November 8, 2016 | Written by: Scott Stockwell
Share this post:
Francesco D’Orazio, VP of Product, Pulsar took the stage. Pulsar is a next generation audience insight company. They’re looking both at people who are talking – and intrestingly, also those who aren’t talking. Pulsar started out working with the familiar favourites Facebook and Twitter. But they now go beyond that and look at Mintel, TGI, clickstream and some of the search and advertising data. Many of the top companies are working with Pulsar and their modular approach – but what do they do?
Pulsar help companies with:
- Audience understanding
- Trend analysis
- Planning insights
- Marketing measurement
- Category mapping
- Brand health
Pulsar works in these ares using audience intelligence – understanding the moments that are influential to audiences. And also actionable insights – to help brands make changes that work for their audiences. The company uses natural language processing (NLP) to understand the conversations that audiences are having.
Francesco D’Orazio explains the need for a modular approach to AI
Understanding complex conversations
But the conversations that we have can be challenging to understand. We’re multi-dimensional and often what we say in meaning is not what we say literally. We’re also creating a lot of multi-media, videos, pictures, emojis, and making sense of this kind of information needs cognitive capabilities. To see what’s in a video, you need a way to analyse it to understand what’s in the picture. Context is important too – what we say, and how we talk to friends, is very different to the conversations that we have with co-workers.
Pulsar designs their offerings as a modular hub of AI services. This allows them to add new modules when they’re available. At the moment, Pulsar operates ‘core services’ and ‘modular services’. Core services include concept targeting, sentiment analysis and conversation analysis. The modular services include image classification, text extraction and emotional analysis. This means that users can choose which applications to apply to which data sets – to make the best use of time and resources.
A picture is worth 1,000 words – in a tag cloud
Franceso gave an example that looked at many images of a Tesla and at the number of times that tagged classification of the image were mentioned – a little like a tag cloud of the contents of an image. For Tesla, there were many images of drivers looking out from the steering wheel. A similar approach for Landrover had many images with dogs in the back of the cars. This is really useful for brands to understand how they are viewed and make informed decisions on who they target and how they present themselves to those audiences. This can have a big impact on the effectiveness of advertising campaigns for example.
“Leningrad ghetto god”
The early days of text recognition in images was not great – the phrase “give it up to God” in a photo caption was first analysed as ‘Leningrad ghetto god”. Nowadays, Watson can look at the text in images and know what it says with great accuracy. Burger King have recently used this in a campaign with McDonalds – and being able to see the words of the brands in images has been very important.
Why a modular approach is important
A modular architecture has allowed Pulsar to integrate the best AI around. Services can be quick productized. Many use cases can be supported with the modular flexibility. It also keeps the team on their toes with more and more new AI applications becoming available as these technologies expand. The bottom line – technology is changing, and being able to respond by building services that use the latest capabilities is critical. A modular approach to AI is just one example of what companies should be doing to stay ahead of the competition.