The limitations of social listening, particularly in languages that are not English, are plain to see across projects and insights when you’re trying to use commercially available software for research.
One of the primary frustrations is the accuracy of the machine-learning results from those efforts, particularly in Arabic – they don’t work.
It’s why we built Sila.
So why invest in a platform like Sila? Because understanding nuances in a language are essential to unlock consumer opinion, which allows insight. It’s all part of our audience-first approach.
Specifically, one nuance, in particular, is fascinating – Arabic-language dialect understanding.
Understanding dialects in Arabic help us get insight in exciting ways – and that’s beyond the conversation that you think you might want to hear.
In a demographically fragmented Gulf, we’re able to understand what people’s origins say about their opinions – for better or worse. Why? It allows us to build a demographic profile and understand certain societal biases.
Afghanistan dialect understanding
We’ve used recent events in Afghanistan to see how, outside of any consumer trends, we can begin to understand the consumer reactions and whether there’s a link to perception between dialects.
In undertaking this analysis, we used Sila’s dialect analysis model. We used Twitter as the base (this seemed right for a topical subject like the Afghan conflict) to understand the conversation and who was driving it.
To start, the short and short of it is that people aren’t interested for very long in today’s news cycles. Although, at the time many claimed that we shouldn’t forget, it seems we did very quickly:
So what about the dialects spoken? As we’d expect, given the conversation was driven by Saudi Arabia, the Saudi dialects were number one in the analysis. Interestingly, when we slice that data individually, Kuwait and UAE had more standard Arabic on the topic.
We see, though, representation across home dialects in Kuwait and Saudi Arabia, but in the UAE, as the country itself, is diverse.
When we apply the dialect emotion detector, we see that the public emotion displayed by news outlets might not have been giving us the complete picture. People were more surprised than anything about the structural failures (or victories, depending on your point of view).
Aside from the almost complete picture, this paints of people’s predictions were the most prevalent feeling towards the situation. In a matter of a month, the conversation almost tailed off.
Although they weren’t angry or sad, the sentiment is very much net-negative. From universal negativity, the positivity then started to emerge in conversations as we started to share the stories of hope that had come from the crisis.
We also can’t discount those celebrating the defeat of the USA. This was a theme that was prevalent in the discussion, to a minor extent, that can also be interpreted in another way and richness to analyse news events.
What can we learn from dialect study?
Anything topical becomes easily dissected when a language is accessible. When it isn’t, or it has significant nuance like dialect or slang, then specialised analysis gives us insight almost immediately.
Aside from a sector that takes direct and literal benefit from Sila, like media, there are learnings applicable in many consumer categories.
By analysing dialect we can:
- More accurately understand the demographics of an online audience in any one geography – particularly in expat-heavy markets;
- Applying that learning to understand the motivations of an audience. What are the cultural differences to a product or category apply? Or has the ‘host’ country homogenised those views?
- Understanding the dialect helps us understand the level of engagement that communities have within, and external to, countries.
- Correlating dialect to customer experience – in many sectors, particularly travel, we see a direct correlation between dialect and customer satisfaction.
What about sentiment and emotion?
As we then overlay that with sentiment and emotion, we can build a stronger picture that can help frame:
- Focus areas for new product development;
- Correlation between emotion and dialect creates an overlay with demographics to better understand how people really feel about a topic.
- Content strategies to appeal to the right audience at the right time, with a language that they are familiar with and understand.
- Improve performance.
This technology will only improve as more data is analysed, and from world events to customer experiences, will help us better understand our end consumer or customer in ways that only traditional research could’ve reached previously.
In real-time we can analyse events, complex societal issues and topics to build a picture of a consumer, in the Arabic-speaking world, that helps drive business forward.
