Here is an interview with Dr Stuart Battersby, Chief Technology Officer and Founder of Chatterbox.co – a sentiment analysis engine that utilizes machine learning (yes that all familiar buzz word) to social media conversations (even more prevalent buzz words). However what makes Chatterbox.co stand apart from social media err,chatter is it’s clever use of hard core computational performance of machine learning with insightful and creative business use cases.
We have interviewed sentiment analysis API creators (Viralheat) before here and there, so developers can take note , compare and contrast- especially since Chatterbox proudly promises a 20% lift over industry average.
Ajay-How does Chatterbox Sentiment Analysis for Social Media API work?
Stuart- When we were designing the Social Technology Engine, of which sentiment analysis is one module, we faced a challenge. There were lots of existing sentiment analysis tools on the market that were aimed a structured text such as news articles, web pages and blog posts – however the challenge with social is greater.
There is very little or no sentence structure, and we needed to build a system that could effectively handle the slang and niche language that is at scale and across geographies. Backed by academic research we built and validated a system that uses machine learning and natural language processing to actually learn the language people use in social media. As a simple example the engine learns that a smiley face is a positive ‘word’ without requiring manual intervention. As we use machine learning, this approach is very effective across industries and languages. We support most western language and also offer Mandarin Chinese. We know the approach works; our customers have independently validated our accuracy at 83.4%, much higher than the industry average of 65.2%.
Ajay- You have two broad pricing plans starting from £250.00 / month.Your API pricing is a fremium model starting from free 500 calls daily going up to a premium 500 pounds a month. How do you intend to get Developers enthusiastic about using your APIs as well as creating mashups with it?
Stuart- Our products are well suited for both developers and Enterprises alike. As a company we are experiencing significant traction across the Enterprise, but it is important to us that the technology we deliver is flexible. By offering a freemium model for the API service we are able to let developers test the water and open up their creative side. We believe in the technology and so are happy for others to experiment with it, and it is an approach that is working. The conversation rates from freemium to cost effective paid for plans is high, with applications of the technology ranging from Enterprise systems, through integration with 3rd party dashboards to academic and research purposes.
Ajay- How does Chatterbox intend to differentiate and distinguish itself from some of the other APIs existing in the social media sentiment analytics market ?Is it just accuracy or are there other benchmarks as well?
Stuart- There are two key points here:
1) Yes, accuracy is essential. Systems with poor accuracy result in extensive and costly manual correction. By using machine learning tailored specifically for social intelligence, our system is nearly 20% more accurate than the market average.
2) The depth and flexibility of the engine is very compelling. We can adapt our technology to each vertical we deal with, providing a bespoke service that embraces the Enterprise’s own terminology. On top of this we deliver more fine grained analysis – we provide metrics on excitement, anger and worry – all essential for competitive intelligence. Our roadmap doesn’t just stop here. We have phrase analysis, topic mining, buying intent and audience interest profiling which dig deeper and give feedback on what the content of social messages are and who the audience are at scale. These are already in our main product lines and will be rolled out as APIs soon.
Ajay- Which sector or domain do you believe critically need social media analytics and sentiment analysis the most?
Stuart-Being able to build a technology engine that is modular, flexible and seamlessly integrates at the Enterprise level across sales, service and brand is essential in today’s market. We are the only vendor that is addressing three distinctively different areas within the social intelligence landscape. With this in mind, the solution is built to cover all major sectors and a clear social intelligence strategy should be part of all organisations. Being able to slice and dice our modules means that we can deliver a bespoke service at the push of a button.
Ajay- Describe your anger detection module in your sentiment analysis ?
Stuart- There is a big difference between a negative comment and an angry comment. Addressing angry comments are critical for customer services, protecting brand reputation and understand any frustration with your products. Anger is where the fires start online. As Chatterbox use natural language processing we are able to exploit the fact that consumers use different language when they’re angry to when they’re generally negative. Our algorithms can distinguish between the two in real time, and feed this back into Enterprise communication systems.
Chatterbox uses machine learning to derive insights from social data. Chatterbox’s Social Technology Engine encapsulates IP into a modular system that operates across verticals, markets and languages. Stuart holds a PhD in Human Interaction and has a background in Computer Science.
Anger Management in Social Media and Accurate Sentiment Analytics ? Just another (Chatterbox) API call away!