A Moving Experience.

Motivation

Statistical and neural network ASR systems normalize text before training models with it. It works in two directions: our ‘tokenizer’ not only splits on whitespace, but also normalizes text as it is spoken. Numbers in digits, dates, time expressions etc. will be written out in words, spelling will be normalized, and some other minor rewrites will be done in order to bring all training material into our token space. On the other hand, the end user does not want to see ‘first of July nineteen ninety two’, so will need the inverse normalization to ‘July 1, 1992’ as a step we call ‘formatting’.

Currently both the tokenizer and the formatter are largely driven by grammar rules. Both require the input to be in a fixed format, normal text for the tokenizer and in ‘token space’ for the formatter. However, nowadays we are dealing with hybrid systems more and more and the output of these might not necessarily be the expected input. On top of this, we are increasingly dealing with multilingual systems where the input can be in several languages.
 

Target Languages & Phenomena

Research Questions

Is it possible to do formatting with the same accuracy as with the rule-based grammars? (On the basis of our current regression tests) Is the system able to think out of the box, to deal e.g. with wrong/non-existing phone numbers? What is the amount of training data needed? How well do multilingual systems perform? Would it help to feed the grammars to the LLM? Is it possible to divide the system into two parts, i.e. identifying the concept (e.g. a French phone number) and the actual formatting? If so, is it better to use LLMs or grammar rules to do the actual formatting? Is the system able to deal with non-standard input, i.e. partially formatted text for the formatter and partially normalized text for the tokenizer?

Cerence Inc. (Nasdaq: CRNC and www.cerence.com) is the global industry leader in creating unique, moving experiences for the automotive world. Spun out from Nuance in October 2019, Cerence is a new, independent company that has quickly gained traction as a leader in the automotive voice assistant space, working with all of the world’s leading automakers – from Ford and Fiat Chrysler to Daimler, Audi and BMW to Geely and SAIC – to transform how a car feels, responds and learns. Its track record is built on more than 20 years of industry experience and leadership and more than 500 million cars on the road today across more than 70 languages. 

As Cerence looks to the future and continues an ambitious growth agenda, we need someone to join the team and help build the future of voice and AI in cars. This is an exciting opportunity to join Cerence’s passionate, dedicated, global team and be a part of meaningful innovation in a rapidly growing industry.

EQUAL OPPORTUNITY EMPLOYER

Cerence is firmly committed to Equal Employment Opportunity (EEO) and to compliance with all federal, state and local laws that prohibit employment discrimination on the basis of age, race, color, gender, gender identity, gender expression, sex, sex stereotyping, pregnancy, national origin, ancestry, religion, physical or mental disability, medical condition, marital status, citizenship status, sexual orientation, protected military or veteran status, genetic information and other protected classifications. Cerence Equal Employment Opportunity Policy Statement.

All prospective and current Employees need to remain vigilant when it comes to executing security policies in the workplace. This includes:

- Following workplace security protocols and training programs to familiarize with the ways to maintain a safe workplace.
- Following security procedures to report any suspicious activity.
- Having respect for corporate security procedures to allow those procedures to be effective.
- Adhering to company's compliance and regulations.
- Encouraging to follow a zero tolerance for workplace violence.

- Basic knowledge of information security and data privacy requirements (e.g., how to protect data & how to be handling this data).

- Demonstrative knowledge of information security through internal training programs.