Frequently asked questions
Are Pluribo summaries really written by computers rather than people?
Yes, we're absolutely certain. We developed the summarization algorithm ourselves.
How is that even possible? How does Pluribo work?
Essentially Pluribo does exactly what you'd do if you had the time: it reads every word, learns what people liked and disliked, and then figures out the key lessons. To do this automatically, Pluribo has developed novel algorithms for opinion aggregation, sentiment analysis, and natural language generation. These techniques are explained in further detail on our technology page.
If Pluribo summaries are written automatically then how do you know they are correct?
We frequently calibrate our algorithm's output to known product data in order to insure a high level of accuracy. Also, since our summaries take into account a large number of reviews, even if the algorithm accidentally misinterprets a few of them, the overall summaries are still very robust. Look at our methodology page to learn about all our other efforts to insure accuracy.
How frequently do you update the summaries?
Whenever we encounter a new review for a product, the summary is immediately updated. In general, we search Amazon every 7 days for new product reviews.
I tried to get a summary for a certain product but none was available. What happened?
If there is no summary, it's either because we do not currently cover that product category or because there were too few reviews available for an accurate summary.
Which Amazon product categories do you currently cover?
At present, we cover most Amazon electronics categories. We try to add new product categories every week, but our priority is to maintain summary quality as we grow.
The summary text for a certain product sounds negative, but the extracted quotes are all positive. What's going on?
Ah, an astute observation. Pluribo evaluates a product by comparing it to similar products. This can lead to a situation where the product might have somewhat positive reviews, but competing products have much more enthusiastic reviews-- so the product is actually kind of bad compared to its peers. The ability to do such peer comparison is critical to maintaining a high level of accuracy for Pluribo. It is also an example of something that's not easy to do on your own!
The summary text for a certain product sounds positive, but the extracted quotes are all negative. What's going on?
See the answer to the previous question. The same logic applies, except that the product has above average reviews.
Why is it that sometimes the extracted quotes contradict each other?
Well, people often disagree with each other in their reviews. Some people might rave about a feature, while others might be highly critical. We try hard to find you the best consensus, but the reviews aren't always unanimous. To give you a sense of the variability in the reviews, we try to find some representative quotes.
Why do the summaries sometimes not mention features that are of obvious importance (i.e., image quality for a camera)?
We only report a feature if it somehow distinguishes an item from alternative products (i.e., if it is really good or really bad). For example, there is little point in mentioning that a camera has "average" image quality because the fact is that most cameras have average image quality. It's best just to mention the things that make a product special.
I found a mistake. What now?
That's great! Please tell us so that we can make the system even better. Thanks for your help.
Is Pluribo available as an Internet Explorer plugin?
Not yet, but we're working on it. In the interim, consider switching to Firefox. It's better anyway.
Can Pluribo summarize reviews on sites beyond Amazon (e.g., TripAdvisor, Yelp, etc.)?
While we currently cover only Amazon, Pluribo's summary engine is flexible enough to summarize reviews on nearly any topic or from any source. We hope to expand our service to other sites in the near future.
I'm a developer. Can I integrate Pluribo's summaries into my service?
We are working with select beta partners to test out an API that allows developers to mashup Pluribo's summaries in any way they wish. Please send us a note at api at pluribo dot com to inquire about joining the beta program.
Still have questions? Please email us at info@pluribo.com.