At Modjo, our team of data scientists has developed a sophisticated detection algorithm to identify the language of each call.
The accuracy of this algorithm is remarkable, however, there may be instances where the identified language is not exact.
This can be due to two main reasons:
No audio: the call recording does not contain any sound in its entirety.
Poor audio quality: the call recording is disrupted by various background noises.
Example: The following call was conducted in English but was identified in French.
> The reason for this is that the call quality was not very clear: all participants were in the same room and using a speaker, which resulted in a significant background noise.
Conclusion
The main source of the problem is usually the audio input, which is either unclear or missing from most of the recording. As a result, the algorithm struggles to recognize the language correctly.
Related article:
How is a call composed?