Amazon’s customer reviews are an indispensable feature on the online mega-store, allowing shoppers to get a quick read on products without having to turn to magazine reviews or external sites like epinions. But many products, particularly electronics, have hundreds of reviews. Who has time to read through them all?
Pluribo, a new startup out of New York, has just released a Firefox extension that looks to filter through the noise of Amazon customer reviews. The plugin automatically culls through every review on a given product and generates a concise two sentence summary that highlights the most common positive and negative comments. For the time being automatic product reviews are restricted to electronics, but the site plans to implement support for other products, like books and kitchen appliances, in the near future.
Beyond providing a condensed summary, Pluribo also tries to justify its conclusions. To see the plugin’s explanation, users need only mouse-over one of the underlined terms, which will generate a list of excerpts from the relevant reviews. For example, hovering over “This product scratches easily” will show a list of quotes that contain phrases related to scratches, scuffs, and other similar terms.
So does it work? Sort of, but I wouldn’t go by Pluribo’s word alone. The system seemed to work well enough for ruling out obviously bad choices, but when it comes down to making a final selection, it’s best to turn to the reviews themselves. Pluribo doesn’t always seem to catch the most obvious problems, and the quotes it uses to justify its conclusions don’t always make sense. There’s also no way to tell which products Pluribo has indexed without opening each individual product page.
It’s also unlikely that many people will be willing to install a Firefox extension to get reviews from Amazon alone(the company says more stores are on the roadmap). Even with its problems, Pluribo is a handy tool and one that could gain some traction once it expands its supported product base.
As avid Amazon fans, we think this is a great idea! Love the start ups that actually have something worthy to offer.
Sounds interesting, but you’d think Amazon would develop this feature themselves. I find that the most favorable and most critical reviews that amazon selects are most of what I need to make a purchasing decision.
Jason said…
The plugin automatically culls through every review on a given product and generates a concise two sentence summary that highlights the most common positive and negative comments.
There are some free codes available on the internet already for text-summarization, which does the same thing as the Pluribo’s plugin.
any idea what their revenue model is?
well their patent pending technology can be licensed out that is one revenue
but what is their revenue model for the firefox plugin?
well amazon can never buy this guys out since a independent company is what will make their reviews appear less bias if they buy them out then everything will point towards good
so what will be their exit stratergy maybe they are hoping yahoo will come and buy them out since yahoo is the only one that i know of who has ever bought a firefox plugin
the plugin is good though
Hi, this is Samidh from Pluribo. Thanks for your feedback.
@Falafulu - I assure you we wouldn’t have built Pluribo if there were already good solutions out there. In particular, the problem we are tackling is called “multi-document summarization”. It requires the understanding of multiple sources of information and synthesizing them down into one summary. There’s definitely nothing out there that can do this well. It remains a difficult research problem. It is our aspiration that we can work with the academic research community to continue to try to crack this problem.
Cliff’s Notes for Amazon reviews? Isn’t this what the star ratings are for?
Full text reviews help readers understand how a product was tested, what the reviewer wanted in the product and whether the reviewer’s IQ exceeds room temperature. All Pluribo adds is a language of grunts (”have screen…” “screen good” “me want screen”)
It is a cool stuff, amazon reviews especially for kitchen appliances and customer electronics is one of the best reviews I can find on the web
Perhaps, firefox extension is overkill
It is not that hard to analyze many reviews when you are at home running your firefox a your linux workstation or laptop
If they would the same multi-document summarization, but for iPhone/other mobile platforms so geeks like me who never buy anything > UDS50 without checking Amazon reviews, could easily get the same summary doing shopping in BestBuy or Target, then it would be really good.
I often turn home without buying anything because I am not sure that the merchandize is good. Such mobile application would save my time and money
BTW why shops do not do the same? Small screen to check amazon and other reviews would be a great help for customers
@Paul - The screenshots are actually a little misleading. The text summaries are very fluent. The “grunts” that you noted are actually just snippets from a context menu, not the main part of the content.
As for how this is better than the 5-star system, we actually wrote a blog topic exactly about this. Check out http://blog.pluribo.com/2008/0.....ive-stars/
Pluribo is really a very fine tool. Every Amazon shopper will benefit from it. amazon shoppers should download the plug in and try it out. The installation is so simple!
Pluribo summary is succinct and yet very meaningful. The summary describes the product’s relative goodness with respect to its competition in the same category. The numeric score feature is very useful in making a purchase decision; it eliminates the need for going through many reviews. The “numeric score” feature so useful to an user - has been totally overlooked by the reviewer! As Pluribo’s support for other categories grow, this will be a “must-have” tool for every Amazon Shopper.
Samidh said…
multi-document summarization…
Samidh , I understand multi-document summarization really well, if fact you can use the traditional SVD (singular value decomposition) , NNMF (non-negative matrix factorisation) as shown here ( Multi-document Summarization Based on Cluster Using Non-negative Matrix Factorization) , the HMM (hidden Markov Model) as described here (CLASSY Query-Based Multi-Document Summarization) and more. These codes are available on the internet already (not necessary for text summarisation application) and it won’t be too long before someone adopts them into a free summariser, since the hard part is already been done (the development of those algorithms) and all it needs is a UI to be developed and some text processing codes and that’s it.
I haven’t done development in this area, but I am quite familiar with the algorithms used in text summarisation, since I am already using HMM, SVD, NNMF and so on for development in certain projects.
Also I have come across a project (I can’t recall the title of the paper) where the authors are working/developing their multi-document summarizer as an open source. As I said that it won’t be too long before it appears on the internet.
Samidh , I am not giving your product a negative comment, it is just that you & your team must be vigilant. Always look out for new algorithms from the literatures (or develop new ones for your own use which is proprietary), because you well know that there is something called unique product in software as the amount of computing peer review researches in different topics outstrips their adoptions in technology implementations. I mean that there are more algorithms lying there in the literatures being unused , then startups or companies adopting them for product development, so one has to always watch out either by scouring the literatures continually or develop his/her own proprietary algorithm (which must be better than the currently available ones - either those in commercial use or those unused ones available in the literatures) or otherwise , there is no point in developing inferior proprietary ones.
Anyway, congratulation for taking a step to commercialize your product. Perhaps, you might want to add an extra functionalities into your product , such as text classification or something like that, which makes it unique. Having a summarisation capability only by itself, might look a little dangerous since there might be open source project/s emerging that do exactly the same thing. I am confident that there will be multi-document summarization open source project that will emerge (soon or later).
Just a thought.
Hi Jason Kincaid,
I just submitted a comment that I believe went straight to your spam folder, because it didn’t appear. My comment contains 2 URLs in it.
Cheers.
Hi Jason Kincaid,
I just submitted a comment and I think it went straight to your spamfolder.
Samidh,
I wonder why Pluribo didn’t do the summarization for book first since majority tend to use Amazon for book review. I have installed it but found that it still very much work in progress.
Sorry for the noise - test comment to see how much spam I get at TechCrunch-Comments@tc50.otherinbox.com