can computers understand time?

28
[email protected] School of Computer Science Leeds, 01/05/2014 presentation L@L PGR Seminar Can computers understand time? Michele Filannino

Upload: michele-filannino

Post on 18-Dec-2014

103 views

Category:

Technology


1 download

DESCRIPTION

Human brain has evolved to master, among the others, the capacity of extracting flows of events out of a speech or a written text. This temporal sense, mainly unconscious, allows us to summarise, organise, remember and combine different pieces of information working out new insights and discoveries. The temporal dimension is an inescapable and easy truth for us, but enabling machines to fully deal with time is a challenging task. Computers are still incapable of detecting temporal incompatibilities, summarising workflows or identifying causes and consequences of facts. My research wants to answer the following questions: Can computers understand time? And what possibilities will that unlock?

TRANSCRIPT

Page 1: Can computers understand time?

[email protected] of Computer Science

Leeds, 01/05/2014

presentation L@L PGR Seminar

Can computers understand time?

Michele Filannino

Page 2: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

can computers understand time?

■ what does it mean?

■ why is it challenging?

■ what possibilities will it unlock?

■ what is there for linguists?

2

Page 3: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

perspectives

■ computer science

■ information extraction

■ computational linguistics

■ text mining

■ machine learning

3

Page 4: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

perspectives

■ computer science

■ information extraction

■ computational linguistics

■ text mining

■ machine learning

4

Page 5: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

what does it mean?

■ source: written texts

■ goal: identify, interpret, communicate and

reason about events

■ easy for people

■ hard for machines

Temporal aspects of events provide a natural

mechanism for organising information

Page 6: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

Source: ISO-TimeML

linguistic key concepts

■ events: phrases denoting eventuality and states

● inflected verbs and nouns: spoken, deliver, will be published

■ temporal expressions: phrases denoting a temporal

entity such as an interval or a time point

● 01/05/2014, March 15, the next week, Saturday, at that time,

yesterday, 5 o’clock, 3 days, every 4 hours

■ links: temporal relation between two phrases

● BEFORE, AFTER, INCLUDES, ENDS, DURING, BEGINS

6

Page 7: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

let’s try!

■ On 28th Feb. 2010, Deutsche Bank released a note saying

that China's current economic policies would result in an

enormous surge in coal consumption over the next

decade.

■ In 1978, Steve Furber was appointed the Rolls-Royce

Research Fellow in Aerodynamics at Emmanuel College,

Cambridge and was awarded a PhD two years later on the

fluid dynamics of the Weis-Fogh principle.

7

Page 8: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

■ On 28th Feb. 2010(T), Deutsche Bank released a note

saying that China's current economic policies would

result in an enormous surge in coal consumption over the

next decade(T).

■ In 1978(T), Steve Furber was appointed the Rolls-Royce

Research Fellow in Aerodynamics at Emmanuel College,

Cambridge and was awarded a PhD two years later(T) on

the fluid dynamics of the Weis-Fogh principle.

8

temporal expressions

let’s try!

Page 9: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

9

temporal expressions

events

■ On 28th Feb. 2010(T), Deutsche Bank released(E) a note

saying(E) that China's current economic policies would

result(E) in an enormous surge(E) in coal consumption over

the next decade(T).

■ In 1978(T), Steve Furber was appointed(E) the Rolls-Royce

Research Fellow in Aerodynamics at Emmanuel College,

Cambridge and was awarded(E) a PhD two years later(T)

on the fluid dynamics of the Weis-Fogh principle.

let’s try!

Page 10: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

10

temporal expressions

events link them

■ On 28th Feb. 2010(T), Deutsche Bank released(E) a note

saying(E) that China's current economic policies would

result(E) in an enormous surge(E) in coal consumption over

the next decade(T).

■ In 1978(T), Steve Furber was appointed(E) the Rolls-Royce

Research Fellow in Aerodynamics at Emmanuel College,

Cambridge and was awarded(E) a PhD two years later(T)

on the fluid dynamics of the Weis-Fogh principle.

let’s try!

Page 11: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

visual representation

11

now28 Feb. 2010

released, saying

2 years

1978 1980

awarded

appointed

2020

surge

Page 12: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

Source: Temporal Information Extraction and Shallow Temporal Reasoning, D. Roth et al. 2012

why is it challenging?

