The following is a critique of Sharon Givon PhD thesis titled “Predicting and Using Social Tags Improve the Accuracy and Transparency of the Recommender Systems.” Sharon Thesis reviews recommendation systems by exposing the weakness of Collaborative Filtering (CF) methods while illustrating the benefit of applying a textual content approach in the production of different products.
Summary of the Thesis
The thesis analyses and implements the capabilities and techniques of current recommender systems by analyzing the problems associated with CF materials and then outlining the frameworks that can be implemented to convert into the use of textual content. For instance, Givon identifies that such devices have ramp-up problems especially due to incompatibility of metadata as well as the lack of transparency resulting from ambiguity in the recommendations produced by the system as the main barriers. The Thesis further notes that social tags are ideal additives to correcting such anomalies. It illustrates the difference between CF based explanations and tag based ones. By using products that differ in size and type, she tackles the problems listed above from a skewed perspective of improving the user experience. It is evident that Sharon considers the difference in ratings as having a profound effect on the recommender systems since such inter-user comparisons differ from others.
Purpose and Objective of the Study
The study is premised on the notion that recommender systems differ in their ratings and therefore need to be fitted with textual content as a way of improving the predictions channeled to the target user. Sharon’s objective is to familiarize the reader with the difference that arises from using social tags as opposed to sticking with the common type of explanations.
The analysis provided by the author aptly captures the working mechanisms of most systems. A refusal to form recommendations because of the lack of adequate data associated with some items as highlighted by Givon traces the problems linked to CF systems. Under such circumstances, it is possible for automatic predictions to be made by internalizing the available textual content. It is imperative that the different tagging systems perform their functions hence the need to structure the technological applications in line with their perceived roles. The text recognizes the importance of creating applications that can handle heavy traffic while also having mechanisms of detecting bugs. The author implores that textbooks form a perfect example of using this approach because the reliability of the above technique is quite high. In making a chronological design that links textual data to the enhancement of transparency of recommender systems, it becomes evident that the social tags used help to increase efficiency in the recommender system. Givon’s analysis of the role played by social tags illustrates that the effects of their sizes and varying types needs consideration as well. For instance, in the annotation of book texts, the algorithms that arise show different levels of scalability.
Sharon’s Literature is detailed describing the types of recommender systems, the procedure of collecting and generating social tags, using these tags to improve recommendations as well as offering explanations about understanding the various recommendations. In making the claims that transparency and accuracy are better attained through the elimination of CF systems, Givon offers an in-depth assessment of the text classifications thereby enhancing the quality of the topic. However, it is missing simplified graphical representations that would have enabled a better comprehension of the text. It lacks an objective discussion about the CF systems due to her opposing opinion to its viability. Some of the sources used such as social media sites are not authoritative because such findings are not scientific. It is a phenomenon that is fluid and bound to change according to the prevailing trends hence the recommendations of such a system may need alterations frequently. Sharon should have provided industrial procedural manual excerpts to accompany the findings in order to reinforce the ideas presented. Addition of such details as directions governing the setting up of social tags would be helpful in identifying the usefulness of abandoning CF systems without social tags. She should also use less jargon as a means of simplifying the content thereby making it easily readable. Some of the expressions are identifiable with technical experts to the detriment of the larger demographic, especially the youth who form a large portion of the consumer and user base.
Research Approach and Methodology
Sharon contributes to literature by introducing the term “folksonomy” which illustrates the classification system whose derivation is traceable to tagging methods. In fact, the process of associating free texts to specific items bolsters the claim that the normal recommender system is less effective without the social tags. It is therefore a useful additive by Givon as it increases the publication’s appeal. The research approach employed involves conducting an online survey whereby users are required to provide feedback depending on their experiences with varying labels. In the design of the tags, the focus of her work is to proof the viability of using sizeable social tags for the different consumers regardless of their past online usage. The methodology involves the application of Feature Space in which popular parts of speech and named entities are evaluated using information retrieval platforms. Having weighted the words alongside their respective tags, the strength of their viability is therefore tested using the above technique. Her contribution is thus based on improving internet-based research for recommender systems.
Model Implementation and Data Analysis
Sharon Givon’s study is based on the above approach. It involves the production of various algorithms in line with the focus on social tags. The use of a training set of 96 books and a testing set of 50 books is done in conjunction with the Boos Texter principle that checks the number and rate of iterations. The parameters applicable in these tests thus form the basis of the online survey in determining the best user experience. Upon doing so, it emerges that the use of nouns is better that proper nouns. The results are also higher when the tags used are popular with the masses due to their ease of remembrance. Selection of tags is also provided as a major pointer of the expected traffic hence the recommender system is best served by social tags with greater appeal. The author asserts that the size of the icons also matter and this is reflected in the huge number of users who easily use legible ones. Book annotation is another aspect of the study whose data generation is deemed difficult but necessary. The Mean Average Precision (MAP) method used in the research is convenient for this type of work because it has a minimal margin of error. For instance, Sharon states that tag sets as large as 10,000 tags are automatically assigned to book texts successfully thereby proving the viability of this technique. Givon use the Relevance Model which is better than the BOOS Texter due to the large-scale annotation needed for the numerous books. The rate of recommendation generation is higher and the use of social tags is preferable to human-assigned ones since the former is automatic.
Recommender systems that use the CF principle place greater demands on the components as well as users due to the laxity in generating recommendations. In tackling this problem, Sharon Givon conducts research into the possibility of using social tags as motivators thereby offering insight into the mechanisms of these products. The use of jargon does not deter the reader from perusing the text since the definitive tone of its contents is appealing. By using an online sampling technique in conjunction with the Relevance Model, Givon is able to illustrate the ease of such transactions upon the inclusion of social tags for a better user experience. It is noteworthy that she realizes the importance of enhancing the ratings of such products hence improving recommendations enriches content generation. Whereas the provision of useful information to people is a key objective, making such content easy to access to the same users is the main goal of this study and the author offers assertions to reinforce this notion. The thesis is thus an insightful account of accuracy and transparency improvement within recommender systems as it exhausts their challenges while offering solutions that are practical and functional. Sharon’s text is authoritative as well due to the scientific proofs documented within.
Top-quality papers guaranteed
100% original papers
We sell only unique pieces of writing completed according to your demands.
We use security encryption to keep your personal data protected.
We can give your money back if something goes wrong with your order.
Enjoy the free features we offer to everyone
Get a free title page formatted according to the specifics of your particular style.
Request us to use APA, MLA, Harvard, Chicago, or any other style for your essay.
Don’t pay extra for a list of references that perfectly fits your academic needs.
24/7 support assistance
Ask us a question anytime you need to—we don’t charge extra for supporting you!
Calculate how much your essay costs
What we are popular for
- English 101
- Business Studies