text stringlengths 49 577 | label stringclasses 7
values | metadata list |
|---|---|---|
Over the last few years dramatic improvements have been made , and a number of comparative evaluations have shown , that [[ SMT ]] gives competitive results to << rule-based translation systems >> , requiring significantly less development time . | COMPARE | [
21,
21,
26,
28
] |
This is particularly important when building [[ translation systems ]] for << new language pairs >> or new domains . | USED-FOR | [
6,
7,
9,
11
] |
This is particularly important when building [[ translation systems ]] for new language pairs or << new domains >> . | USED-FOR | [
6,
7,
13,
14
] |
This is particularly important when building translation systems for [[ new language pairs ]] or << new domains >> . | CONJUNCTION | [
9,
11,
13,
14
] |
[[ STTK ]] , a << statistical machine translation tool kit >> , will be introduced and used to build a working translation system . | HYPONYM-OF | [
0,
0,
3,
7
] |
[[ STTK ]] , a statistical machine translation tool kit , will be introduced and used to build a working << translation system >> . | USED-FOR | [
0,
0,
18,
19
] |
[[ STTK ]] has been developed by the presenter and co-workers over a number of years and is currently used as the basis of CMU 's << SMT system >> . | USED-FOR | [
0,
0,
24,
25
] |
[[ It ]] has also successfully been coupled with << rule-based and example based machine translation modules >> to build a multi engine machine translation system . | CONJUNCTION | [
0,
0,
7,
13
] |
[[ It ]] has also successfully been coupled with rule-based and example based machine translation modules to build a << multi engine machine translation system >> . | USED-FOR | [
0,
0,
17,
21
] |
It has also successfully been coupled with [[ rule-based and example based machine translation modules ]] to build a << multi engine machine translation system >> . | USED-FOR | [
7,
13,
17,
21
] |
This paper presents an [[ unsupervised learning approach ]] to building a << non-English -LRB- Arabic -RRB- stemmer >> . | USED-FOR | [
4,
6,
10,
14
] |
The << stemming model >> is based on [[ statistical machine translation ]] and it uses an English stemmer and a small -LRB- 10K sentences -RRB- parallel corpus as its sole training resources . | USED-FOR | [
6,
8,
1,
2
] |
The stemming model is based on statistical machine translation and << it >> uses an [[ English stemmer ]] and a small -LRB- 10K sentences -RRB- parallel corpus as its sole training resources . | USED-FOR | [
13,
14,
10,
10
] |
The stemming model is based on statistical machine translation and << it >> uses an English stemmer and a small -LRB- 10K sentences -RRB- [[ parallel corpus ]] as its sole training resources . | USED-FOR | [
22,
23,
10,
10
] |
[[ Monolingual , unannotated text ]] can be used to further improve the << stemmer >> by allowing it to adapt to a desired domain or genre . | USED-FOR | [
0,
3,
11,
11
] |
Our [[ resource-frugal approach ]] results in 87.5 % agreement with a state of the art , proprietary << Arabic stemmer >> built using rules , affix lists , and human annotated text , in addition to an unsupervised component . | COMPARE | [
1,
2,
16,
17
] |
Our << resource-frugal approach >> results in 87.5 % [[ agreement ]] with a state of the art , proprietary Arabic stemmer built using rules , affix lists , and human annotated text , in addition to an unsupervised component . | EVALUATE-FOR | [
7,
7,
1,
2
] |
Our resource-frugal approach results in 87.5 % [[ agreement ]] with a state of the art , proprietary << Arabic stemmer >> built using rules , affix lists , and human annotated text , in addition to an unsupervised component . | EVALUATE-FOR | [
7,
7,
16,
17
] |
Our resource-frugal approach results in 87.5 % agreement with a state of the art , proprietary << Arabic stemmer >> built using [[ rules ]] , affix lists , and human annotated text , in addition to an unsupervised component . | USED-FOR | [
20,
20,
16,
17
] |
Our resource-frugal approach results in 87.5 % agreement with a state of the art , proprietary Arabic stemmer built using [[ rules ]] , << affix lists >> , and human annotated text , in addition to an unsupervised component . | CONJUNCTION | [
20,
20,
22,
23
] |
Our resource-frugal approach results in 87.5 % agreement with a state of the art , proprietary << Arabic stemmer >> built using rules , [[ affix lists ]] , and human annotated text , in addition to an unsupervised component . | USED-FOR | [
22,
23,
16,
17
] |
Our resource-frugal approach results in 87.5 % agreement with a state of the art , proprietary Arabic stemmer built using rules , [[ affix lists ]] , and << human annotated text >> , in addition to an unsupervised component . | CONJUNCTION | [
