The communication and collaboration tools that we have at our disposal help us to be more productive in the workplace and lead richer social lives. Despite the clear benefits, however, it is worth remembering that our increasing reliance on these tools is not without cost. Many of us now experience information overload: too much data flowing in from too many sources at every hour of every day. And the newfound obsession with all things real-time suggests that this is not likely to become less of an issue going forward.
Fortunately, a number of startups are building tools to help us cut through the data stream to get straight to the information that matters. For example, on the business applications side, Xobni has created an Outlook plugin which helps us more effectively navigate and organize our emails and contacts. And on the consumer side, Israel’s own My6Sense (see our previous coverage here) analyses usage patterns to build a personal ‘digital intuition’ that recommends highly relevant RSS feeds.
A more recent entrant is MoodBase, a Hod Hasharon-based startup which helps users discover more relevant content based not on user history or the social graph, but on sentiment. The company, which was founded in October 2008 by a group of high-tech industry veterans, is best described as offering a news aggregation service which classifies content according to the emotions that it elicits.
MoodBase’s core Natural Language Processing (NLP) technology, for which the company has lodged a provisional patent application, determines the ‘emotional consensus’ of content based on reader/viewer reactions expressed in the comment sections of news sites and social networks. It then classifies content items into relevant categories such as Funny, Weird and Romantic, and summarizes stories within these categories into single words such as Cute, Handsome and Wrong.
President Obama’s Nobel Peace Prize award provided a timely example of how MoodBase can offer users recommended stories according to the emotions elicited. As a company blog post notes, several news sources discussing the issue were classified by MoodBase into three different mood categories: Weird, Funny and Right. Evidently, journalists discussing an identical topic can elicit very different emotions in their readers.
The idea of analyzing the sentiment contained in unstructured data is not new. Marketing agencies have for some time now used software like Lexalytics to detect (with varying degrees of accuracy and detail) the emotions expressed in articles mentioning a particular brand or product. And on the consumer side, there is no shortage of activity in this space either, with applications like Tweetfeel and Twendz analyzing the sentiment of tweets across different topics in – you guessed it – real-time.
However, after exploring MoodBase for only few minutes it becomes clear that this service has a more rigorous technological foundation than most other sentiment analysis tools – MoodBase goes far beyond the relatively simple positive/negative classification problem. Eran Schaffer, CEO of MoodBase, suggests that only very few comments are required to determine a suitable category, and points out that the algorithms have been designed to disregard spam comments (which could no doubt have a distortive impact on the perceived sentiment).
Schaffer (formerly at Vizrt) and his co-founders – Oren Glickman (former head of Research at Shopping.com) and Roee Fridman (formerly at IDF’s Unit 8200) – envision numerous applications for this technology going forward. The focus is on becoming a major destination site, but not necessarily one that is limited to news aggregation. This technology has the potential to be applied in other areas like product, restaurant and travel recommendations.
The company acknowledges that the concept of navigating content according to mood is polarizing – some people ‘get it’ while others simply do not. For this reason, MoodBase is focused on addressing the segment of the English-speaking population that thinks primarily in terms of sentiment. For example, though technically-inclined readers might find this hard to believe, some consumers go shopping looking for a ‘fun’ camera, not one with a certain number of megapixels or one belonging to a particular brand.
While a key challenge for MoodBase is to achieve greater distribution – a challenge common to all consumer web companies – at this early stage the company is already offering an API to third-party developers. The API provides developers with access to the core NLP technology and has the potential to increase MoodBase’s reach. It might also shed light on market opportunities that the company has not considered at this point. The company is also in discussions with potential partners (large destination sites) that could benefit by providing sentiment analysis functionality to their users, and there is scope to offer this technology as a white label solution.
The MoodBase concept raises some interesting philosophical questions: does the average opinion matter, and if it does not, how should one go about determining whether an individual’s voice is credible or not? These are questions that the company will no doubt have to deal with going forward if it is to continue to advance the state of the art in sentiment analysis. Nonetheless, by going beyond the dichotomy of positive/negative, which has largely defined consumer sentiment analysis tools to date, the company has the potential to change the way people think about discovering content. MoodBase is a welcome development in this era of information overload, as web users can now cut through the data deluge according to emotion. But remember: this is just one opinion.
