From Session Detection to Mission Detection Matthias Hagen Jakob - - PowerPoint PPT Presentation

from session detection to mission detection
SMART_READER_LITE
LIVE PREVIEW

From Session Detection to Mission Detection Matthias Hagen Jakob - - PowerPoint PPT Presentation

From Session Detection to Mission Detection Matthias Hagen Jakob Gomoll Anna Beyer Benno Stein Bauhaus-Universit at Weimar matthias.hagen@uni-weimar.de OAIR 2013 Lisbon, Portugal May 23, 2013 Hagen, Gomoll, Beyer, Stein From Search


slide-1
SLIDE 1

From Session Detection to Mission Detection

Matthias Hagen Jakob Gomoll Anna Beyer Benno Stein

Bauhaus-Universit¨ at Weimar matthias.hagen@uni-weimar.de

OAIR 2013 Lisbon, Portugal May 23, 2013

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 1

slide-2
SLIDE 2

What is the user searching?

manhattan

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 2

slide-3
SLIDE 3

Without context . . .

source: [http://usatravel.about.com/od/Pictures-And-Maps/ss/Amazing-Aerial-Views-Of-America.htm]

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 3

slide-4
SLIDE 4

What if you knew the previous queries? party ideas cocktail recipes caipirinha manhattan

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 4

slide-5
SLIDE 5

What if you knew the previous queries? party ideas cocktail recipes caipirinha manhattan

source: [https://commons.wikimedia.org/wiki/File:Manhattan Cocktail2.jpg]

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 4

slide-6
SLIDE 6

What if you knew the previous queries? party ideas cocktail recipes caipirinha manhattan

Improves

Intent understanding Retrieval precision Disambiguation

source: [https://commons.wikimedia.org/wiki/File:Manhattan Cocktail2.jpg]

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 4

slide-7
SLIDE 7

A typical query log

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 istanbul archeology 2013-04-21 12:02:54 istanbul archeology 2013-04-21 18:31:21 weather new york 2013-04-21 18:45:23 constantinople 2013-04-21 18:45:36 footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 derby eterno 2013-04-21 22:42:48 constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 5

slide-8
SLIDE 8

Physical sessions

(gaps ≤ 30 minutes)

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — — — — — istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — — — — — istanbul archeology 2013-04-21 18:31:21 weather new york 2013-04-21 18:45:23 constantinople 2013-04-21 18:45:36 footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 — — — — — — — — — — — — — — — — — benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — — — — — derby eterno 2013-04-21 22:42:48 constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 6

slide-9
SLIDE 9

Physical sessions → interleaved intents

Query Time Intent ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — — — — — istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — — — — — istanbul archeology 2013-04-21 18:31:21 history weather new york 2013-04-21 18:45:23 weather constantinople 2013-04-21 18:45:36 history footbal lisbon 2013-04-21 19:14:01 sports football lisbon 2013-04-21 19:14:11 sports — — — — — — — — — — — — — — — — — benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — — — — — derby eterno 2013-04-21 22:42:48 constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 6

slide-10
SLIDE 10

Actual search intent switches

Query Time Intent ancient turkey 2013-04-20 20:02:44 history history istanbul 2013-04-20 20:24:17 istanbul archeology 2013-04-21 12:02:54 istanbul archeology 2013-04-21 18:31:21 — — — — — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 weather — — — — — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 history — — — — — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 sports football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 history constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 7

slide-11
SLIDE 11

Long-term tasks

Query Time Intent ancient turkey 2013-04-20 20:02:44 history history istanbul 2013-04-20 20:24:17 istanbul archeology 2013-04-21 12:02:54 istanbul archeology 2013-04-21 18:31:21   — — — — — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 weather — — — — — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 history         — — — — — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 sports football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 history constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 7

slide-12
SLIDE 12

Multitasking and search missions

Observations

[Spink et al., 2006; Jones and Klinkner, 2008]

