Pages

Monday, October 29, 2018

F# Stanford NLP is running

After some configuration, I can successfully run the first NLP project with F#. Special thanks to Sergey's post! The post is very informative. His solution is based on the F# interactive while I prefer to use the project-based solution.

Sergey points out that one of the common problems to setup is the path problem. His claim is so true. I had stuck in this problem for days. Here is the process I followed.


  • Open Visual Studio 2017 and create an F# console application. 
    • I tried .net core app; it does not work as the IKVM has the dependency on the .NET framework
  • compile the F# console application and remember the debug folder location
  • Open NuGet and retrieve Stanford NLP CoreNLP. The current version is 3.9.1
    • Current Stanford NLP is 3.9.2. I suggest you download 3.9.1 version
  • download the Standard NLP 3.9.1 zip file
  • unzip the 3.9.1 file to the F# console app debug folder
  • go the unzipped folder and find the model JAR file

  • download WINRAR to unzip the JAR file to a folder, this folder should contain a folder called "EDU"
  • copy the "EDU" folder up to debug folder, so the structure in the "DEBUG" folder is like the following.
  

The F# file I was using is listed below. Set the "EDU" folder to the debug folder can save you from setting the CurrentDirectory. 


1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
// Learn more about F# at http://fsharp.org
// See the 'F# Tutorial' project for more help.

open System
open System.IO
open java.util
open java.io
open edu.stanford.nlp.pipeline

[<EntryPoint>]
let main argv = 
    let text = "Kosgi Santosh sent an email to Stanford University. He didn't get a reply.";

    // Annotation pipeline configuration
    let props = Properties()
    props.setProperty("annotators","tokenize, ssplit, pos, lemma, ner, parse, dcoref") |> ignore
    props.setProperty("ner.useSUTime","0") |> ignore

    let pipeline = StanfordCoreNLP(props)

    // Annotation
    let annotation = Annotation(text)
    pipeline.annotate(annotation)

    // Result - Pretty Print
    let stream = new ByteArrayOutputStream()
    pipeline.prettyPrint(annotation, new PrintWriter(stream))
    printfn "%O" <| stream.toString()
    stream.close()

    printfn "%A" argv
    0 // return an integer exit code

Executing the NLP program seems taking a lot of memory. My program uses 2G memory and takes a while to show the result. Hopefully, your computer is faster enough. :)

Thursday, October 25, 2018

F# Enum usage II

I want to take a quick note on the F# enum usage again. The supporting of the space and unique character in F# language and editor is a great feature can make your development work much more comfortable. I am in the Web API front these days.

One of the requirement is to provide options to end users. If an application only takes "excellent choice", "good option", "ok choice", and "bad and never go there" as options, I'd like to offload these values check to the type system instead of handling the error in my code.

If I can declare the enum like the following

type enum MyEnum =
    ``excellent choice`` = 0
    | ``good option`` = 1
    ....

After declaring the enum-as-string attribute on the attribute [JsonConverter(typeof(StringEnumConverter))], the output and input validation is solved in one shoot.