-
Book Overview & Buying
-
Table Of Contents
-
Feedback & Rating

AI Blueprints
By :

Our method for building a natural language question-answering service is straightforward. Three primary components are involved, as shown in Figure 4. Our method will be as follows:
The user first provides the query in text or voice. If voice is used, we can make a request to the Google Cloud Speech-to-Text API (https://cloud.google.com/speech-to-text/) to get the text version of the speech.
Next, we need to figure out what the user is asking about. There are lots of ways to ask the same question, and we wish to support several different kinds of questions in both example domains. Also, a question may contain internal phrases, or entities, such as the name of a particular Pokémon, or the name of a college course. We will need to extract these entities while also figuring out what kind of question is being asked.
Once we have the question and entities, we then compose a new kind of query. Since we are using Prolog to represent much of the domain knowledge...