WEAKLY SUPERVISED DEEP LEARNING FOR CUSTOMER REVIEW SENTIMENT ANALYSIS
Abstract : In recent times the concept of smart cities have gained grate popularity. The ever increasing population has led to chaotic city traffic. As a result of the process of searching a parking lot becomes tedious. It is time consuming task leading to discomfort. The fuel consumption is on an increasing side due to such scenarios. The increase in vehicular traffic creates a negative impact on the environment. In the wake of smart city times these issues lead us to the need of a ‘smart’ solution. In order to resolve these issues and satisfy the increasing demand for the parking areas, parking management organizations are trying to implement better and technologically advanced solutions. A smart car parking application will enable real time parking availability monitoring and reservation thereby providing better services to the end users as well as reduce the workload of the parking administrator. A mobile application is also provided that allows an end user to check the availability of parking space and book a parking slot accordingly. The project also describes a high –level view of the system architecture.
• Lexicon-based methods typically take the tack of first constructing a sentiment lexicon of opinion words e.g. “wonderful”, “disgusting”.
• Furthermore, lexicon-based methods cannot well handle implicit opinions.
• Objective statements such as “I bought the mattress a week ago, and a valley appeared today”. As pointed out in this is also an important form of opinions.
• Factual information is usually more helpful than subjective feelings.
• Lexicon-based methods can only deal with implicit opinions in an ad-hoc way.
• Feature engineering also costs a lot of human efforts, and a feature set suitable for one domain may not generate good performance for other domains.
• This kind of algorithm needs complex lexicon construction and rule design.
• The existing systems cannot well handle objective statements; it only handles single word based sentiment analysis.
• A novel deep learning framework for review sentence sentiment classification is proposed.
• The framework treats review ratings as weak labels to train deep neural networks.
• For example, with 5-stars scale ,we can deem ratings above 3-stars as positive and below 3-star as negative weak labels.
• The framework generally consists of two steps. In the first step, rather than predicting sentiment labels directly, we try to learn a Learning space.
• The Proposed work leverages the vast amount of weakly labeled review sentences for sentiment analysis.
• It is much more effective than the previously developed works.
• The proposed work finds the sentiment not only based on the rating that user gives but also taking into consideration of reviews that they are post.
• In fact it mainly takes an account of review, even though user gave ratings.
|