Siri
Download as PPTX, PDF
1 like258 views
Siri works by first encoding sounds into a digital format, then connecting to cloud-based language models to comprehend speech and either handle requests locally or query the network. It recognizes letters from the speech using local and cloud models, estimates words, and creates possible interpretations. It then performs the most confident interpretation by determining intent and carrying out the function on the iPhone, checking its understanding if the speech is ambiguous.
Siri
- 2. Initial Sounds First, the sounds of your speech are immediately encoded into a compact digital form, preserving its information.
- 3. Language Comprehenders The signal from the iPhone connects with the Internet and communicates with a cloud-based model that consists of a series of language comprehenders.
- 4. Speech Evaluation Simultaneously, a local model evaluates the speech on the device. Then, with the cloud models, the device figures out if it needs the network to move forth or can handle the process locally. In other words, a network query could be sending a text or searching the web, while a local process would be playing a song or setting an alarm. If it is deemed local, then it does not use the cloud anymore.
- 5. Language to Letters Using both local and server models, the device recognizes which letters constitute which parts of the speech. Now that the speech is converted into letters-vowels and consonants, a language model can estimate the comprised words of the speech. Then, the system creates a list of possible interpretations of the words that your speech might mean.
- 6. Final Action After this, it is all downhill. Of the list of possible interpretations, the most confident result is used. The computer determines the intent of the speech, and performs the function in the iPhone. If your speech is too ambiguous at any point during the process, the computer will defer and make sure that the computer- determined intent is correct.