Sentiment analysis, also known as opinion mining, has a history dating back to the 1960s when psychologist Stuart Oskamp made the first attempt to classify text based on sentiment using a simple counting method.
In the 1980s and 1990s, sentiment analysis gained popularity in market research, then in the early 2000s, machine-learning algorithms were introduced, allowing for more accurate analysis of large datasets. With the emergence of social media in the mid-2000s, sentiment analysis became an important tool for monitoring and understanding public opinion on platforms such as Twitter and Facebook.
In recent years, sentiment analysis has become more widely-used across a variety of sectors, including finance, healthcare and politics. Advances in natural language processing (NLP) and machine learning have led to more accurate sentiment analysis and the ability to analyse more complex types of text, such as social media posts and product reviews. Today, sentiment analysis is a rapidly-growing field and continues to be an important tool for understanding and analysing public opinion.
With the tenant satisfaction measures (TSMs) coming into effect in April 2023, understanding the sentiment of tenants will be more important than ever. While the survey questions themselves have fixed responses, the verbatim comments could be a valuable source of additional data for housing providers on the reasons behind their feedback, enabling data-led strategies on what improvements and changes need to be made.
Scalability – Sentiment analysis can be applied to large volumes of text data, making it useful for processing not only verbatim feedback but also social media feeds and other forms of user-generated content at scale.
Speed and efficiency – Sentiment analysis algorithms can quickly and accurately classify the sentiment of a large volume of text, saving time and effort compared with manual analysis.
Insights – Sentiment analysis can provide valuable insights into customers’ opinions, helping housing providers make data-driven decisions.
Accuracy – Sentiment analysis algorithms are not always accurate and may misclassify sentiment, especially in cases of sarcasm or irony.
Contextual limitations – Sentiment analysis algorithms often struggle to account for the nuances of language and sector-specific context, leading to errors in classification.
Bias – Sentiment analysis algorithms may be biased towards certain sentiments or groups of people, depending on the training data used.
Lack of granularity – Sentiment analysis can only provide a high-level analysis of sentiment and may not capture the nuances of sentiment within a piece of text.
Emotion vs. sentiment
Sentiment and emotion are related concepts but they are not the same thing. Understanding both emotions and sentiments in natural language processing (NLP) is important because they provide different and complementary information.
Sentiment refers to the polarity of a text, whether it is positive, negative or neutral. Sentiment analysis is useful for quickly gaining a general understanding of the attitude expressed in a piece of text.
Emotion, on the other hand, refers to the feelings or states of mind evoked by a text, such as anger, joy, fear or sadness. Emotion analysis is more nuanced than sentiment analysis because it attempts to capture the complex and multi-dimensional nature of human emotions.
Understanding emotions in addition to the sentiment can provide deeper understandings and insights into the impact an experience had or is having on a customer.
The future of sentiment analysis in housing
Multi-lingual sentiment analysis – With such a growing, diverse cultural composition across UK housing providers, multi-lingual sentiment analysis has become more important. Advanced machine-learning models that can accurately analyse sentiment in multiple languages will be in high demand.
Real-time sentiment analysis – The ability to analyse sentiment in real-time, integrated with chatbots, voice assistants and recommendation engines, providing support to both agents and customers.
Emotion detection – Sentiment analysis has moved beyond simply identifying positive or negative sentiment to detecting emotions such as anger, joy or sadness. This will enable businesses to provide more personalised customer experiences and make more informed business decisions.
Caroline Thomas is a senior CX service designer at FourNet.