Physical sessions: interleaved intents (multitasking) Long-term tasks: several sessions (search missions)

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 8

slide-13
SLIDE 13

Multitasking and search missions

Observations

[Spink et al., 2006; Jones and Klinkner, 2008]

Physical sessions: interleaved intents (multitasking) Long-term tasks: several sessions (search missions)

Traditional session detection

Only consecutive queries → Missions impossible

Example

history istanbul 2013-04-20 20:24:17 same istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — new

  • football lisbon

2013-04-21 19:14:11 — — — — — — — — — new

  • constantinople

2013-04-21 23:09:02

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 8

slide-14
SLIDE 14

Our topic . . . Pre-retrieval session + mission detection

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 9

slide-15
SLIDE 15

Our topic . . . Pre-retrieval session + mission detection

Remark: Runtime is crucial!

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 9

slide-16
SLIDE 16

Typical query similarity features

Temporal thresholds 5 minutes

[Silverstein et al., 1999]

15 minutes

[He and G¨

  • ker, 2000]

30 minutes

[Downey et al., 2007]

120 minutes

[Buzikashvili and Jansen, 2006]

user specific

[Murray et al., 2006]

Lexical similarity term overlap

[Kotov et al., 2011]

n-gram overlap

[Zhang and Moffat, 2006]

Levenshtein distance

[Jones and Klinkner, 2008]

reformulation patterns

[Huang and Efthimiadis., 2009]

Semantic similarity ESA

[Lucchese et al., 2011]

Search results

[Radlinski and Joachims, 2005]

Linked Open Data

[Hollink et al., 2011]

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 10

slide-17
SLIDE 17

Previous methods

Feature combinations

More accurate than single features One of the best: Geometric method (time + lexical)

[Gayo-Avello, 2009]

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 11

slide-18
SLIDE 18

Previous methods

Feature combinations

More accurate than single features One of the best: Geometric method (time + lexical)

[Gayo-Avello, 2009]

Shortcomings

All features evaluated simultaneously → runtime Geometric method ignores semantics → accuracy

Examples

Substring test suffices. football football lisbon Geometric method fails. benfica vs sporting derby eterno

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 11

slide-19
SLIDE 19

Our previous cascading method . . .

[Hagen et al., 2011]

source: [http://wp.ltchambon.com/wp-content/uploads/2010/09/Cascade-de-Tufs-Baume-les-messieurs-Jura.jpg]

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 12

slide-20
SLIDE 20

. . . well . . . it looked more like this

[Hagen et al., 2011]

source: [http://www.solarshop.com/solarpix/Solar Cascade 4 Tier GreenL.jpg]

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 13

slide-21
SLIDE 21

. . . well . . . it looked more like this

[Hagen et al., 2011]

source: [http://www.solarshop.com/solarpix/Solar Cascade 4 Tier GreenL.jpg]

Step 1: Subset test ց Step 2: Geometric method ց Step 3: ESA similarity ց Step 4: Search results

Basic idea

Increased feature cost (runtime) from step to step. Expensive features only if previous steps“unreliable.”

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 13

slide-22
SLIDE 22

Our improved cascade

source: [http://www.solarshop.com/solarpix/Solar Cascade 4 Tier GreenL.jpg]

Basic idea still

Cheap features first. Step 0: Time gaps ց Step 1: Substring test ց Step 2: Lexical similarity ց Step 3: ESA similarity ւ Step 4: LOD similarity ւ Step 5: Search results

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 13

slide-23
SLIDE 23

Step 0: Time gaps

(gaps ≤ 90 minutes)

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — istanbul archeology 2013-04-21 18:31:21 weather new york 2013-04-21 18:45:23 constantinople 2013-04-21 18:45:36 footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — derby eterno 2013-04-21 22:42:48 constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 14

slide-24
SLIDE 24

Step 1: Substring test

(one query contained in the other)

