Technology

narratif is not the same as "social listening software"

(and why this is a good thing)

Almost all social listening systems offer very little assistance to the user when they need to specify and maintain the definitions of what topics to monitor, how these topics are worded, and how they change.

This often comes as a surprise to new users. One would think that social listening software would advise you on what current or breaking topics would be relevant to you, and on what keywords and tags are currently used to express concepts relevant to you.

This is mostly not the case.

Given the fact that Twitter conversations, tags, and vocabulary change like the weather, this poses a big problem to an average user. What mostly happens is that one or two designated social media experts in a work group are dedicated to maintaining an awareness of what is actually happening on the media, by whatever means they can find, and they must learn how to program this as complex boolean searches and classifiers into their expensive social listening software. This is time consuming, costly, and can form a bottleneck for the other account managers in the organisation.

It also means that you can be blind-sided by breaking stories.

Some might call this feeding a dog, and having to bark yourself.

We think that this is not good enough. narratif technology makes it easy for users to specify ideas that are relevant, either using one or two search terms, or some descriptive text (such as a blog, a Wikipedia page, a value statement, etc.). In fact, our Trends story board of the top conversations on Twitter doesn't require you to search at all.

narratif uses real-time machine learning to expand and optimise your query to reach the conversations you would have missed. It then uses the same machine learning to collate the relevant tweets into topics, or stories.

So narratif generally knows Twitter better than you do at any given time, and automatically curates a content feed for you.

 

narratif uses real-time machine learning to expand and optimise your query to reach the conversations you would have missed. It then uses the same machine learning to collate the relevant tweets into topics, or stories.

Sentiment vs Framing

narratif doesn't force the meaning of stories into just two sentiment dimensions for some very good reasons. The main one is that we don't need to - we show you a few much more nuanced story frames that tell you what the story means for you or your client.

narratif's engine automatically generates schemas on the fly at query time. Keywords are weighted based on their relevance.

narratif's engine automatically generates schemas on the fly at query time. Keywords are weighted based on their relevance.

During the Charlie Hebdo events in Paris, we were watching the events and resulting conversations unfold using Discover. We could see stories that were tragic, stories that were angry, hopeful, etc. We wondered how you could squash all the nuanced emotions and frames into just favourable or unfavourable, and if you did, how would you then understand what each story might mean for your client.

Standard sentiment classifiers are known to be rather inaccurate. Maybe they were somewhat useful to reduce the complexity of a list of hundreds or thousands of mixed tweets; but narratif curates the tweet flow into just a few story cards, and you can easily understand the detailed framing for each story without bothering with an overly simplistic 'sentiment' metric.