In the Heart of Silicon Valley
Driving the conversational connection to computer intelligence
Exponentially increasing computer processing power
Trends such specialized chips, cloud computing, and increasing processing at the “edge” (e.g., smartphones) are accelerating the amount computer power available to support conversational technology at a rate even faster than Moore’s Law. Further indirect support results from improvements in the utility of applications that can use the technology leveraging that increased computer power.
Strains on the Graphical User Interface (GUI)
The point-and-click GUI that led the way in democratizing access to computer power is reaching its limits. Applications have grown increasingly complex, making it often difficult to find the feature desired in menus or requiring many clicks to execute a task. This trend toward complexity is exacerbated on the small screens of smartphones using the large pointing device of a finger.
No manual required
The goal of a Language User Interface (LUI) is to allow a user to simply “say or type what you want.”
Artificial intelligence in general and machine learning using deep neural networks have improved the basic technologies required to support technologies such as Natural Language Processing and Speech Recognition.
Digital personal assistants
The personal assistants provided by major technology firms for smartphones, home speakers, automobiles, and other digital systems are popularizing the use of speech to interact with digital systems.
Automating customer service
A major trend is greeting a customer with a prompt such as “Please state why you are calling”, eliminating the long list of options that often frustrate customers.
Web sites and apps on smartphones provide the option to text in human language with automated systems for customer service, technical support, and other uses.
Making employees more efficient
The techniques that support customer service can support internal company activities, such as automating access to the human resources department or easier use of enterprise software.
Natural language understanding
Natural language processing—understanding the intent of a user request—has developed rapidly. Part of that rapid improvement is the use of highly computational techniques such as machine learning, but another supporting trend is the increasing size of the databases of speech and text that support that machine learning. When users increase their use of the LUI, it provides more data to improve its performance—a virtuous cycle.
Speech recognition has similarly benefited from machine learning and the increased data available through increased use of the modality. “Linguistic” techniques supplement the basic data analysis. For example, a song titles in data for a music selection application can be replaced by a “word” SONG-TITLE, which refers to a list of song titles available. This allows a new song to be added to a SONG-TITLE list without needing specific examples of the song during the machine learning.
Converting text to speech has reached a level of intelligibility that makes it equivalent to recorded voice files, but with much more flexibility in responses.
Today’s language technology works well when the computer response is based solely on the last user request. Maintaining a dialog that retains the context of the entire conversation is a goal that researchers are addressing.
Analyzing big data
“Big data” is often in the form of unstructured text or voice files (e.g., recorded customer service calls). Speech recognition and NLP techniques can uncover trends or find answers to inquiries in such data.