Query Time ancient turkey 2013-04-20 20:02:44 — — — — — — — — — — — — — history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — istanbul archeology 2013-04-21 18:31:21 — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 — — — — — — — — — — — — — football lisbon 2013-04-21 19:14:11 — — — — — — — — — — — — — benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 15

slide-25
SLIDE 25

Step 2: Lexical similarity

(cosine similarity of char 3-/4-grams)

Query Time ancient turkey 2013-04-20 20:02:44 — — — — — — — — — — — — — history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — istanbul archeology 2013-04-21 18:31:21 — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 — — — — — — — — — — — — — benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 16

slide-26
SLIDE 26

Step 3: ESA similarity

[Gabrilovich and Markovitch, 2007]

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — istanbul archeology 2013-04-21 18:31:21 — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 — — — — — — — — — — — — — benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 17

slide-27
SLIDE 27

Step 4: LOD similarity

(heaviest 2-step path in DBpedia)

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — istanbul archeology 2013-04-21 18:31:21 — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 18

slide-28
SLIDE 28

Step 5: Search results

(shared top-10 result)

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — istanbul archeology 2013-04-21 18:31:21 — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 19

slide-29
SLIDE 29

Computed logical sessions

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — istanbul archeology 2013-04-21 18:31:21 — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 20

slide-30
SLIDE 30

Unidentified mission connections

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 — — — — — — — — — — — — — ◭ istanbul archeology 2013-04-21 12:02:54 — — — — — — — — — — — — — ◭ istanbul archeology 2013-04-21 18:31:21   — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36          — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 — — — — — — — — — — — — — ◭ derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 20

slide-31
SLIDE 31

Mission detection

Idea

Run the cascade twice:

1 Session detection on query level 2 Mission detection on session level Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 21

slide-32
SLIDE 32

Example: You are here!

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 istanbul archeology 2013-04-21 12:02:54 istanbul archeology 2013-04-21 18:31:21   — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — ◮ constantinople 2013-04-21 23:09:02 ◭ constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 22

slide-33
SLIDE 33

Example: You are here!

(ignore intermediate queries)

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 istanbul archeology 2013-04-21 12:02:54 istanbul archeology 2013-04-21 18:31:21   — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36 — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — ◮ constantinople 2013-04-21 23:09:02 ◭ constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 22

slide-34
SLIDE 34

Step 1 of the cascade does the job

(substring test)

Query Time ancient turkey 2013-04-20 20:02:44 history istanbul 2013-04-20 20:24:17 istanbul archeology 2013-04-21 12:02:54 istanbul archeology 2013-04-21 18:31:21   — — — — — — — — — — — — — weather new york 2013-04-21 18:45:23 — — — — — — — — — — — — — constantinople 2013-04-21 18:45:36        — — — — — — — — — — — — — footbal lisbon 2013-04-21 19:14:01 football lisbon 2013-04-21 19:14:11 benfica vs sporting 2013-04-21 20:23:04 derby eterno 2013-04-21 22:42:48 — — — — — — — — — — — — — constantinople 2013-04-21 23:09:02 constantinople 2013-04-21 23:27:38

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 22

slide-35
SLIDE 35

How good does it work?

Accuracy Runtime

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 23

slide-36
SLIDE 36

Available evaluation corpora

Gayo-Avello’s corpus

(AOL log, 1 annotator)

11 500 queries But: empty queries, order changed 215 users But: many with ≤ 3 queries 2.7 queries per session But: no mission annotation

Lucchese et al.’s corpus

(AOL log, 1 annotator)

1500 queries But: 97% of queries dropped 13 users

Our new corpus

(basis: Gayo-Avello, 2 annotators)

8800 queries Empty queries removed 127 users Users with ≤ 3 queries removed 6.5 queries per mission

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 24

slide-37
SLIDE 37

Available evaluation corpora

Gayo-Avello’s corpus

(AOL log, 1 annotator)

