How a population feels about a certain topic is not always an easy thing to measure. While surveys and polls have traditionally been the methods of choice for gauging public sentiment, they face severe difficulties in achieving their aims. Door-to-door approaches are expensive and the response rate to phone contact has been steadily decreasing for over a decade . The speed with which information can spread through the internet has resulted in an interconnected public capable of rapidly changing viewpoints. Furthermore, multiple choice surveys might miss important information due to their constrained format and allowing responders to answer questions in their own words reduces the speed questionnaires can be traditionally interpreted.
In order to understand how the public perceives a certain brand, product or topic, non-traditional public sentiment techniques have become essential. Natural Language Processing (NLP) and sentiment analysis techniques are already being used to:
Monitor how a company’s brand or product is being perceived by the public, and identify words and phrases associated with that sentiment.
Understand how a competitor’s brand or product is being perceived by the public, and what differences are being expressed between your brand and theirs.
Forecast stock market trends, with models that can update in real time to reflect the public mood
- Gain a deeper level of insight into the political landscape, where polls are becoming increasingly less accurate.
Here at Peak, we understand the importance of public sentiment insight, and through our platform, we apply state-of-the-art NLP techniques to analyse and interpret large bodies of text. From the real-time insight offered by social media to the bulk analysis of historical questionnaires, documents, news articles, or blog posts, the Peak platform is capable of searching both your databases and the internet for the topics that are important to your business, and analysing how the public have responded.
HOW WE DELIVER RESULTS
The Peak platform applies our proprietary pattern extraction algorithms to preprocess a text-based data set, ranging in size from a single tweet to a full, multiple paged document. We then apply an ensemble of state-of-the-art sentiment analysis algorithms, ranging from bag-of-words and n-gram style scorers to high-level Machine Learning techniques, trained on the most relevant corpuses. Finally, the results are weighted and combined to produce our Peak Sentiment Score.
For historical datasets, Peak delivers detailed reports discussing both the raw analytical results and their interpretation into business insight. For real-time sentiment analysis, Peak delivers live results through our Business Intelligence platform. Of course, we also keep a close eye on how public sentiment changes over time or in response to key events and keep you up to date with how society is talking about what matters to you.
 Assessing the Representativeness of Public Opinion Surveys, Pew Research Centre (2012)