Diagrams added to four more User Guide scenarios (#485)

user-guide/modules/ROOT/pages/task-metaprogramming.adoc

@@ -285,7 +285,9 @@ boost:hana[] adds efficient tuple handling at runtime (for example, easier acces
 | Better Compile-Time Speed	| Yes | No, Slower
 |===
 
-Let's update our sample to include _tag dispatching_, so each type is classified at compile time, and _runtime tuple processing_, so the sample iterates over heterogeneous types at runtime.
+Let's update our sample to include _tag dispatching_, so each type is classified at compile time, and _runtime tuple processing_, so the sample iterates over heterogeneous types at runtime. Not having all metaprogramming processes occur at the same time does add a level to your program complexity!
+
+image::template-time-machine.png[]
 
 [source,cpp]
 ----

user-guide/modules/ROOT/pages/task-natural-language-parsing.adoc

@@ -19,6 +19,7 @@ Natural language processing is a complex field that goes beyond just programming
 * <<Simple English Parsing Sample>>
 * <<Add a Dictionary of Valid Words>>
 * <<Add Detailed Error Reporting>>
+* <<Next Steps>>
 * <<See Also>>
 
 == Libraries
@@ -410,12 +411,20 @@ The sentence must be of the form:
   Noun 2:      - invalid
 ----
 
+== Next Steps
+
 You will notice how adding more features to a natural language parser starts to considerably increase the code length. This is a normal feature of language parsing - a lot of code can be required to cover all the options of something as flexible as language. For an example of a simpler approach to parsing _well-formatted_ input, refer to the sample code in xref:task-text-processing.adoc[].
 
+When designing a natural language system, it is good practice to divide the process into three main steps: *syntax analysis* (parsing), *semantic analysis* (grammar), and *interpretation*. Even though it might be possible to combine some grammar checking with the syntax checking, it is often better practice not to do this but complete the syntax checking first - this makes for greater maintainability and comprehension as the code is developed over time. The final stage, interpretation, usually depends on context. In the following sequence, a phrase as simple as "time flies" contains ambiguity:
+
+image::natural-language-parsing.png[]
+
+The phrase is so widely known it obscures the interpretation of "flies" as a noun. Context is required during interpretation to treat the input as a phrase, and not as something of the form "horse flies". Natural language parsers can remind us of ambiguity when at first we don't see it!
+
+
 == See Also
 
 * https://www.boost.org/doc/libs/latest/libs/libraries.htm#Parsing[Category: Parsing]
 * https://www.boost.org/doc/libs/latest/libs/libraries.htm#Patterns[Category: Patterns and Idioms]
 * https://www.boost.org/doc/libs/latest/libs/libraries.htm#String[Category: String and text processing]
 
-

user-guide/modules/ROOT/pages/task-quantum-computing.adoc

@@ -520,6 +520,8 @@ If you run this code you will get something like:
 
 Again, the result shows the probability distribution of being at any one of the positions at the end of the 2D walk.
 
+image::quantum-walk-simulation.png[]
+
 
 === Other Quantum Algorithms
 

user-guide/modules/ROOT/pages/task-system.adoc

@@ -267,6 +267,8 @@ The following command lines show how to run with defaults, run with options spec
 
 It might be important to record the time taken for system operations, both in testing and in the operation of a system app. So, let's integrate boost:chrono[] to measure the time taken for key filesystem operations, such as creating directories, writing to files, reading files, and deleting files.
 
+image::system-timing.png[]
+
 [source,cpp]
 ----
 #include <boost/filesystem.hpp>
@@ -395,9 +397,7 @@ int main(int argc, char* argv[]) {
 
 ----
 
-Note:: The code measures operation execution time in microseconds.
-
-The following is example output from running the sample.
+The following is example output from running the sample:
 
 [source,text]
 ----