Practical Guide to building Actionable Linguistic Intelligence.

Lesson 1 : The two primary reasons why interactive natural language applications are so hard to build.

The two major obstacles to building interactive natural language applications are:

(1) The absence of a Lingustic Parser that would be up to the task, and

(2) The overwhelming complexity of the task.

1. Is it true that presently available Linguistic Parsers are not adequate for developing Actionable Linguistic Intelligence?

Let me explain. Why do I say that the present tools are inadequate? Didn’t Google recently release a good Linguistic Parser called “Parsey McParseface”?

To see for yourself, you can use the online demos of parsers are also based on Dependency Grammar: Stanford Parser and displaCy.

Try them! You will see that the parser’s output is not usable at all.

The text is enriched with grammatical tags and arrows. As a developer, what are you are supposed to do with it? The output is not helpful for mapping to entities and actions, which is what you need. In other words, with this type of output you cannot even begin to build interactive applications. The output is even harder to deal with than the input itself.

“If you will succeed in breaking the wall with your head — what will you do in the adjacent cell?” (S. J. Lec)

2. How not to be overwhelmed by Inherent Complexity of Natural Language?

Users of interactive applications expect natural language robots to be prepared for a HUGE number of words to be used. No startup or medium-sized company can afford to invest millions of working hours to cover all the necessary lexical fields.

You simply cannot do everything by yourself.

Examples of topics for which you may want to obtain pay-for-play (J) resources include:

“numerical quantities”

“dates and time intervals”

“colors”

“geographical locations”

etc, etc, etc.

Clearly, what is needed is an easy-to-use system for the processing of language that will support the sharing and instant re-using of lexical resources. Yet up until now there has been no such system. Existing tools do not provide an easy way to reuse the work of other developers of interactive applications.

And that alone is enough to preclude any possibility of building a successful and compelling applications with Actionable Linguistic Intelligence.