short-order

short-order is an exerimental natural language conversational agent intended for domains with a fixed vocabulary of entities and a small number of intents. Uses might include ordering food from a restaurant or organizing your song collection.

The first stage of development is a set of tokenizers that detect entities, intents, and quantities. As an example, consider the following utterance, which would typically come from a speech-to-text system:

I would like a Dakota burger with no onions extra pickles fries and a coke

In this example, the utterance has no commas, since they were not provided by the speech-to-text process.

Using ShortOrder, this text might be tokenized as

[ADD_TO_ORDER] [QUANTITY(1)] [DAKOTA_BURGER(pid=4)] [QUANTITY(0)]
[SLICED_RED_ONION(pid=5201)] [QUANTITY(1)] [PICKLES(pid=5200)]
[MEDIUM_FRENCH_FRIES(pid=401)] [CONJUNCTION] [QUANTITY(1)]
[MEDIUM_COKE(1001)]

After tokenization, a parser might be able to group the tokens into a tree that reflects the speaker's intent:

[ADD_TO_ORDER]
    [QUANTITY(1)] [DAKOTA_BURGER(pid=4)]            // Burger, standalone menu item.
        [QUANTITY(0)] [SLICED_RED_ONION(pid=5201)]  //   Remove onion modification
        [QUANTITY(1)] [PICKLES(pid=5200)]           //   Add pickles modification
    [QUANTITY(1)] [MEDIUM_FRENCH_FRIES(pid=401)]    // French Fries, standalone menu item
    [QUANTITY(1)] [MEDIUM_COKE(1001)]               // Coke, standalone

Try It Out

ShortOrder is currently in the earliest stages of development, so documentation is sparse or nonexistant, and the code stability is uneven.

If you are interested in taking a look, you can clone the repo on GitHub or install short-order with npm.

npm install shortorder

As of commit 3166f1e2, there are a number of working samples, based on a ficticious restaurant.

These samples are not included in the short-order npm package. To use them, you must clone the repo from GitHub.

You can find the definition files for the menu, intents, attributes, and quantifiers at

samples\data\menu.yaml
samples\data\intents.yaml
samples\data\attributes.yaml
samples\data\quantifiers.yaml

Relevance Test Sample

This sample runs a suite of test utterances through the tokenization pipeline. The test utterances can be found at samples\data\tests.yaml.

If you've cloned the repo, you can build and run the sample as follows:

npm install
npm run compile
node build/samples/relevance_demo.js

The output is the sequence of tokens extracted for each test utterance:

% node build/samples/relevance_demo.js

14 items contributed 143 aliases.
5 items contributed 22 aliases.
16 items contributed 31 aliases.
60 items contributed 210 aliases.

All tests passed.

0 general - PASSED
   input "Hamburger with extra pickles"
  output "[ENTITY:HAMBURGER,1] [QUANTITY:1] [QUANTITY:1] [ENTITY:PICKLES,5200]"
expected "[ENTITY:HAMBURGER,1] [QUANTITY:1] [QUANTITY:1] [ENTITY:PICKLES,5200]"

1 general - PASSED
   input "Uh yeah I'd like a pet chicken fries and a coke"
  output "[UNKNOWN:Uh] [INTENT:SEPERATORS] [INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:GRILLED_PETALUMA_CHICKEN_SANDWICH,100] [ENTITY:MEDIUM_FRENCH_FRIES,401] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:MEDIUM_COKE,1001]"
expected "[UNKNOWN:Uh] [INTENT:SEPERATORS] [INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:GRILLED_PETALUMA_CHICKEN_SANDWICH,100] [ENTITY:MEDIUM_FRENCH_FRIES,401] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:MEDIUM_COKE,1001]"

2 general - PASSED
   input "Uh yeah I'd like a pet chicken french fries and a coke"
  output "[UNKNOWN:Uh] [INTENT:SEPERATORS] [INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:GRILLED_PETALUMA_CHICKEN_SANDWICH,100] [ENTITY:MEDIUM_FRENCH_FRIES,401] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:MEDIUM_COKE,1001]"
expected "[UNKNOWN:Uh] [INTENT:SEPERATORS] [INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:GRILLED_PETALUMA_CHICKEN_SANDWICH,100] [ENTITY:MEDIUM_FRENCH_FRIES,401] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:MEDIUM_COKE,1001]"

