Voice App testing with Botium speech processing

Yes, it is there.

But the path, mentioned in the error is having an extra “HiTest/step0”.

in the Media Settings, remove the last two path components. Botium is concatenating this path with the path given in the convo files.

Updated it, but still same error. Now the duplication in path is removed.
But the file not found error is shown.
The exact location of the file is as below:
image

Here, the path is like Botium Box–>testsets–>LexbotTestSet…
But in the location specified by the error message, …botiumbox.com is shown in between testsets and LexbotTestSet.Will that be an issue?

image

1 Like

Thanks for pointing us at this, there is indeed a file storage issue in your Box Mini instance. It will be fixed tomorrow by lunchtime (CET).

Ok sure pls, will try after that.Thank you

done

Thank You.I could run the test cases now. At the end of the test results, I could see the audio attachments. I hope they are the bot request reponse.
But could not play it.

As in the above figure, I could see the recordings after test execution.Are they the recordings of the chatbot response?Could you please have a comment on that.

playing them should be possible now, also in the code view (clicking on <>)
response audio seems not be available there.

ya, I could play the audio. It is the audio request. Is there any way we will, get the bot response?
And other doubt is:
If I use the same audio input as utterance, The bot is not recognizing the input. Tried with different intents.

The same audio, if used directly, without uttereance, getting the correct result.

Utterances are for text content only. If you want to have Botium send more than one audio file, you can use wildcard matching in the MEDIA path.

Lex doesn’t return the TTS response by default. It is possible to enable it, but not in Box Mini. But anyway it doesn’t really make sense to use TTS for conversation flow testing, it just adds another layer of complexity.

Could unplease elaborate on the wild card matching in media path

https://botium-docs.readthedocs.io/en/latest/05_botiumscript/index.html#media

got it, Thank You

Is there any way, if we are passing the text input and it will be automatically converted to .wav file…
May be, we have 100s of test cases…so for all, do we have to create .wav files?

Thank you for your support. Could present it before team, and they are happy with this. But an issue is like, when I upload the .wav file, for all the .wav files, I am not getting the exact output.
I used different convertors to get the .wav files, from text.

Using a library of .wav files as input makes mostly sense if there are existing recordings available. But there may be reasons to do this.
In Botium Box there currently is no “batch” text-to-wav converter. You can do it one after the oterh in the Voice Wizard.

Don’t know what I can help with the exact output issue … speech-to-text transcriptions are not always 100 percent accurate, that’s the reason why testing should be performed with multiple voices/dialects and audio quality. In the Voice Wizard, there is an option to apply some noisy effects to existing audio files to prepare a library of voice test files. But it is not unusual, that the speech-to-text will fail in some cases

Ok sure.Will try with that.

I tried.The chatbot is recognizing the sound, and providing the correct output if I am recording the sound by Live Chat and saving the test case.
In the live chat, there is an option to upload the recording. I tried by uploading the sound. I need to check for different accent, which is a requirement for us.
Could you please confirm if there is any requirement to be satisfied, for the sounds that are uploading.

The audio files are uploaded 1:1, there is no pre-processing done by Botium. They have to fulfill the requirements given by the chatbot platform you are using. The Botium live chat recordings are Wav Mono 16bit, which is a format suitable for most voice applications.