Given the recent introduction of Echo, the voice-driven virtual home assistant device from Amazon, we’ve received inquiries about how MindMeld’s voice interface platform compares to Amazon’s Alexa, the brains behind Echo.
We have our own Echo here in the office and have a lot of fun playing around with it. It’s a great platform and does many things well. In particular, we’re excited that it’s opening minds to how useful voice can be as an interface for anything, beyond just mobile phones and tablets.
There are a few separate components to Amazon’s Alexa offering:
- Alexa: Amazon’s intelligent virtual assistant.
- Echo: The cylindrical speaker Alexa inhabits.
- Alexa Voice Service: Enables Alexa to “live” on other hardware/devices, freeing the assistant from only inhabiting the Echo.
- Alexa Skills Kit: An SDK to give Alexa new capabilities.
As a platform for building custom voice functionality, it is the Alexa Skills Kit that may have distant similarity to MindMeld. However, there are key differences in focus that distinguish the two services’ primary use cases. Indeed, in certain situations, Alexa and MindMeld can even complement each other, pairing up to harmoniously deliver a strong, flexible voice solution.
Let’s dive deeper: Alexa has built-in functionality of its own (such as searching Wikipedia, telling jokes, and setting alarms), but with the Skills Kit, Alexa acts like a switchboard operator between discrete applications. For instance, in Amazon’s own example, a developer of a horoscopes app could equip Alexa with the ability to answer queries such as:
- “Alexa, ask Daily Horoscopes for the Gemini horoscope”
- “Alexa, ask Daily Horoscopes about Gemini”
- “Alexa, ask Daily Horoscopes to tell me the horoscope for Gemini”
In these queries, Gemini occupies what Amazon calls a “slot.” In this case, it’s the Sign slot, representing a fixed set of astrological signs. For a query of a particular structure to be possible, developers must specify example utterances following that same form. Developers also need to include at least one example utterance for each of the possible values of fixed-set slots like Sign.
Combined with the Alexa Voice Service, the Alexa Skill Set makes it easy and quick to build a simple app that is highly device-flexible, opening up exciting possibilities for hardware creators and developers of simple IoT applications.
For some applications, hard-coding all possible sentence structures and slot values is time-consuming enough to become an impossibility. This is where MindMeld enters the picture. The MindMeld platform has been built from the ground up to be optimized for large custom content collections. With MindMeld, any domain can be converted into a knowledge graph that defines all relevant concepts, their related ideas, and their likelihood of showing up in a search.
When the custom knowledge graphs are connected with MindMeld’s state-of-the-art NLU technology, the real power of the platform comes to the fore. With MindMeld, several hundred sample queries can power an application with human-like understanding of potentially millions of unique queries. This opens up the possibilities of a voice interface for any complex, large-vocabulary domain.
