dalelane

Keybase proof

I hereby claim:

• I am dalelane on github.
• I am dalelane (https://keybase.io/dalelane) on keybase.
• I have a public key whose fingerprint is 0950 6D8F 8321 BEC7 CD01 8CF5 4872 0994 FE4A ECD8

To claim this, I am signing this object:

dalelane

##########################################################################
#
#  xmldiff
#
#    Simple utility script to enable a diff of two XML files in a way 
#     that ignores the order or attributes and elements.
#
#    Dale Lane ([email protected])
#     6 Oct 2014
#

dalelane

# Converting roman numerals into numbers
#
#   A homework helper by Dale and Grace Lane
#          1-Nov-2014
#   
#    http://dalelane.co.uk/blog/?p=3244


#
# some simple helper functions to make 

dalelane

{ entities:
   [ { type: 'PERSON',
       class: 'SPC',
       level: 'NAM',
       mentions:
        [ { mtype: 'NAM',
            role: 'PERSON',
            class: 'SPC',
            text: 'Dale Lane',
            location: { begin: 0, end: 8, 'head-begin': 0, 'head-end': 8 } },

dalelane

/**
 * Downloads the contents of a news story, then use the 
 *   Watson Relationship Extraction service to identify
 *   the names of all the people mentioned in the story.
 *
 * @author Dale Lane
 */

var async = require('async');
var unfluff = require('unfluff');

dalelane

# Get these from Bluemix and set as environment variables
#    $BLUEMIX_WATSON_MACHTRANS_USER
#    $BLUEMIX_WATSON_MACHTRANS_PASS
#    $BLUEMIX_WATSON_MACHTRANS_URL

export FORMAT="rt=text"  # or json or xml

export LANG_FROM="enus"
export LANG_TO="frfr"

dalelane

{
    "env" : {
        "node": true
    },

    "rules" : {
        "strict": [2, "global"],

        "quotes": [1, "single"],
        "key-spacing": [1, { "beforeColon" : true, "afterColon" : true }],

dalelane

@rnn_dalelane

Inspired by https://medium.com/@samim/obama-rnn-machine-generated-political-speeches-c8abd18a2ea0 I tried training a recurrent neural network (RNN) for myself.

I exported every tweet I've ever posted as @dalelane, and used that as the training text. The idea was to see whether it would generate new tweets that looked like they could be things that I had written.

This is just a first quick attempt, with no attempts to tweak or tune the generation of the training model, or modify any other settings.

This is the kind of thing it's currently outputting:

---

dalelane

@rnn_dalelane

Attempt 2

Like https://gist.github.com/dalelane/f0a1b9ce75509875f91d but this time I tried tweaking some settings to "optimise" it.

Erm... it didn't go well.

---

@andypiper @andypiper @andypiper @hardillb @andypiper @andypiper @andypiper @andypiper @andypiper @andypiper @andypiper @andypiper @andypiper

---

@andypiper @andypiper @andypiper @andypiper @andypiper @andypiper @andypiper @andypiper @andypiper @andypiper @andypiper

dalelane

import requests
import image_slicer


# Gets the contents of an image file to be sent to the
# machine learning model for classifying
def getImageFileData(locationOfImageFile):
    with open(locationOfImageFile, "rb") as f:
        data = f.read()
        return data.encode("base64")