22,
23,
26,
28
] |
Our resource-frugal approach results in 87.5 % agreement with a state of the art , proprietary << Arabic stemmer >> built using rules , affix lists , and [[ human annotated text ]] , in addition to an unsupervised component . | USED-FOR | [
26,
28,
16,
17
] |
Our resource-frugal approach results in 87.5 % agreement with a state of the art , proprietary Arabic stemmer built using rules , affix lists , and [[ human annotated text ]] , in addition to an << unsupervised component >> . | CONJUNCTION | [
26,
28,
34,
35
] |
Our resource-frugal approach results in 87.5 % agreement with a state of the art , proprietary << Arabic stemmer >> built using rules , affix lists , and human annotated text , in addition to an [[ unsupervised component ]] . | USED-FOR | [
34,
35,
16,
17
] |
<< Task-based evaluation >> using [[ Arabic information retrieval ]] indicates an improvement of 22-38 % in average precision over unstemmed text , and 96 % of the performance of the proprietary stemmer above . | USED-FOR | [
3,
5,
0,
1
] |
<< Task-based evaluation >> using Arabic information retrieval indicates an improvement of 22-38 % in [[ average precision ]] over unstemmed text , and 96 % of the performance of the proprietary stemmer above . | EVALUATE-FOR | [
13,
14,
0,
1
] |
Task-based evaluation using Arabic information retrieval indicates an improvement of 22-38 % in [[ average precision ]] over << unstemmed text >> , and 96 % of the performance of the proprietary stemmer above . | EVALUATE-FOR | [
13,
14,
16,
17
] |
The paper assesses the capability of an [[ HMM-based TTS system ]] to produce << German speech >> . | USED-FOR | [
7,
9,
12,
13
] |
In addition , the [[ system ]] is adapted to a small set of << football announcements >> , in an exploratory attempt to synthe-sise expressive speech . | USED-FOR | [
4,
4,
12,
13
] |
In addition , the [[ system ]] is adapted to a small set of football announcements , in an exploratory attempt to synthe-sise << expressive speech >> . | USED-FOR | [
4,
4,
21,
22
] |
We conclude that the [[ HMMs ]] are able to produce highly << intelligible neutral German speech >> , with a stable quality , and that the expressivity is partially captured in spite of the small size of the football dataset . | USED-FOR | [
4,
4,
10,
13
] |
Furthermore , in contrast to the approach of Dalrymple et al. -LSB- 1991 -RSB- , the treatment directly encodes the intuitive distinction between [[ full NPs ]] and the << referential elements >> that corefer with them through what we term role linking . | CONJUNCTION | [
23,
24,
27,
28
] |
Finally , the [[ analysis ]] extends directly to other << discourse copying phenomena >> . | USED-FOR | [
3,
3,
8,
10
] |
How to obtain [[ hierarchical relations ]] -LRB- e.g. superordinate - hyponym relation , synonym relation -RRB- is one of the most important problems for << thesaurus construction >> . | PART-OF | [
3,
4,
23,
24
] |
How to obtain << hierarchical relations >> -LRB- e.g. [[ superordinate - hyponym relation ]] , synonym relation -RRB- is one of the most important problems for thesaurus construction . | HYPONYM-OF | [
7,
10,
3,
4
] |
How to obtain hierarchical relations -LRB- e.g. [[ superordinate - hyponym relation ]] , << synonym relation >> -RRB- is one of the most important problems for thesaurus construction . | CONJUNCTION | [
7,
10,
12,
13
] |
How to obtain << hierarchical relations >> -LRB- e.g. superordinate - hyponym relation , [[ synonym relation ]] -RRB- is one of the most important problems for thesaurus construction . | HYPONYM-OF | [
12,
13,
3,
4
] |
A pilot system for extracting these << relations >> automatically from an ordinary [[ Japanese language dictionary ]] -LRB- Shinmeikai Kokugojiten , published by Sansei-do , in machine readable form -RRB- is given . | USED-FOR | [
11,
13,
6,
6
] |
The << features >> of the [[ definition sentences ]] in the dictionary , the mechanical extraction of the hierarchical relations and the estimation of the results are discussed . | USED-FOR | [
4,
5,
1,
1
] |
The features of the [[ definition sentences ]] in the << dictionary >> , the mechanical extraction of the hierarchical relations and the estimation of the results are discussed . | PART-OF | [
4,
5,
8,
8
] |
This is evident most compellingly by the very low [[ recognition rate ]] of all existing << face recognition systems >> when applied to live CCTV camera input . | EVALUATE-FOR | [
9,
10,
14,
16
] |
This is evident most compellingly by the very low recognition rate of all existing << face recognition systems >> when applied to [[ live CCTV camera input ]] . | USED-FOR | [
