Hasn't Google Solved the Problem?

The way that today's speech recognition technologies work to recognize various words is based upon what words they have access to in some sort of dictionary or database. When you use Google's Search by Voice, for example, their software is actively tapping into its available database to try to understand whatever you are saying in real time. This approach is the one most commonly used but it is limited by what is made available to the software through the database. Google attempted to remedy this by using their "Big Data" approach. In other words, they simply amassed as many words and as much data to put into their database as possible. The problem with this is that, at some point, the database will become too large to quickly sift through. Their solution to this issue was to implement a logic that we as humans use everyday: grammar. This logic works great when quickly recognizing actual words in structured sentences, but not so much with names. When it comes to recognizing spoken names, Google reverted back to their "Big Data" approach. They did, however, utilize this approach in a slightly different way than the way they did for normal words. Being that names can be derived from different languages, Google used their "Big Data" approach to account for different pronunciations and accents by collecting data from recordings of many different people. Their software takes whatever name you say to it and compares it against all of their collective data. Using this approach has limited the range of names that their software can recognize simply to the names of famous people or characters and popular or common names. Google's approach and their "Search by Voice" software is very limited in its capabilities and pertains only to a small fraction of the names in the world. Currently, their software's ability to recognize spoken names serves some purpose, but by no means solves the problem.