a ka Robochef v 20 Objectives To convert semistructured NLP into a structured machine readable format Chop the carrots with a large knife gt CHOP CARROTS KNIFE Data ltRECIPEgt ID: 184725
Download Presentation The PPT/PDF document "Information Extraction From Recipes" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Information Extraction From Recipes
a
ka
Robochef
v 2.0Slide2
Objectives
To convert semi-structured NLP into a structured, machine readable format
“Chop the carrots with a large knife” ->
CHOP( CARROTS) (KNIFE)Slide3
Data
<RECIPE>
<TI>Lychee Sherbet</TI>
<IN>24 lychees</IN>
<IN>1
pk
unflavored
gelatine
</IN>
<IN>1/4 cup cold water</IN>
<IN>1/3 cup milk</IN>
<IN>1/2 cup sugar</IN>
<IN>1/3 cup milk</IN>
<IN>1 cup half and half</IN>
<IN>1 teaspoon lemon juice</IN>
<PR>Peel and seed lychees. Squeeze lychees through 2 pieces of cheesecloth to obtain 1 cup of juice. Sprinkle
gelatine
over cold water and let stand for 5 minutes. Scald milk, add soaked
gelatine
, and stir until thoroughly dissolved. Add sugar, mixing well. Cool. Add milk and half and half. Stir in lychee and lemon juice. Freeze in ice-cream freezer.</PR>
</RECIPE>Slide4
NER task
1
cup
half and half
<AMT>
<UNIT>
<FOOD>
Squeeze
lychees through 2 pieces of cheesecloth …
<ACT>
<FOOD>
<UTENSIL>Slide5
Group Words
Words like “batter”, which are not explicit ingredients but combinations of ingredients
Words like “dry ingredients”, which refer to a large set of ingredientsSlide6
Features
Current Word
.
Previous Word, Next word.
Previous
Label.
Parts of Speech for current word, previous word, and next word.
The appearance of numbers [0-9]
. Ends with “-ed”. No vowels. If the word is not contained in the list of ingredients. If the words is not in the ingredients AND it is a nounWord Length. First occurrence. The appearance of “mix-“ or “ingredient-”Slide7
Sematic Role Labeling
Squeeze
lychees
through 2 pieces of
cheesecloth
Squeeze
(
lychees)(cheesecloth)Arguments assigned to actions using distance in parse treeSlide8
Coreference Resolution
Resolving all instances of “lychees” is easy
Use the ingredients list
Resolving “dough” to “
flour”,”water
” is harder.
Treat it as a machine translation problem
Use EM
Didn’t get time to evaluate