Download presentation
Presentation is loading. Please wait.
Published bySusanti Widjaja Modified over 6 years ago
1
A Machine Learning Approach to Coreference Resolution of Noun Phrases
12/4/2018
2
Outline The notion of Coreference A Machine learning approach
Extraction of Markables Extracted Features Training Data Classifier Construction Testing Result analysis 12/4/2018 2
3
The notion of Coreference Definition
The grammatical relation between two words that have a common referent (WordNet) In linguistics, Coreference is the phenomenon where two expressions in an utterance both refer to the same thing (Wikipedia) A Coreference resolution process output pairs of noun phrases (coreferences) 12/4/2018 5
4
John arrived. He looked tired.
A typical example of anaphoric expression are pronouns such as he in the text John arrived. He looked tired. 12/4/2018
5
The notion of Coreference Usage
Information Retrieval Question answering Shallow parsing And more… 12/4/2018 6
6
The notion of Coreference Example
(Eastern Air)a1 Proposes (Date For Talks on ((Pay)c1-Cut)d1 Plan)b1. (Eastern Airlines)a2 executives noticed (union)e1 leaders that the carrier wishes to discuss selective ((wage)c2 reductions)d2 on (Feb. 3)b2. ((Union)e2 representatives who could be reached)f1 said (they)f2 hadn’t decided whether (they)f3 would respond. By proposing (a meeting date)b3, (Eastern)a3 moved one step closer toward reopening current high-cost contract agreements with ((its)a4 unions)e3. 12/4/2018 10
7
Outline The notion of Coreference A Machine learning approach
Extraction of Markables Extracted Features Training Data Classifier Construction Testing Result analysis 12/4/2018 11
8
Extraction of Markables Preprocessing
12/4/2018 14
9
Outline The notion of Coreference A Machine learning approach
Extraction of Markables Extracted Features Training Data Classifier Construction Testing Result analysis 12/4/2018 15
10
Extracted Features 12 suggested features for markables pairs
Distance (How far the two markables are) i/j is a Pronoun (he, him, himself, his…) String match feature (base strings match) j is a Definite noun phrase (the) j is a Demonstrative noun phrase (this, that, these, those) Number agreement (i and j are both plural/singular) 12/4/2018 19
11
Extracted Features cont.
12 suggested features for markables pairs Semantic class agreement (i and j are of the same WordNet class) Gender agreement (i and j are of the same gender) Both proper name (i and j are proper names) Alias (i and j match. e.g. 1st jan and for dates) Apposition (j is an apposition of i. e.g. Mubarak, Egypt's president) 12/4/2018 22
12
Extracted Features Example
12/4/2018 25
13
Outline The notion of Coreference A Machine learning approach
Extraction of Markables Extracted Features Training Data Classifier Construction Testing Result analysis 12/4/2018 26
14
Training Data MUC-6/7 conference corpora Creating positive examples
Creating negative examples 12/4/2018 27
15
Outline The notion of Coreference A Machine learning approach
Extraction of Markables Extracted Features Training Data Classifier Construction Testing Result analysis 12/4/2018 28
16
Classifier Construction
Classifier types: neural network, SVM, KNN, Decision tree (selected) Decision tree structure: Each node of the tree is a question about one of the features. According to the answer, the path is chosen. When a leaf is reached, its label is returned. 12/4/2018 31
17
Outline The notion of Coreference A Machine learning approach
Extraction of Markables Extracted Features Training Data Classifier Construction Testing Result analysis 12/4/2018 32
18
Testing After a classifier is built, it is tested against a pre-annotated example. Then, the results are compared with the “true” anotation. The measures are Recall (how many of the real coreferences were returned) and Precision (how many of the coreferences returned, are true ones). 12/4/2018 34
19
Testing Example (Ms. Washington)73's candidacy is being championed by (several powerful lawmakers)74 including ((her)76 boss)75, Chairman John Dingell)77 (D., (Mich.)78) of (the House Energy and Commerce Committee)79. (She)80 currently is (a counsel)81 to (the committee)82. (Ms. Washington)83 and (Mr. DingeU)84 have been considered (allies)85 of (the (securities)87 exchanges)86, while (banks)88 and ((futures)90 exchanges)89 have often fought with (them)91. 12/4/2018 37
20
Testing Example Classification
12/4/2018 40
21
Outline The notion of Coreference A Machine learning approach
Extraction of Markables Extracted Features Training Data Classifier Construction Testing Result analysis 12/4/2018 41
22
Result analysis Decision Tree
12/4/2018 44
23
Result analysis Recall & Precision
12/4/2018 45
24
Result analysis misconceptions
The Decision tree shows that only 8 features are being used. When used with 3 features (alias, apposition, string match) the scores (f-measure) were only 1-2.3% worse then when used with all of them only 3 features really contribute. 12/4/2018 47
25
Result analysis misconceptions – cont.
66.3% of the positive results followed the path of the first tree node – string matching. 70% of the total precision problems are caused by string matching: Directors also approved the election of Allan Laufgraben, 54 years old, as president and (chief executive officer)1 and Peter A. Left, 43, as chief operating officer. Milton Petrie, 90-year-old chairman, president and (chief executive officer)2 since the company was founded in 1932, will continue as chairman. 12/4/2018 49
26
Result analysis conclusions
The great achievement according to the authors – the fact that a learning method, over “shallow features” achieves the same performance as top-of-the-art systems. A HUGE majority of the results (and errors) is determined by 1-3 features. Learning over such a small amount of features isn’t really learning. So the achievement does not look like one. Not to me, though. 12/4/2018 52
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.