11 500 queries But: empty queries, order changed 215 users But: many with ≤ 3 queries 2.7 queries per session But: no mission annotation

Lucchese et al.’s corpus

(AOL log, 1 annotator)

1500 queries But: 97% of queries dropped 13 users

Our new corpus

(basis: Gayo-Avello, 2 annotators)

8800 queries Empty queries removed 127 users Users with ≤ 3 queries removed 6.5 queries per mission

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 24

slide-38
SLIDE 38

Available evaluation corpora

Gayo-Avello’s corpus

(AOL log, 1 annotator)

11 500 queries But: empty queries, order changed 215 users But: many with ≤ 3 queries 2.7 queries per session But: no mission annotation

Lucchese et al.’s corpus

(AOL log, 1 annotator)

1500 queries But: 97% of queries dropped 13 users

Our new corpus

(basis: Gayo-Avello, 2 annotators)

8800 queries Empty queries removed 127 users Users with ≤ 3 queries removed 6.5 queries per mission

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 24

slide-39
SLIDE 39

Session detection accuracy

F-Measure on our corpus

(6630 queries, 25 % training)

Geometric method 0.821 Original cascade 0.853 Improved cascade 0.946

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 25

slide-40
SLIDE 40

Session detection accuracy

F-Measure on our corpus

(6630 queries, 25 % training)

Geometric method 0.821 Original cascade 0.853 Improved cascade 0.946

Performance per step

decides F-Measure time Step 0 23.87% 0.807 0.033 ms Step 1 48.72% 0.845 0.002 ms Step 2 13.28% 0.925 0.178 ms Step 3 0.60% 0.930 0.237 ms Step 4 0.11% 0.930 12.770 ms Step 5 2.03% 0.946 13.359 ms

Remark: Without Steps 4–5 about 8500 queries per second!

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 25

slide-41
SLIDE 41

Mission detection accuracy

Accuracy on our corpus

(6630 queries, 25 % training)

865 continuations identified (269 missed: 157 horizon effects) 307 sessions wrongly assigned a continuation

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 26

slide-42
SLIDE 42

Mission detection accuracy

Accuracy on our corpus

(6630 queries, 25 % training)

865 continuations identified (269 missed: 157 horizon effects) 307 sessions wrongly assigned a continuation

Without semantic steps!

807 continuations identified 113 sessions wrongly assigned a continuation

Remark: Missions often picked up with identical query.

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 26

slide-43
SLIDE 43

Mission detection accuracy

Accuracy on our corpus

(6630 queries, 25 % training)

865 continuations identified (269 missed: 157 horizon effects) 307 sessions wrongly assigned a continuation

Without semantic steps!

807 continuations identified 113 sessions wrongly assigned a continuation

Remark: Missions often picked up with identical query.

Observations

Session detection benefits from semantics Mission detection better without semantics

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 26

slide-44
SLIDE 44

Almost the end: The take-home messages!

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 27

slide-45
SLIDE 45

What we have done

Results

Improved cascade Cheap features first Applicable to missions Semantics rather costly LOD not really useful yet Large mission corpus

Future Work

Speed up semantics Prune LOD graph Wikipedia index Search results

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 28

slide-46
SLIDE 46

What we have (not) done

Results

Improved cascade Cheap features first Applicable to missions Semantics rather costly LOD not really useful yet Large mission corpus

Future Work

Speed up semantics Prune LOD graph Wikipedia index Search results

Hagen, Gomoll, Beyer, Stein From Search Session Detection to Search Mission Detection 28

slide-47
SLIDE 47

What we have (not) done

Results

Improved cascade Cheap features first Applicable to missions Semantics rather costly LOD not really useful yet Large mission corpus

Future Work

Speed up semantics Prune LOD graph Wikipedia index Search results

Thank you

  • Hagen, Gomoll, Beyer, Stein

From Search Session Detection to Search Mission Detection 28