Digital gaming has a history of more than 50 years. The industry started in the late 1960’s when the game titles such as Pong, Centipede and Odyssey were introduced to consumer markets. Digital gaming is now a wide spread phenomenon and at least 70% of the US and Europe households say that they play video games using different consoles such as PC, Xbox, PS4, Nintendo etc. It is reported that in 2011, the total revenue of the industry amounted to about 17 billion USD. Each game is reviewed and rated on the internet by users who played the game and the reviews are often contrasting based on the sentiments expressed by the user. Analyzing those reviews and ratings to describe the positive and negative factors of a specific game could help consumers make a more informed decision about the game.
Sentiment analysis is one of the important research area of Natural Language Processing (NLP) and have become important alternatives for automatically extracting the affective information found in reviews in the form of texts. Sentiment analysis has gain much attention from few years. Our objective is to build an NLP model to analyze gamers’ opinions expressed in games reviews. We will collect data from different gaming websites, twitter, YouTube, Facebook and many other social platforms. This study will help the game developers to identify the interest of users from a game, the users who think of buying a game after reading the reviews on different websites and it is also beneficial for parents who can restrict their children from playing different deadly games.
The popularity of social media is increasing rapidly because it is easy in use and simple to create and share images, video even from those users who are technically unaware of social media. There are many web platforms that are used to share non-textual content such as videos, images, animations that allow users to add comments for each item. There are many ways in which social network data can be leveraged to give a better understanding of user opinion such problems are at the heart of natural language processing (NLP) and data mining research.
Internet is a rich source for game reviews. Suppose being able to analyze the reviews and understand what exactly the liked or disliked. By using text mining, we can find the words that are mostly used in the reviews and how it affects the game reputation. We can analyze each term in the text and see which other terms is strongly related to. Doing so we can scale the customer liking or disliking with the game which may affect the revenue generated by the game either positively or negatively. Using sentiment analysis, we can build models on the existing reviews and be able to predict the new reviews as good or bad. Game developers can use this analysis to improve the quality of the games to meet the expectations of the general public and to maximize revenue. Also, parents can prevent their children from playing deadly games which might ultimately take their children to death.
1- Related Work:-
In Turney reviews are classi?ed by calculating the summation of polarity of the adjectives and adverbs contained within text. This study utilized game reviews, where text indicates positive and negative sentiment respectively. A discrepancy between the accuracy of the game reviews was observed with the game reviews getting a higher accuracy. This was attributed to the fact that game reviews, whilst being positive, can have a lot of adjectives and adverbs that do not fully relate to the overall enjoyment of the game and can actually be more a description of the scenes within the game itself. The PMI-IR (Pointwise Mutual Information – Information Retrieval) algorithm was used to classify documents. This algorithm works by taking the relevant bigrams from the document then using the near function on a search engine to see how many times this bigram appears near a word that expresses strong positive or negative sentiment, a large number of matches indicates a stronger polarity. Consider word presence vs frequency where word presence is found to be more e?ective than word frequency for sentiment analysis. Word position within a given sentence can also be e?ective, where such information can be used to decide if a particular word has more strength at the beginning or the end of a given sentence. Considers objective tweets as well as those that are positive and negative sentiment. This paper discuss the method for collecting corpus data, this again is similar to by using emoticons for positive and negative sets Generally there is a far greater di?erence in the objective and subjective texts than positive and negative sets, such di?erences show that using POS tags can be a strong indicator of the di?erence between types of text. The objective and subjective comparison shows that the interjections and personal pronouns are strong indicators of subjective texts whilst common and proper nouns are indicators of objective texts. Subjective texts are often written in ?rst or second person in past tense whilst objective texts are often written in third person. The di?erence between the positive and negative sets do not give a strong indication, however they are good indicators in the di?erence between the amount of superlative adverbs and possessive endings, both indicating positive sentiment whilst the negative set often contains more verbs in the past tense as people are often expressing disappointment. use multi nominal naive Bayes classi?er to compare unigrams, bigrams and trigrams they conclude that bigrams give the best coverage in terms of context and expression of sentiment.
2- Sentimentor: Sentiment Analysis Tool:-
Sentimentor is a web based tool which uses naive Bayes Classi?er to classify live Twitter data based on positivity, negativity and objectivity. Sentimentor has an interface which enables the user to analyze the word distributions.
3- Data Collection and Preprocessing:-
Twitter API was used for the data extraction process. Negative, positive and objective texts were collected by following the same procedures as in Tokenization process from and were followed for the data preprocessing task. The steps followed included the removal of any urls and usernames (usernames follow the @symbol) and removal any characters that repeat more than twice turning a phrase such as OOMMMGGG to OOMMGG, which is applied by a regular expression. Finally, the stopset words were removed from the data. The stopset is the set of words such as “a”, “and”, “an”, “the”, these are words that do not have any meaning on their own. The second phase is associated with determining the POS tag for each word. The OpenNLP library was used for POS tagging and the extraction of unigrams and bigrams.