1. Matt exercised during his lunch break.

2. He stretched, lifted weights, and ran.

3. He showered, got dressed and returned work.

12

Page 13: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

1. Matt exercised(E) during his lunch break(E).

2. He stretched(E), lifted(E) weights, and ran(E).

3. He showered(E), got dressed(E) and returned(E) work.

Source: Temporal Information Extraction and Shallow Temporal Reasoning, D. Roth et al. 2012

linguistic knowledge

13

exercised

lunch break

Page 14: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

1. Matt exercised(E) during his lunch break(E).

2. He stretched(E), lifted(E) weights, and ran(E).

3. He showered(E), got dressed(E) and returned(E) work.

Source: Temporal Information Extraction and Shallow Temporal Reasoning, D. Roth et al. 2012

linguistic knowledge

14

stretch, lift, run

lunch break

exercised

Page 15: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

1. Matt exercised(E) during his lunch break(E).

2. He stretched(E), lifted(E) weights, and ran(E).

3. He showered(E), got dressed(E) and returned(E) work.

Source: Temporal Information Extraction and Shallow Temporal Reasoning, D. Roth et al. 2012

linguistic knowledge

15

shower, dress, return

lunch break

exercised

stretch, lift, run

Page 16: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

1. Matt exercised(E) during his lunch break(E).

2. He stretched(E), lifted(E) weights, and ran(E).

3. He showered(E), got dressed(E) and returned(E) work.

Source: Temporal Information Extraction and Shallow Temporal Reasoning, D. Roth et al. 2012

common sense knowledge

16

shower, dress, return

lunch break

exercised

stretch, lift, run

Page 17: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

1. Matt exercised(E) during his lunch break(E).

2. He stretched(E), lifted(E) weights, and ran(E).

3. He showered(E), got dressed(E) and returned(E) work.

Source: Temporal Information Extraction and Shallow Temporal Reasoning, D. Roth et al. 2012

common sense knowledge

17

lunch break

exercised shower, dress, return

stretch, lift, run

Page 18: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

1. Matt exercised(E) during his lunch break(E).

2. He stretched(E), lifted(E) weights, and ran(E).

3. He showered(E), got dressed(E) and returned(E) work.

Source: Temporal Information Extraction and Shallow Temporal Reasoning, D. Roth et al. 2012

common sense knowledge

18

lunch break

exercised shower dress return

stretch, lift, run

Page 19: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

1. Matt exercised(E) during his lunch break(E).

2. He stretched(E), lifted(E) weights, and ran(E).

3. He showered(E), got dressed(E) and returned(E) work.

Source: Temporal Information Extraction and Shallow Temporal Reasoning, D. Roth et al. 2012

domain knowledge

19

stretch liftrun stretch

lunch break

exercised shower dress return

Page 20: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

Source: http://start.csail.mit.edu/answer.php?query=

what possibilities will it unlock?

20

Page 21: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

Source: https://www.google.it/search?q=google+stock+price

better search results

21

Page 22: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

Source: GMail

events in emails

22

Page 23: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

Source: i2b2/2012 Clinal NLP training data

clinical timeline extraction

23

ADMISSION DATE: 2011-02-06; DISCHARGE DATE: 2011-02-08; HISTORY OF PRESENT ILLNESS: Mr. Pohl is a 53 - year-old male with history of alcohol use and hypertension. Blood alcohol level was 383. Agitated in emergency room requiring 4 leather restraints, received 5 mg of Haldol, 2 mg of Ativan. He became hypotensive in the emergency room with a systolic blood pressure in the 80 's and had decreased respiratory rate. He received a normal saline bolus of 2 litres of good blood pressure response. The patient was then admitted to the medical Intensive Care Unit for observation and then transferred to our service on medicine when the blood pressures remained stable overnight...

06/02/2011 07/02/2011 08/02/2011

General

Tests

Treatments

Problems

admission discharge

BAL 383

Haldol 4mgAtivan 2mg

hypotensive

SBP ~80decreased respiratory rate

Saline bolus 2l

transfer

stable

SBP stable

hands tremor improved

blood pressure medications

Page 24: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

URL: http://www.cs.man.ac.uk/~filannim/mantime.html

what is there for linguists?

24

Page 25: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

Note: Online DEMO soon available

concept’s temporal footprint

25

Dante Alighieri(1265-1321)

Galileo Galilei(1564-1642)

Page 26: Can computers understand time?

/ 2601/05/2014, Leeds

presentation L@L PGR Seminar

TE = Temporal expressions

open problems

■Are TE, events and links enough? how to spot them?

● prepositions, surrounding words, digits, temporal dictionary,

verbs, noun phrases

■ “I’ve played Tennis for 10 years” vs.

“I’ve played Tennis for 4 hours.”

■How these systems can automatically adapt to

different languages?

■ statistics != linguistics

26

Page 27: Can computers understand time?

Thank you.

Page 28: Can computers understand time?

Contact:

[email protected]

?QUESTIONS