Next Generation Search – Lemon Finder

Next Generation Search Lemon Finder
By monitoring online user groups, it may be possible to predict product defects and recalls. I’m Jim Metzner and this is the Pulse of the Planet.

Wang: By going through users’ discussions about their problems with cars, we were able to identify early warning signs.

Alan Wang is an associate professor in business information technology at the Pamplin College of Business at Virginia Tech. He and his team have developed an algorithm that monitors online discussion groups to find serious product defects.

Wang: Many times, the user-generated content may contain early warning signs for a product defect. In 2008, an owner of a 2001 Honda Civic posted a note about an airbag problem in an accident, saying that the car’s airbag had exploded, shooting metal pieces all over the place. In early 2015, Honda was fined 70 million dollars by the federal government for under-reporting deaths, injuries and the warranty claims stemming from serious defects like the one just mentioned.

There’s thousands of postings like that, so it’s very likely that we can detect some early warning signs, especially for product defects that may have a catastrophic consequence.
We assume that, if a posting is made for a complaint or a problem, it should be longer than postings that do not talk about a complaint. We think that postings that present negative sentiment are more likely to be a product-defect related posting.

Professor Wang’s team also identified certain key words or phrases that could flag a potential defect.

Wang: In (the) car industry, people may use words like “engine, airbag, door” to describe problems.

In tests with online car forums, the algorithm was able to identify product defects with an accuracy of 70%. I’m Jim Metzner and this is the Pulse of the Planet.

Next Generation Search - Lemon Finder

Is it possible to predict product defects and recalls by monitoring online user groups?
Air Date:03/03/2017
Scientist:
Transcript:

Next Generation Search Lemon Finder
By monitoring online user groups, it may be possible to predict product defects and recalls. I'm Jim Metzner and this is the Pulse of the Planet.

Wang: By going through users' discussions about their problems with cars, we were able to identify early warning signs.

Alan Wang is an associate professor in business information technology at the Pamplin College of Business at Virginia Tech. He and his team have developed an algorithm that monitors online discussion groups to find serious product defects.

Wang: Many times, the user-generated content may contain early warning signs for a product defect. In 2008, an owner of a 2001 Honda Civic posted a note about an airbag problem in an accident, saying that the car's airbag had exploded, shooting metal pieces all over the place. In early 2015, Honda was fined 70 million dollars by the federal government for under-reporting deaths, injuries and the warranty claims stemming from serious defects like the one just mentioned.

There's thousands of postings like that, so it's very likely that we can detect some early warning signs, especially for product defects that may have a catastrophic consequence.
We assume that, if a posting is made for a complaint or a problem, it should be longer than postings that do not talk about a complaint. We think that postings that present negative sentiment are more likely to be a product-defect related posting.

Professor Wang's team also identified certain key words or phrases that could flag a potential defect.

Wang: In (the) car industry, people may use words like "engine, airbag, door" to describe problems.

In tests with online car forums, the algorithm was able to identify product defects with an accuracy of 70%. I'm Jim Metzner and this is the Pulse of the Planet.