Industrial applications of semantic analysis and the study of natural language by AI ('Artificial Intelligence'), especially in the field of Voice of the Customer and user support, are associated with sentiment analysis.

The conclusion is often insufficient

Some marketing pitches would have you think that sentiment analysis is so progressive and mature that it is reliable...others would say “Bullshit”. Graphs are colourful, easy to read, beautiful in strategic reports. Of course understanding has progressed but if we limit ourselves only to “feeling”, the conclusion is often insufficient. It is necessary to combine it with other analysis.

Systems are improving but they are not perfect yet

There is often a confusion with the use of the term “sentiment”. In French, experts use the term “opinion”, while in English "sentiment" and "opinion" are almost identical in the NLP field.

Let’s start again.

The definition of opinion analysis is: systems are improving but they are not perfect yet.

Some methods create an illusion of accuracy, of reliability. Even worse, others sell the concept as the magic index on which you should base your decisions.

At Phedone, we propose the analysis of personal involvement

At Phedone, we propose the analysis of personal involvement because it is an easy concept to understand for both the analyst and the client. We add and combine with other indexes from AI, statistics and business rules.

An analysis starts with a categorisation, then we add a colouring. This makes sorting much easier.

In almost real time, without assistance from level #0 or #1 support, or from experts and managers, 61% of the themes are generated and used in production.

This does not replace the analyst or the support but it does speed up the process thanks to our knowledge of the business. You get a first analysis in less than 10 minutes instead of 3 months. In 3 minutes, you get 88% of the information previously extracted manually in 2 months.

Conclusion

We combine different in-house indexes to find information that is not visible without Phedone.

At Phedone, our job is to sort out the important from the rest.

Here are some results of the Phedone solution:

63% of automatically detected categories are automated in the support tools
90% accuracy of automatic qualification
Between 20 and 50% of staff time is freed up.

We use AI, statistics and business rules to study the voice of the customer, the employees and now the patient and the medical staff in general.

No training is required, the head of the company, the infrastructure and information system department, and the quality department can immediately detect trends and weak signals. They can even sort and direct support requests automatically.

Natural Language ProcessingSentiment analysis, can we believe in it?

Discover the advanced AI thatautomates the processing of your customer returns

Get notified when we’re launching.