Summarizing user reviews for computers and electronics

Users have told us that product reviews are one of the most important research tools they use when shopping for products.  The Internet has lots of reviews, but they're usually spread across many sites, which makes finding and reading all of them time consuming.

Our team saw this as a great opportunity to help.  As alluded to in a previous post, we recently launched a new feature that finds, analyzes and summarizes reviews from across the Web in a single place.  Our search results for popular computers and electronics now include graphical summaries of reviews from people across the Web.  The summaries highlight the most discussed aspects (e.g. size, battery life) of each product and the percentage of positive and negative opinions expressed.

For example, try the query hp printers and click a printer such as the LaserJet 2600n.

You can see a graphical summary of user reviews in the form of green bars on the left and click on the different aspects discussed such as "Price", "Print Quality", or "Speed".  Here's a screenshot of positive comments about the printer's speed.  From users' comments I can quickly see that a majority of users on many different review sites think that the printer is fast.

However, when I read negative comments I learn that many users complain about the printer speed, especially for its startup speed and black and white prints.  It may have been difficult to read every review for the product to find these comments, but the summary automatically lifts out comments about this important aspect of the printer's speed.

See our previous blog post for some other computers and electronics searches to try.  Click results with star ratings to see summarizations for products with many reviews.

As you can imagine, summarizing reviews can be tricky.  There are some cases where it's easier to classify comments as positive or negative and identify the aspect discussed.  For example, the following comments are concise and clearly mention aspects of the product such as "easy to use" or "support."

• "Very easy to use"
• "Cons: HP will not be supporting this per their website."

However, consider the following cases.  These comments are trickier and there are an infinite combination of these and other types of issues in natural language.

• "The best budget laser color created by hp" - this comment was correctly associated with other positive opinions about "Price" even though it doesn't contain the word "price." This requires understanding the fact that the word "budget" is synonymous with a good price.
• "I love it, you just have to be careful with it because it's somewhat fragile" - this comment is part positive and part negative requiring careful handling.
• "I ordered the product online and waited anxiously" - this comment says nothing about the quality of the product and needs to be ignored during summarization.
• "The manual duplex is real easy to use" - this comment mentions "manual," which is a word with multiple meanings. One meaning of the word is product documentation, which is a commonly discussed aspect of products. However, this comment is talking about a printing capability called "manual duplex". This type of word ambiguity can cause the comment to be misclassified as discussing product documentation.

Because of these challenges, it took a lot of effort and help from Microsoft researchers to develop this feature for popular computers and electronics.  Also, the summarization isn't always perfect and occasionally misclassifies comments so we're working hard to continually improve the accuracy of this technology.

We hope that this feature saves you time and helps you find the most interesting user opinions about a product without requiring multiple searches and scanning through pages of reviews.

Please let us know what you think!