Our Services

Excel Spreadsheet Solutions

Bringing years of solving issues with spreadsheets in many technical arenas, Abacii offers services for improving the performance of your Excel spreadsheets. The tasks to achieve this goal include the following:

  1. Scan and gain an understanding of the intention and operation of all existing worksheets, formulas and macros.
  2. Construct a plan and schedule for correcting any known and/or discovered issues or inconsistencies in spreadsheet performance.
  3. Execute the planned solution by correcting formulas and/or macros, or by creating new spreadsheets, formulas or macros.
  4. Test all changes to ensure consistency and effectiveness of the implemented solution.

Speech Grammar Tuning

A speech system uses grammars to interpret what callers say. These grammars can often be improved, making the system more accurate and more reliable. This process is known as tuning.

A tuning project will generally involve these steps:

  1. Calls are recorded from an active speech system.
  2. These calls are transcribed.
  3. The existing grammars are used to interpret the recorded calls. These interpretations are tested against the transcribed calls, allowing a speech scientist to measure the accuracy of the existing system.
  4. A speech scientist works on the grammars in order to improve their accuracy, regularly testing them against the transcribed calls.
  5. The improved grammars are inserted into the active speech system.

We handle steps 2, 3, and 4. Our clients switch on the call recording option, collect the data, and send the data to us. We process the data, transcribe the calls, and tune the grammars. We then send the new grammars back to our clients, along with a full performance analysis.

Speech Grammar Writing

We can write grammars for new speech systems.

Possible goals for a new system include:

  • Routing a caller to the correct department (billing, insurance, etc)
  • Checking a caller’s bank balance
  • Checking the status of a caller’s order

Our clients design conversations that the system needs to handle in order to accomplish these goals.

An example conversation:

Caller: “Hi I want to know if my order has been dispatched”
System: “Of course. Could you tell me your order number, please?”
Caller: “Yes, it’s 78145”
System: “Thank you. Your order has been dispatched. It should arrive tomorrow between 3 and 5 pm”

We then write grammars that implement these conversations in a form that the speech system can use. The grammar is usually more complex than the designed conversation, because speech can vary greatly. For example, the first line of the conversation above could have taken any of the forms below:

Caller: “Hi I haven’t received my order”
Caller: “I need to find out if my order has been dispatched”
Caller: “Can you tell me if my order has been sent out?”

A well-written grammar will efficiently handle many possible variations of the conversation.

Speech Grammars in Spanish

We are able to handle Spanish-language data at every stage of our process: Call transcription, tuning existing grammars, and writing new grammars. Most commonly, we are asked to write Spanish-language versions of existing English-language systems, so that a speech system can offer callers a choice between these two languages.

Technical Notes

We work with these grammar types:

  • Standard Grammars, detailed above in the Speech Grammar Tuning and Speech Grammar Writing sections.
  • Dynamic Grammars, which look up information in order to improve their accuracy. For example, the system might ask the caller for a customer reference number, look up a list of existing orders (and the associated order numbers) for this customer, and then ask the caller to specify an order number. The system now only has to match the caller’s speech to an order number on the looked-up list, rather than to any possible order number.
  • Statistical Language Model (SLM) Grammars, which are built via statistical processing of the call data. These are suited for use where a caller may begin a conversation about many different subjects. For example, the system may greet the caller by asking “How may I help you?”, which can allow much more variation in caller responses.

We work with these speech recognition engines:

  • Nuance Recognizer 9
  • Nuance Recognizer 10
  • Microsoft Speech Platform

If you have a particular technical requirement not covered here, we would be happy to discuss it, and to see whether we can adjust to handle it.

If you are interested in any of the services shown here, please contact us at: 
enquiries@abaciivoice.com