Opinion Mining/Sentiment Analysis
Chapter One
Preamble of the study
Sentiment analysis application areas are for example, a brand tracking what bloggers are saying about a new product or service. As more consumers make purchases online, this potential customers go through reviews left by other customers who have purchased a similar product often basing their decision to buy on the ratings. A five star product is more likely to be preferred over a 2 star product. Tracking this opinions is in the interest of the product manufacturer as well. However, the application of sentiment analysis is not restricted only to consumer goods sector. Sentiment analysis tasks could be applied to get political opinions of anybody. It could be applied to newspapers stories or editorials. This when applied to different news sources could help highlight different opinion holders in media. This knowledge can then be used for targeted adverts. Also, businesses can track new product perception, detect flames and general brand perception.
Chapter Two
Literature Review
Sentiment Analysis
The web is growing rapidly. The rise of Web applications that promote interactions between users is at the forefront of this growth. Web 2.0 platforms, those sites that fuel this growth include facebook, twitter, review aggregation sites such as epinions and rotten tomatoes have revolutionized the way and amount of data that is created and consumed on a daily basis. The mainly static pages offering little interaction between the users has been transformed to user- driven community based social networks. This also extends to e-commerce sites where users can post comments and opinions about products or services. Forums, wikis and blogs relating to different domains and interest groups generate huge amounts of data as well. This data contains valuable information that could be analyzed for the sentiment and opinions expressed. Both businesses and customers want to get opinions but for different reasons. Businesses can use this information to learn what is liked or not about products. The general sentiment expressed on a recently released product could be used for product or service upgrades or to counter the opinions raised as the case may be. Customers or consumers on the other hand, want to learn of past experiences of others before making decisions to buy or not. People generally seek for opinions of friends and family when thinking of purchasing a good of service. In the days gone by, word of mouth was the only option or adverts on bulletin boards. However, nowadays the e-commerce sites that sell the product also carry reviews by past users of the product. Many other special interest offer profess recommendations such as www.cnet.com, www.zdnet.com for technology related issues, www.rottentomatoes.com, www.imdb.com for movie and series ratings.
A study conducted by comScore and Kelsey Group (2007) revealed the impact online consumer generated review on the price consumers were willing to pay for a service that is delivered offline. Over 2000 US residents were surveyed in the study. It was revealed that consumers were willing to pay at least 20 percent more for services that receive a 5- star rating (excellent) rather than for the same service receiving a 4-star rating (good).
Of those that consulted an online survey, 41% of restaurant reviewers subsequently visited a restaurant. This is closely followed by 40% for hotel reviewers that subsequently stayed at a hotel. The study also found out that reviews generated by users were more influential than professionally sponsored reviews.
This study was conducted in 2007. With the rapid growth of technology and ubiquity of internet connected mobile devices now, it is likely that the number of internet users relying on such reviews has grown significantly. Independent review aggregation sites also exist. These sites such as UK based which offer advice on a wide range of products and services and conducts extensive tests by independent professionals providing objective information to consumers.
Chapter Three
Research Design and Methodology
Yahoo BOSS API
Yahoo provides a service that allows developers to integrate yahoo search into their web or desktop applications. The API supports various search operations. The following boolean operations are supported – AND, OR (|, &)9. However, the NEAR operator is not supported. Previous work by (Turney 2001) has also shown that NEAR performs better than the AND operator for proximity queries. Therefore using the API is not suitable for queries.
Google API
Google also does not currently support near proximity search. There was a tool, Google API proximity search (GAPS)10 which was based on the SOAP Search API. The SOAP Search API11 is now deprecated, therefore GAPS does not work. The newer api – Ajax Search API does not support the NEAR operator.
References
- Noto N. A., The Effect of Property Tax Policies on Property Values and Rents,
- Cambridge: Mass Lincoln Publishers, 1981, pp. 98-100.
- Kayaga, L. Tax Policy Challenges Facing Developing Countries: A case study of Uganda” M.Sc. thesis, Queens University Kingston, Ontario, Canada, 2007, pp 10-12.
- Manly, T.S., Thomas, D.W and Ristema, C.M, Attracting nonfilers through amnestyprogrames. Journal of American Taxation Association, vol. 27 no. s-1, 2005, pp. 75-95.
- Kun, C., Melih, K., Sangjae, L and Gyoo, G., User evaluation of tax filing websites. Journal of online information Review, vol. 32, no. 6, 2008, pp. 842-859