3 general - PASSED
   input "Can I get a cheeseburger with swiss"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:CHEESEBURGER,2] [QUANTITY:1] [ENTITY:SWISS_CHEESE_SLICE,5102]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:CHEESEBURGER,2] [QUANTITY:1] [ENTITY:SWISS_CHEESE_SLICE,5102]"

4 general - PASSED
   input "I'll have two six piece wings"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:6_WINGS,410]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:6_WINGS,410]"

5 general - PASSED
   input "I'll have five dozen wings"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:5] [ENTITY:12_WINGS,411]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:5] [ENTITY:12_WINGS,411]"

6 general - PASSED
   input "Get me a coffee with two creams and one sugar"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:MEDIUM_COFFEE,1101] [QUANTITY:1] [QUANTITY:2] [ENTITY:CREAM,1194] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:SUGAR,1190]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:MEDIUM_COFFEE,1101] [QUANTITY:1] [QUANTITY:2] [ENTITY:CREAM,1194] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:SUGAR,1190]"

7 general - PASSED
   input "Large iced tea unsweet"
  output "[ENTITY:LARGE_UNSWEET_TEA,1072]"
expected "[ENTITY:LARGE_UNSWEET_TEA,1072]"

8 bugreport - PASSED
   input "can I have two hamburgers"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:HAMBURGER,1]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:HAMBURGER,1]"

9 bugreport - PASSED
   input "Can I get a coffee I'd also like two hamburgers"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:MEDIUM_COFFEE,1101] [INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:HAMBURGER,1]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:MEDIUM_COFFEE,1101] [INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:HAMBURGER,1]"

Suites:
  general: 8/8
  bugreport: 2/2

Priorities:
  1: 10/10

Overall: 10/10

You can run the Spanish version as follows:

npm install
npm run compile
node build/samples/relevance_demo_spanish.js

It will produce output like

% node build/samples/relevance_demo.js

14 items contributed 133 aliases.
5 items contributed 35 aliases.
16 items contributed 37 aliases.
60 items contributed 253 aliases.

Failing tests:
0 general - PASSED
   input "Hamburguesa con Pickles Extra"
  output "[ENTITY:HAMBURGER,1] [QUANTITY:1] [ENTITY:PICKLES,5200] [QUANTITY:1]"
expected "[ENTITY:HAMBURGER,1] [QUANTITY:1] [ENTITY:PICKLES,5200] [QUANTITY:1]"

1 general - PASSED
   input "Si me gustaría unas pollo grillado petaluma papas fritas y una coca"
  output "[INTENT:SEPERATORS] [INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:GRILLED_PETALUMA_CHICKEN_SANDWICH,100] [ENTITY:MEDIUM_FRENCH_FRIES,401] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:MEDIUM_COKE,1001]"
expected "[INTENT:SEPERATORS] [INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:GRILLED_PETALUMA_CHICKEN_SANDWICH,100] [ENTITY:MEDIUM_FRENCH_FRIES,401] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:MEDIUM_COKE,1001]"

2 general - PASSED
   input "Si me gustaria unas pollo grillado petaluma papas pequeñas y una coca"
  output "[INTENT:SEPERATORS] [INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:GRILLED_PETALUMA_CHICKEN_SANDWICH,100] [ENTITY:SMALL_FRENCH_FRIES,400] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:MEDIUM_COKE,1001]"
expected "[INTENT:SEPERATORS] [INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:GRILLED_PETALUMA_CHICKEN_SANDWICH,100] [ENTITY:SMALL_FRENCH_FRIES,400] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:MEDIUM_COKE,1001]"

3 general - PASSED
   input "Puedo pedir una hamburguesa con queso suizo"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:CHEESEBURGER,2] [ENTITY:SWISS_CHEESE_SLICE,5102]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:CHEESEBURGER,2] [ENTITY:SWISS_CHEESE_SLICE,5102]"

4 general - PASSED
   input "Quiero dos alitas de seis"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:6_WINGS,410]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:6_WINGS,410]"

5 general - FAILED
   input "Quiero cinco alitas de doce"
  output "[INTENT:ADD_TO_ORDER] [UNKNOWN:cinco] [ENTITY:12_WINGS,411]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:5] [ENTITY:12_WINGS,411]"

