyses for sentiment presented in the review. To consider the correct sense of words in different contexts is a challenging issue in sentiment analysis. This main issue in sentiment analysis is named word sense disambiguation. Word sense disambiguation is the automatic way to determine or disambiguate the exact sense of the word in context. In a general context, word sense disambiguation identifies the intended meaning of a word in a sentence. However, sentiment analysis, refers to the process of determining whether a word in a sentence is either appeared in a positive, negative, or neutral sense.
Example
a. The battery life is a long time.
b. The hotel is very small.
c. I have a very small laptop which I can carry easily.
For example, in battery context “long time” expresses the positive sense. Similarly, in the hotel domain, “small” express negative opinion however the same word small is expressed in a positive sense in mobile and laptop domains. So, there are numerous words in a review that express some opinions about the product like large, short, lightweight, long, removed, cheap, clean, etc. but it is hard to find senses of those words without considering the context of uses. This shows that 4 how much the context of the word is important in order to determine the polarity of the sentence. Most of the errors in sentiment analysis are because the word appears in the sentence in one sense, however, the polarity obtained from lexical is of another sense. Some authors proposed sense disambiguation methods; however, locating sense in the sentence and identifying it exact matches in lexical do not provide good accuracy. However, most of these lexical dictionaries contain only the words with their associated polarities, rather than the words polarities with different senses. On the other hand, it is very difficult and time-consuming to obtain all senses of words manually for a domain and this process will be repeated again for another domain. Therefore an automatic way is needed that not only obtain all the senses of words in a particular domain but also provide a method to identify the sense of the word in a sentence. There are some words such as intensifiers and reducers which increase or decrease the polarity of other words. These words can only be handled if the relationship between intensifier/reducer and modified word are identified accurately and if the polarity of the modified word is correctly disambiguated. However, most of the existing works do not handle such kinds of words.
The main issue and challenge in sentiment analysis is word sense disambiguation. Word sense disambiguation is the process of identifying which sense (Positive, negative, or neutral) of a word is used in a sentence or in review when the word has multiple senses. The Discussed proposed method consists of two main steps: one is the development of a lexical dictionary and another one is local context disambiguation. It is proposed for the development of own dictionary which contains a semantic relationship between feature and opinioned words like adjective, verb, and adverb depending on context. Here proposed different rules for Feature Scope Identification and Building seed of word sense which are the two main parts for lexical resource development. Validation of those important sub-steps results with another method, it is found that the result of the proposed rule has high accuracy and correct result. Still, There can be extended work for negation handling which is also an important issue for sentiment analysis in this research area.
Nice articl, thanks