CLIENT ISSUE
| OUR SOLUTION / SERVICE
|
"What is Speech Recognition? What is it used for and where?" | |
"Could I introduce voice in my existing services and legacy / DTMF IVRs and how?" | We analyse your existing digital CRM channels, esp. your Call Centre processes and any IVRs you may have in place and specify those sub-processes and channels that would benefit from voice and chat automation, those that should use a mixture of both speech and touchtone / DTMF (and where in the call flow exactly), but also those processes that shouldn't be touched! |
"OK, I have decided I want to introduce Voice as part of our Digital Strategy. How do I get going?" | We carry out Discovery Workshops with you to understand your existing CRM processes and your target user groupsWe study your existing marketing strategy and your branding and design and flesh out the most appropriate IVR | Voicebot | Chatbot persona for brand consistency and stickiness.We design detailed Call Flows and Conversations exhibiting your business and application logic, incl. error recovery and repair strategies to deal with misunderstandings and unexpected input or behaviour.
We perform Usability tests after the first prototype has been developed to assess the intuitiveness and efficiency of the new interface.We specify all the different system prompts, i.e. everything that your system may need to say to the user at any one point (in the best & worst scenarios, incl. error messages)We identify the most appropriate voice talent(s) to record all those messages (if you don't want to use Text-to-Speech / TTS) and sit down with them at the studio and coach them through the recordings in order to get the right communication style and tone
We write speech recognition and NLU grammars trying to capture all the things that the majority of your users are likely to say or write at each step.We carry out design logic and functionality tests to identify any bugs, errors and inconsistencies in the logic, the message wording, the pronunciation or the recognition grammars.We adjust and tune everything that needs optimising (prompts, pronunciations, grammars, dialogue strategies and ASR / NLU parameters).
|
"We have actually already started introducing speech in our IVR, but now need to ensure the usability and effectiveness of the design, the recognition grammars and the system prompts." | We review all your current call flows and dialogue | VUI designs, IVR prompts and speech recognition grammars and check for: inconsistencies and holes in the application logic or the system message wording, unnecessary complexities in the call flow or the prompt formulation, superfluous and missing steps, overgeneration and grammars with incomplete coverage. We optimise the original design (e.g. the order of questions to the caller), reword system prompts and tune speech recognition grammars and ASR parameters. |
"We already have a speech IVR in production. We now want to improve its usability and acceptability, improve call completion rates, and shorten call duration times." | As above: we review all your existing call flows and VUI designs, IVR prompts and speech recognition grammars; minimise any unnecessary complexities and superfluous steps in the dialogue flow and optimise the order of questions to the caller; simplify and clarify the prompt wording; eliminate overgeneration and tune the recognition grammars and, if needed, the ASR parameters.
Those observations, recommendations and modifications are also based on our independent testing of your IVR and an analysis of live call recordings using Speech Analytics Software. |
"We already have a live speech IVR. We now want to quickly identify and resolve any recognition or usability issues." | We listen to, annotate and analyse live calls that you have recorded from your system (Speech Analytics) and also perform our own testing resulting in a prioritised list of recommendations for changes, additions or things to take out, that will have an immediate positive effect on the IVR. E.g., a slight change in the message wording may prevent the need to change the whole recognition grammar at that step! Or deactivating barge-in at certain points in a call may immediately improve speech recognition accuracy, because the system will be processing complete caller utterances and not just their endings! |
"We have built a Voice Assistant Skill and want to quickly identify and resolve any recognition or design issues." | As above, we will test your skill to its limits, so that your users don't have to, but also check any user logs or recordings you have access to to identify any low-hanging fruit, i.e. suggestions for improvement to the Voice Design, the error repair strategies, the NLU or your messaging to the user. E.g., a slight change in the prompt wording can do away with changing the whole grammar at that step! |
"We have extensive experience building Chatbots, but want now to get into Voice and Voice Assistants. How do we go about it?" | We have already helped Digital Agencies get into the world of Voice, e.g. by explaining to them the differences between text input and speech input and the associated complexities of NLP and NLU versus ASR. We even prepared one Agency's VUI Design training course by adjusting their existing Chatbot course material to the special challenges of Speech. |
Your Query / Issue? | We always provide both short-term (easy to implement) tuning recommendations and longer-term solutions, depending on your time and budget requirements |