4- Evaluation of the data set:-
Overall singular noun (NN) is the most common POS tag, occurring 29.08% across the whole corpus. Preposition or conjunction (IN) occur 10.28% of the time with it being clear that there is a signi?cant di?erence between the occurrences in all sets. To better understand the di?erences between sets we have calculated the percentage di?erence between the percentage distributions of each POS tag. This has been done for the di?erence between the objective and subjective sets and between the positive and negative sets. The common nouns and proper nouns are a strong indicator of the subjective set by looking at common noun plural (NNS) nouns proper singular (NNP) and noun common singular (NN). According to Pak we expect writers of subjective text to be talking in the ?rst or second person, we can partially con?rm this by looking at the di?erence of verb present tense not third person singular (VBP) and verb past tense (VPD) however verb, present participle (VBG) contradicts this as it prevails in the objective set. This could have happened because the selected news outlets might have more comment on news than original reporting or this could be a di?erence in the POS tagger, however this is of little concern because the di?erence is relatively negligible. Likewise we can expect objective texts to be in third person the results for verb present tense 3rd person singular (VBZ) can con?rm this. The List item marker (LS) has a -100% di?erence as this doesn’t occur in the objective set this tag isn’t present in Pak 3 data. The symbols that have not been removed by the tokenizer are a potential source of error as these represent signi?cant di?erence between sets. The POS tagger used can detect foreign words (FW) which have a Strong indication of the text being subjective, the reasoning for this is because news outlets would only be expected to use correctly structured English, standard user tweets may contain a mix of languages even though the Twitter search was limited to English tweets. The strongest indicators for negative sentiment is Currency ($) and quotation marks while an individual is highly likely to express their ?scal issues in a negative sentiment but as there are only 19 occurrences of currency in the system this is not a good indicator of what set the text belongs to, also the inclusion of quotation marks here is likely going to introduce error into the system. Wh-adverb – negative (WRB), particle (RB, RP) genitive marker (POS) are all strong indicators on negative sentiment state that (POS) may be an indicator of positive sentiment, the results we have collected contradict this. Superlative adverb (RBS) , proper noun singular (NNP) , adjective superlative (JJS) , Noun (NNS) common plural are all indicating strong positive sentiment.
The naive Bayes classi?er was used for classi?cation this decision is primarily based on ?ndings that the naive Bayes classi?er show good performance results.
:: Where C is the class positive, negative or objective sets, m is the twitter message and f is a feature. In our experiments the features are POS tags, unigrams or bigrams.
We have tested our classi?er against a training set which contains 216 manually tagged tweets. We have provided the test results for unigrams and bigrams both with and without the use of POS tags previously mentioned tests. The test with the highest accuracy is the one using bigrams without POS tags with an accuracy of 52.31% and the lowest is Unigrams without POS tags at 46.76%. Accuracy would be far higher if we were to carry out these tests using binary classi?cation and it should be stated that this is one of the further complexities of using microblogging data as appose to using reviews as these are not expected to be objective. The use of bigrams has show an increase in performance with or without the use of POS tags. This also reduces the amount of false positives in the objectivity classi?er however there is also notable increase in false positives by the positive classi?er, the negative classi?er does not seem to be e?ected much by this. Overall the use of POS tags has had a negative e?ect on the accuracy of the calssi?cation proccess, this is caused by the Ambiguity of POS tag occurances across sets this is most likely also the case because we using the summation of POS tags in a given phrase and not looking for binary occurance. It may potentially beni?t the classification proccess to give less weighty to the POS tags or to experiment with di?erent n-grams of POS tags. We have con?rmed previous works ?nding to be correct in there conclusion that bigrams give better results than unigrams. The overall performance of the system is satisfactory, however we would still like to further improve this as outlined in our future work section.
7- Conclusion and Future Work:-
In this paper we have presented a way in machine learning techniques can be applied to large sets of data to establish membership, in this case positivity, negativity and objectivity. We have looked at common process in NLP that can help us derive the meaning or context of a given phrase. We have demonstrated how to collect an original corpus for sentiment classi?cation and the re?nement that is needed with such data. We have applied a naive Bayes classi?er to this set conduct sentiment analysis and have found this process to be successful. On analysis of our results we have con?rmed that bigrams o?er better performance when conducting the classi?cation process. We has also con?rmed some of ?ndings when looking at the di?erences between the objectivity and subjectivity set, the same can’t be be said for the positive and negative sets which prove to be far more ambiguous. We have discovered that collecting data across a short amount of time may be a potential source of error when determining sentiment this is due to the fact that opinions can shift over time as can the meaning of words. The classi?cation process itself has been successful with and accuracy of 52.31% however it is felt that this could be further improved, this is outlined in future work. One of our future works is to experiment with di?erent classi?ers on our dataset. We also intend on developing an application which carries our textual analysis on movies analyzing what an audience is expressing and adjusting the enviroment accordingly.