20,
23,
14,
16
] |
In this paper , we present a [[ Bayesian framework ]] to perform multi-modal -LRB- such as variations in viewpoint and illumination -RRB- << face image super-resolution >> for recognition in tensor space . | USED-FOR | [
7,
8,
21,
23
] |
In this paper , we present a Bayesian framework to perform multi-modal -LRB- such as variations in [[ viewpoint ]] and << illumination >> -RRB- face image super-resolution for recognition in tensor space . | CONJUNCTION | [
17,
17,
19,
19
] |
In this paper , we present a Bayesian framework to perform multi-modal -LRB- such as variations in viewpoint and illumination -RRB- [[ face image super-resolution ]] for << recognition >> in tensor space . | USED-FOR | [
21,
23,
25,
25
] |
In this paper , we present a Bayesian framework to perform multi-modal -LRB- such as variations in viewpoint and illumination -RRB- face image super-resolution for << recognition >> in [[ tensor space ]] . | FEATURE-OF | [
27,
28,
25,
25
] |
Given a [[ single modal low-resolution face image ]] , we benefit from the multiple factor interactions of training tensor , and super-resolve its << high-resolution reconstructions >> across different modalities for face recognition . | USED-FOR | [
2,
6,
22,
23
] |
Given a single modal low-resolution face image , we benefit from the [[ multiple factor interactions of training tensor ]] , and super-resolve its << high-resolution reconstructions >> across different modalities for face recognition . | USED-FOR | [
12,
17,
22,
23
] |
Given a single modal low-resolution face image , we benefit from the multiple factor interactions of training tensor , and super-resolve its [[ high-resolution reconstructions ]] across different modalities for << face recognition >> . | USED-FOR | [
22,
23,
28,
29
] |
Given a single modal low-resolution face image , we benefit from the multiple factor interactions of training tensor , and super-resolve its << high-resolution reconstructions >> across different [[ modalities ]] for face recognition . | FEATURE-OF | [
26,
26,
22,
23
] |
Instead of performing << pixel-domain super-resolution and recognition >> independently as two separate sequential processes , we integrate the tasks of [[ super-resolution ]] and recognition by directly computing a maximum likelihood identity parameter vector in high-resolution tensor space for recognition . | HYPONYM-OF | [
19,
19,
3,
6
] |
Instead of performing pixel-domain super-resolution and recognition independently as two separate sequential processes , we integrate the tasks of [[ super-resolution ]] and << recognition >> by directly computing a maximum likelihood identity parameter vector in high-resolution tensor space for recognition . | CONJUNCTION | [
19,
19,
21,
21
] |
Instead of performing << pixel-domain super-resolution and recognition >> independently as two separate sequential processes , we integrate the tasks of super-resolution and [[ recognition ]] by directly computing a maximum likelihood identity parameter vector in high-resolution tensor space for recognition . | HYPONYM-OF | [
21,
21,
3,
6
] |
Instead of performing pixel-domain super-resolution and recognition independently as two separate sequential processes , we integrate the tasks of << super-resolution >> and recognition by directly computing a [[ maximum likelihood identity parameter vector ]] in high-resolution tensor space for recognition . | USED-FOR | [
26,
30,
19,
19
] |
Instead of performing pixel-domain super-resolution and recognition independently as two separate sequential processes , we integrate the tasks of super-resolution and << recognition >> by directly computing a [[ maximum likelihood identity parameter vector ]] in high-resolution tensor space for recognition . | USED-FOR | [
26,
30,
21,
21
] |
Instead of performing pixel-domain super-resolution and recognition independently as two separate sequential processes , we integrate the tasks of super-resolution and recognition by directly computing a [[ maximum likelihood identity parameter vector ]] in high-resolution tensor space for << recognition >> . | USED-FOR | [
26,
30,
36,
36
] |
Instead of performing pixel-domain super-resolution and recognition independently as two separate sequential processes , we integrate the tasks of super-resolution and recognition by directly computing a << maximum likelihood identity parameter vector >> in [[ high-resolution tensor space ]] for recognition . | FEATURE-OF | [
32,
34,
26,
30
] |
We show results from << multi-modal super-resolution and face recognition >> experiments across different imaging modalities , using [[ low-resolution images ]] as testing inputs and demonstrate improved recognition rates over standard tensorface and eigenface representations . | USED-FOR | [
16,
17,
4,
8
] |
We show results from << multi-modal super-resolution and face recognition >> experiments across different imaging modalities , using low-resolution images as testing inputs and demonstrate improved [[ recognition rates ]] over standard tensorface and eigenface representations . | EVALUATE-FOR | [
24,
25,
4,
8
] |
We show results from multi-modal super-resolution and face recognition experiments across different imaging modalities , using low-resolution images as testing inputs and demonstrate improved [[ recognition rates ]] over standard << tensorface and eigenface representations >> . | EVALUATE-FOR | [
24,
25,
28,
31
] |
In this paper , we describe a [[ phrase-based unigram model ]] for << statistical machine translation >> that uses a much simpler set of model parameters than similar phrase-based models . | USED-FOR | [
7,
9,
11,
13
] |
In this paper , we describe a [[ phrase-based unigram model ]] for statistical machine translation that uses a much simpler set of model parameters than similar << phrase-based models >> . | COMPARE | [
7,
9,
25,
26
] |
In this paper , we describe a << phrase-based unigram model >> for statistical machine translation that uses a much simpler set of [[ model parameters ]] than similar phrase-based models . | USED-FOR | [
21,
22,
7,
9
] |
During << decoding >> , we use a [[ block unigram model ]] and a word-based trigram language model . | USED-FOR | [
6,
8,
1,
1
] |
During << decoding >> , we use a block unigram model and a [[ word-based trigram language model ]] . | USED-FOR | [
11,
14,
1,
1
] |
During decoding , we use a << block unigram model >> and a [[ word-based trigram language model ]] . | CONJUNCTION | [
11,
14,
6,
8
] |
During training , the << blocks >> are learned from [[ source interval projections ]] using an underlying word alignment . | USED-FOR | [
8,
10,
4,
4
] |
During training , the blocks are learned from << source interval projections >> using an underlying [[ word alignment ]] . | USED-FOR | [
14,
15,
8,
10
] |
We show experimental results on << block selection criteria >> based on [[ unigram counts ]] and phrase length . | USED-FOR | [
10,
11,
5,
7
] |
We show experimental results on block selection criteria based on [[ unigram counts ]] and << phrase length >> . | CONJUNCTION | [
10,
11,
13,
14
] |
We show experimental results on << block selection criteria >> based on unigram counts and [[ phrase length ]] . | USED-FOR | [
13,
14,
5,
7
] |
This paper develops a new [[ approach ]] for extremely << fast detection >> in domains where the distribution of positive and negative examples is highly skewed -LRB- e.g. face detection or database retrieval -RRB- . | USED-FOR | [
5,
5,
8,
9
] |
This paper develops a new approach for extremely fast detection in domains where the distribution of positive and negative examples is highly skewed -LRB- e.g. [[ face detection ]] or << database retrieval >> -RRB- . | CONJUNCTION | [
25,
26,
28,
29
] |
In such domains a [[ cascade of simple classifiers ]] each trained to achieve high detection rates and modest false positive rates can yield a final << detector >> with many desirable features : including high detection rates , very low false positive rates , and fast performance . | USED-FOR | [
4,
7,
24,
24
] |
In such domains a cascade of simple << classifiers >> each trained to achieve high [[ detection rates ]] and modest false positive rates can yield a final detector with many desirable features : including high detection rates , very low false positive rates , and fast performance . | EVALUATE-FOR | [
13,
14,
7,
7
] |
In such domains a cascade of simple classifiers each trained to achieve high [[ detection rates ]] and << modest false positive rates >> can yield a final detector with many desirable features : including high detection rates , very low false positive rates , and fast performance . | CONJUNCTION | [
13,
14,
16,
19
] |
In such domains a cascade of simple << classifiers >> each trained to achieve high detection rates and [[ modest false positive rates ]] can yield a final detector with many desirable features : including high detection rates , very low false positive rates , and fast performance . | EVALUATE-FOR | [
16,
19,
7,
7
] |
In such domains a cascade of simple classifiers each trained to achieve high detection rates and modest false positive rates can yield a final << detector >> with many desirable [[ features ]] : including high detection rates , very low false positive rates , and fast performance . | FEATURE-OF | [
