Automated Sentiment Reports
SM2 automatically determines the sentiment of a conversation, so you know exactly how people are feel about your brand. Using natural-language processing and Bayesian analysis, SM2 discovers the sentiments around each discussion and aggregates these to provide an overview. It compares the words in each search result to a lexicon dictionary and determines whether positive or negative sentiment should be applied to the words.
The reason for each determination can be found under the expanded options 'View Results', 'Full Details;, and 'Analysis Tools' for each result. The Brand Reference Sentiment is intended to provide a high level overview of the conversation.
By clicking on the hand icon, you can easily change the assigned positive or negative sentiment or click on 'ID number' to see why the sentiment was assigned. Clicking the radio button will display the entire thread of positive or negative mentions.
By clicking on the hand icon, you can easily change the assigned positive or negative sentiment or click on 'ID number' to see why the sentiment was assigned. Clicking the radio button will display the entire thread of positive or negative mentions.
Customizable dictionary increases Accuracy
Many find automated sentiment to be approximately 70 percent accurate. So SM2 offers you the opportunity to customize the dictionary! Terminology varies widely depending on the industry. In customizing the dictionary, every factor can be taken into account including colloquiums and specific words of interest/concern. Customizing the dictionary in SM2 is a simple as reviewing a sample of data and adding/deleting words in the dictionary!
Tone Analysis Reports
Sentiment is determined by specific words in relation to the brand. Tone is an overview of the sentiment of an entire conversation. It is more representative of natural language usage.
Tone is displayed in SM2 using a Likert scale that ranges from one to five. A three represents a neutral tone, one is the most negative and five is the most positive.
SM2 determines Tone by considering the amount of positive and negative sentiment and the length of the post.
For example: If a conversation has 4 negative words and 2 positive words then the net sentiment would be 2 negatives. For a 140 character Tweet this would be considered very negative on the Likert scale. But having 2 negatives in a 1,000 word blog post would be considered neutral. The weighting also takes into account the average length of the conversations in the data set.
While sentiment is a more granular evaluation of a conversation, tone is an overview of the sentiment of an entire post, giving an indication of natural language usage. For example, a post can have a count of three positive occurrences and two negative. The post would have a raw positive score of three and two raw negative.
The tone of each result is displayed using a Likert scale with a range of one through five. A three represents a neutral tone, one is the most negative and five is the most positive. To convert the raw scores to the Likert scale, the data needs to be converted.
In the human language, we have a certain amount of positive and negative sentiment so an overview of the results needs to be analyzed to determine the average positive and negative scores. The net positive score is determined by taking the average positive score and subtracting the raw positive score. Likewise the net negative score is determined by subtracting the raw negative score from the average negative score.
To determine the net overall tone, the net negative score is subtracted from the net positive score.
Net Positive Score - Net Negative Score = Net Overall Tone
The last step is to map the net overall tone to the Likert Scale. The weight given is relative to the length of the post. For example, one word in Twitter is more powerful than one word in a blog post that consists of 1,000 words; SM2 weights the value on the length of the result. That determines the standard deviation and a one through five is assigned depending on how negative or positive the overall tone is compared to what's determined average for the conversation.
In short, Tone is the average Sentiment. It's dependent on the difference between the positive sentiment and negative sentiment and depends on the length of the conversation.
Content Emotions Reports
The Emotions in Content report shows the 16 categories into which the SM2 lexicon dictionary is subdivided. SM2 compares each search result against the lexicon dictionary and a number is determined for each subcategory of the lexicon dictionary. This information can be helpful in gauging the emotion in response to advertising.
For example, imagine you've recently launched a video campaign via an advertising agency. All sentiment is good, but when you look at Content Emotions, a large amount of the feedback is death related. It serves as an early indicator for brand perception. Another use would be for pre-launch of products: you can gauge what customers think about them by analyzing the conversations around the prelaunch information.