6 general - PASSED
   input "Dame un cafe con dos cremas y un azucar"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:MEDIUM_COFFEE,1101] [QUANTITY:1] [QUANTITY:2] [ENTITY:CREAM,1194] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:SUGAR,1190]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:MEDIUM_COFFEE,1101] [QUANTITY:1] [QUANTITY:2] [ENTITY:CREAM,1194] [INTENT:CONJUNCTION] [QUANTITY:1] [ENTITY:SUGAR,1190]"

7 general - PASSED
   input "Te sin edulcorante grande"
  output "[ENTITY:LARGE_UNSWEET_TEA,1072]"
expected "[ENTITY:LARGE_UNSWEET_TEA,1072]"

8 bugreport - PASSED
   input "Quiero dos hamburguesas"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:HAMBURGER,1]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:HAMBURGER,1]"

9 bugreport - PASSED
   input "Puedo pedir un cafe también quiero dos hamburguesas"
  output "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:MEDIUM_COFFEE,1101] [INTENT:CONJUNCTION] [INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:HAMBURGER,1]"
expected "[INTENT:ADD_TO_ORDER] [QUANTITY:1] [ENTITY:MEDIUM_COFFEE,1101] [INTENT:CONJUNCTION] [INTENT:ADD_TO_ORDER] [QUANTITY:2] [ENTITY:HAMBURGER,1]"

Suites:
  general: 7/8
  bugreport: 2/2

Priorities:
  1: 9/10

Overall: 9/10

REPL Sample

This sample provides a Read-Eval-Print-Loop that runs the tokenizer on each line entered.

If you've cloned the repo, you can build and run the sample as follows:

npm run compile
node build/samples/repl_demo.js

% node build/samples/repl_demo.js

Welcome to the ShortOrder REPL.
Type your order below.
A blank line exits.

14 items contributed 143 aliases.
5 items contributed 22 aliases.
16 items contributed 31 aliases.
60 items contributed 210 aliases.

% i'd like a dakota burger fries and a coke

INTENT: ADD_TO_ORDER: "i'd like"
QUANTITY: 1: "a"
ENTITY: DAKOTA_BURGER(4): "dakota burger"
ENTITY: MEDIUM_FRENCH_FRIES(401): "fries"
INTENT: CONJUNCTION: "and"
QUANTITY: 1: "a"
ENTITY: MEDIUM_COKE(1001): "coke"

% actually make that a pet chicken with extra pickles

INTENT: CANCEL_LAST_ITEM: "actually"
INTENT: RESTATE: "make that"
QUANTITY: 1: "a"
ENTITY: GRILLED_PETALUMA_CHICKEN_SANDWICH(100): "pet chicken"
QUANTITY: 1: "with"
QUANTITY: 1: "extra"
ENTITY: PICKLES(5200): "pickles"

%
bye

Stemmer Confusion Matrix Sample

In some cases, the stemmer can stem words with different meanings to the same term. One can check for these problems in their menu.json, quantifiers.json, and intents.json files by producing a stemmer confusion matrix.

node build\samples\stemmer_confusion_demo.js

14 items contributed 143 aliases.
5 items contributed 22 aliases.
16 items contributed 31 aliases.
60 items contributed 210 aliases.
"fri": [fried,fries]
"that": [that,that's]
"thank": [thank,thanks]

In the example above, we see that the words fries and fried are treated as the same term, causing the phrase, "I'll have a pet chicken fries and a coke" to be interpreted as "pet chicken fried", instead of a "pet chicken" and "French fries".

% I'll have a pet chicken fries and a coke

INTENT: ADD_TO_ORDER: "I'll have"
QUANTITY: 1: "a"
ENTITY: FRIED_PETALUMA_CHICKEN_SANDWICH(101): "pet chicken fries"
INTENT: CONJUNCTION: "and"
QUANTITY: 1: "a"
ENTITY: MEDIUM_COKE(1001): "coke"

One can address this problem with a different stemmer or lemmatizer. One simple work-around is to wrap the default stemmer in a function that has special handling for certain words like fried and fries:

function hackedStemmer(term: string): string {
    const lowercase = term.toLowerCase();
    if (lowercase === 'fries' || lowercase === 'fried') {
        return lowercase;
    }
    return Tokenizer.defaultStemTerm(lowercase);
}

lulzzz / shortorder Goto Github PK

shortorder's Introduction

short-order

Try It Out

Relevance Test Sample

REPL Sample

Stemmer Confusion Matrix Sample

Tokenizer Design Notes

shortorder's People

Contributors

Watchers

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent