11 Reaching understanding through parsing and intent resolution

Chapter 10 covered the first component of a conversational system: speech-to-text (STT). This chapter is step 2: natural language processing (NLP) and natural language understanding (NLU). If chapter 10 focused on understanding what a user said, chapter 11 focuses on what those words mean, as NLU takes the STT representations and interprets the intended meaning behind those words to determine how to respond. Here, you learn what “meaning” means to a voice system, why it’s difficult, and how to reach the best result.

If your recognition goes wrong, your meaning interpretation probably will too: you can’t “make sense” out of the wrong words. Accurate understanding lets you respond sensibly to users and give them what they ask for. It lets you create more complex and sophisticated VUIs that handle ambiguity and pronoun references, and it lets you deal with mistakes intelligently when they occur (chapter 13). In this chapter, you learn about some of the methods involved in NLP and NLU. NLP assigns structure to a text string, while NLU uses that additional information to more accurately determine the user’s meaning and assign an intent that the VUI can act on and respond to.

11.1  From words to meaning

11.1.1    NLP

11.1.2    NLU

11.2  Parsing

11.3  Machine learning and NLU

11.4  Ontologies, knowledge bases and content databases

11.5  Intents

11.5.1    Intent tagging and tagging guides

11.5.2    Middle layers: semantic tags versus system endpoints

11.6  Putting it all together

11.6.1    Matching wide or narrow

11.6.2    Multiple grammars, multiple passes

11.7  A simple example

11.7.1    The Stanford Parser revisited

11.7.2    Determining intent

11.7.3    Machine learning and using knowledge

11.8  What’s next?

11.9  Summary