Online Reviewing Made Super Easy By Mining User Generated Content

94 views Leave a comment

With a transition from word of mouth to an ‘electronic word of mouth’ selling culture, businesses currently feel a need to possess an expanded arsenal of user feedback (preferably positive) in sequence to symbol their repute and participation in a Web. Though a 1% sequence of Internet is reputed to be dead, a suit of lurkers, i.e., people who observe user-generated calm in a Web though contributing, is still high. A 2010 Pew Internet consult reveals that usually a entertain (24%) of Americans have ever posted comments or reviews online about a things they buy. Thus, businesses strategize innovative inducement programs in a unfortunate try to hoard patron reviews.

“Writing reviews is too tedious” is a many renouned reason given by people who never or frequency write online reviews, followed by “I forgot”. While businesses can take caring of a after by promulgation sign emails to business periodically, they still need a complement to facilitate a examination essay charge for a customers.

Dr. Gautam Das during University of Texas during Arlington and his students, Mahashweta Das and Azade Nazi yield a novel resolution to this problem. They precedence user feedback accessible in a Web for products by past users in sequence to brand a set of suggestive phrases, i.e., tags, that when suggested to a user would assistance her examination a product. The user would fast select from among a set of returned tags to clear her feedback for a product though carrying to spend a lot of time essay a review.

As one of a initial step towards solution, they occupy content mining techniques in sequence to remove suggestive phrases or tags with view labels from user feedback in a form reviews. “It is a lightweight camera with some extraordinary features” is reduced to a tags {lightweight camera, extraordinary features}, where both tags have certain sentiment.

The inventors delineate a problem as a general-constrained optimization problem. A core plea in this pattern is defining a essential properties of a tags to be returned that would offer to examination a product effectively. They cruise aptitude (i.e., how good a outcome set of tags describes a product to a user), coverage (i.e., how good a outcome set of tags covers a opposite aspects of a product), and polarity (i.e., how good view is trustworthy to a outcome set of tags) in sequence to capacitate a user to satisfactorily examination a product.

A user can examination a product in opposite ways. A user can demonstrate her extended opinion about a opposite aspects of a product which, in turn, can possibly be certain or negative. Again, a user can demonstrate both certain and disastrous opinion for a same product feature. For example, a user might write a examination for a camera as “The design peculiarity of this camera is good and so is a sharpness and tone correctness of a pictures, though a battery life is short.”, while another user of a same camera might write “Though a additional shade with touchscreen and gesture-control facilities saps battery life, it’s ideal for fashion-conscious snap shooters.”. The initial examination contains certain feedback for a camera’s picture peculiarity and disastrous feedback for a camera’s battery life. The second examination contains both certain and disastrous feedback for a camera’s modernized facilities such as dual-screen, touchscreen and gesture-control. The ubiquitous problem plan considers dual opposite definitions of coverage of product facilities by tab in sequence to capacitate a opposite real-world scenarios. They rise unsentimental algorithms with fanciful end to solve a problem well and effectively.

The group conducts an Amazon Mechanical Turk user investigate in sequence to countenance if users cite and advantage from tags returned by a due resolution for reviewing products. They cruise 12,600 reviews accessible in Walmart by 11,500 users for 140 digital cameras as training information and beget tags for 6 new cameras.

The user investigate was conducted in dual phases. In a initial partial of a study, an strenuous 71% of a users reviewed a 6 cameras selecting tags returned by a due resolution instead of essay a examination from blemish thereby validating that they find a returned tags suggestive and adequate. The second partial of a investigate dictated to establish if a feedback left by a users compare a tags returned by a solution. Domain experts complicated a formula and suggested that 77% of a users submitted feedback that matches tags returned by a due system.

The work was supposed for announcement during a reward general peer-reviewed investigate conferences SIGMOD/PODS by Association for Computing Machinery (ACM) and during VLDB, both of that ranks among a tip Computer Science conferences of all times. The proof of a work will be presented during a 42nd International Conference on Very Large Databases in New Delhi, India on 6th September, 2016.

Sources:

  1. Azade Nazi, Mahashweta Das, Gautam Das. The TagAdvisor: Luring a Lurkers to Review Web Items. 34th ACM SIGMOD International Conference on Data Management.
  2. Rajeshkumar Kannapalli, Azade Nazi, Gautam Das and Mahashweta Das. ADWIRE: Addon for Web Item Reviewing System. 42nd International Conference on Very Large Data Bases.
  3. Web Item Reviewing Made Easy by Leveraging Available User Feedback. arXiv preprint: 1602.06454. 2015