September 26, 2017 | Written by: Jalal Mahmud
Categorized: Cognitive Enterprise | Developers | Watson APIs
Share this post:
– We’re excited to announce new changes to the Watson Tone Analyzer service that reflect user feedback
– Three key changes include removing social tones, combining anger and disgust tones and outputting only dominant tones for text
– These changes are active as of 25 September 2017
Learn about Watson Tone Analyzer
We are pleased to announce new updates and feature additions to our IBM Watson Tone Analyzer service.
What you told us
Since our initial release last year, we collected feedback from users. We’ve used the feedback to inform the changes to the Watson Tone Analyzer service. You told us the following:
- Some tones are more frequently used than others
- Tones that are differentiated and actionable are most useful
- Some users were not using the scores from the service appropriately due to confusion. For example, some used tone signals that scored very low or were weak to make important decisions as part of their applications
Changes we made and why we made them
To improve your experience and ability to make the best decision with the help of Watson, we’ve made the following updates:
Change: Social tones were removed from the service output. They consist of the Big Five personality dimensions: extraversion, agreeableness, openness, conscientiousness and emotional range.
Rationale: These tones were not being used. Removing them from the service simplifies and amplifies the experience.
Change: We combined anger and disgust emotional tones.
Rationale: We found our users had perception issue while differentiating anger from disgust. We believe this combined tone ‘anger’ would support the same use cases what would have been supported by anger and disgust as separate tones.
Change: We output only the dominant tones for a given text. Dominant tones are those that have scores of at least 0.5.
Rationale: In our earlier API version, we returned scores for all tones. However, we found that users were making incorrect interpretations of the scores and using weak tone signals (e.g., a score of 0.2 for anger) in their applications. That often led undesirable outcome in the applications. In this version, we want to ensure that only strong/dominant tone signals (i.e. score of at least 0.5) are given as output from the service. This eliminates confusions regarding tone scores and incorrect use of the tones. This is also consistent with our customer engagement endpoint of tone analyzer where we only return dominant tones (https://www.ibm.com/blogs/watson/2017/07/ibm-expands-watson-tone-sensitivity-for-customer-service/).
These changes will be available in our posted API beginning September 25, 2017. We are also excited to announce support of French in the tone API. You can send French text to Tone API for analysis by setting the content-language to fr.
For more details about Tone Analyzer Service, the science behind it, how to use the APIs, and example applications explore the documentation for Tone Analyzer.
Explore our Watson Tone Analyzer documentation and new features.