Automated sentiment analysis uses advanced processes such as natural language processing, text analysis and computational linguistics to identify the predominant sentiment in mentions.
Based on the expressions detected in the text, Bunker Listening evaluates whether words with positive or negative connotations predominate and assigns the corresponding sentiment. If there is no clear inclination towards positive or negative, the system categorises the mention as neutral.
Example of every type of sentiment:
Green for positive
Grey for neutral
Red for negative
It is important to keep in mind that even if a mention is classified as negative due to the use of certain words, this perception may not match your personal stance on the issue. Automatic assignment is based on linguistic patterns and does not always reflect subjective or contextual interpretations.
Factors that affect automatic sentiment’s accuracy
Although automated sentiment analysis is highly accurate (around 70-80%), there are factors that can affect its accuracy:
Lack of context
The analysis of isolated sentences can lead to incorrect interpretations. For example:
Isolated text: ‘Nothing at all’ could appear negative.
Extended context: If this response follows the question ‘What didn't you like about our customer service?’, it takes on a positive tone, indicating satisfaction.
Sarcasm & irony
Sarcasm is a challenge for machines, as its detection depends on factors such as intonation or context, beyond the words used. For example:
‘Oh yeah, so polite’ or “It’s totally fine our reservations were canceled.” Although the words seem positive, the real tone can be negative.
Oppositional conjunctions
When a mention contains mixed sentiments, it can be difficult to determine the predominant one:
‘Your customer service is terrible, but your tool is great.’ In this case, the automatic analysis may not accurately reflect the balance between the two sentiments.
Manual adjustment as a solution
When automatic accuracy doesn't meet your needs, you can manually adjust the sentiment of your mentions in Bunker Listening to ensure they more accurately reflect the context and original intent.
This approach allows you to combine technological capability with human judgement for a more accurate and personalised interpretation.