Of the many ways for Northern Light SinglePoint™ users to become aware of relevant content in their market and competitive intelligence portal – dashboards, newsletters, and search alerts, to name a few – one that has proven particularly effective in recent years is SinglePoint’s automated Recommended Reading List, one of the powerful machine learning features in the platform.
Recommended Reading List alerts work better than traditional email alerts in part because the platform (vs. the individual user) writes them. The feature is based on latent semantic indexing. For a given topic – for example, a researcher investigating the impact of sustainability in marketing to Millennials – we model the user and the documents she reads. We can determine which documents cover both concepts (sustainability and Millennials) most intensively. The Recommended Reading List can assess the weighting of the topics and prioritize its recommendations accordingly. A normal search alert would include documents that mention both terms without regard for how (or how much) they factor into the substance of the content.
For each user, SinglePoint creates a model of the documents they have downloaded and use it to look for other documents being added to the research collection that contain those topics with similar weightings. In the actual implementation, the model considers not just singular words, but groups of words with synonyms and related terms that represent the topics. (Think of a “topic” as a group of words that like to hang out together and that represent an idea.)
In short, the Recommended Reading List truly understands what the user wants. The whole process of topic creation is unsupervised – the machine figures it out – and it uses 800 topics, not just two. Now that’s putting machine learning to good use, since computers have no trouble thinking in 800 dimensions, but human beings cannot think that way!
The feedback Northern Light gets from clients when their users employ the Recommended Reading List feature is the document recommendations are uncannily accurate in terms of understanding that the users want.
Given all this, it should not come as a surprise that Recommended Reading List alerts are the fastest growing channel for document views in SinglePoint portals. According to metrics Northern Light has aggregated from all our portal users worldwide, Recommended Reading List alerts are up 53% over the past year, and document downloads driven by those alerts are up 61% during the same period. (The growth numbers over the past three years are almost incalculable.)
The fact that Recommended Reading Lists are fully automated is key, since the vast majority ofusers won’t take the time to proactively build a search alert for themselves. Even “expert searches” prepared by others aren’t used nearly as much, nor as effectively, as the Recommended Reading List.
So, to maximize content consumption and provide users the information they want and need for their research and decision-making, put SinglePoint’s Recommended Reading List feature to work for your organization.