28,
28,
24,
24
] |
Achieving extremely high [[ detection rates ]] , rather than << low error >> , is not a task typically addressed by machine learning algorithms . | COMPARE | [
3,
4,
8,
9
] |
We propose a new variant of [[ AdaBoost ]] as a mechanism for training the simple << classifiers >> used in the cascade . | USED-FOR | [
6,
6,
14,
14
] |
We propose a new variant of AdaBoost as a mechanism for training the simple [[ classifiers ]] used in the << cascade >> . | USED-FOR | [
14,
14,
18,
18
] |
Experimental results in the domain of << face detection >> show the [[ training algorithm ]] yields significant improvements in performance over conventional AdaBoost . | USED-FOR | [
10,
11,
6,
7
] |
Experimental results in the domain of << face detection >> show the training algorithm yields significant improvements in performance over conventional [[ AdaBoost ]] . | USED-FOR | [
19,
19,
6,
7
] |
Experimental results in the domain of face detection show the << training algorithm >> yields significant improvements in performance over conventional [[ AdaBoost ]] . | COMPARE | [
19,
19,
10,
11
] |
The final face detection system can process 15 frames per second , achieves over 90 % [[ detection ]] , and a << false positive rate >> of 1 in a 1,000,000 . | CONJUNCTION | [
16,
16,
20,
22
] |
This paper proposes a [[ method ]] for learning << joint embed-dings of images and text >> using a two-branch neural network with multiple layers of linear projections followed by nonlinearities . | USED-FOR | [
4,
4,
7,
12
] |
This paper proposes a << method >> for learning joint embed-dings of images and text using a [[ two-branch neural network ]] with multiple layers of linear projections followed by nonlinearities . | USED-FOR | [
15,
17,
4,
4
] |
This paper proposes a method for learning joint embed-dings of images and text using a << two-branch neural network >> with [[ multiple layers of linear projections ]] followed by nonlinearities . | PART-OF | [
19,
23,
15,
17
] |
This paper proposes a method for learning joint embed-dings of images and text using a two-branch neural network with [[ multiple layers of linear projections ]] followed by << nonlinearities >> . | CONJUNCTION | [
19,
23,
26,
26
] |
This paper proposes a method for learning joint embed-dings of images and text using a << two-branch neural network >> with multiple layers of linear projections followed by [[ nonlinearities ]] . | PART-OF | [
26,
26,
15,
17
] |
The << network >> is trained using a [[ large-margin objective ]] that combines cross-view ranking constraints with within-view neighborhood structure preservation constraints inspired by metric learning literature . | USED-FOR | [
6,
7,
1,
1
] |
The network is trained using a << large-margin objective >> that combines [[ cross-view ranking constraints ]] with within-view neighborhood structure preservation constraints inspired by metric learning literature . | FEATURE-OF | [
10,
12,
6,
7
] |
The network is trained using a large-margin objective that combines [[ cross-view ranking constraints ]] with << within-view neighborhood structure preservation constraints >> inspired by metric learning literature . | CONJUNCTION | [
10,
12,
14,
18
] |
The network is trained using a << large-margin objective >> that combines cross-view ranking constraints with [[ within-view neighborhood structure preservation constraints ]] inspired by metric learning literature . | FEATURE-OF | [
14,
18,
6,
7
] |
Extensive experiments show that our << approach >> gains significant improvements in [[ accuracy ]] for image-to-text and text-to-image retrieval . | EVALUATE-FOR | [
10,
10,
5,
5
] |
Extensive experiments show that our << approach >> gains significant improvements in accuracy for [[ image-to-text and text-to-image retrieval ]] . | EVALUATE-FOR | [
12,
15,
5,
5
] |
Our << method >> achieves new state-of-the-art results on the [[ Flickr30K and MSCOCO image-sentence datasets ]] and shows promise on the new task of phrase lo-calization on the Flickr30K Entities dataset . | EVALUATE-FOR | [
8,
12,
1,
1
] |
Our << method >> achieves new state-of-the-art results on the Flickr30K and MSCOCO image-sentence datasets and shows promise on the new task of [[ phrase lo-calization ]] on the Flickr30K Entities dataset . | EVALUATE-FOR | [
21,
22,
1,
1
] |
Our method achieves new state-of-the-art results on the Flickr30K and MSCOCO image-sentence datasets and shows promise on the new task of << phrase lo-calization >> on the [[ Flickr30K Entities dataset ]] . | USED-FOR | [
25,
27,
21,
22
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.