This is a table of type bigram and their frequencies. Use it to search & browse the list to learn more about your study carrel.
bigram | frequency |
---|---|
information retrieval | 176 |
seed words | 73 |
social media | 67 |
word embeddings | 63 |
neural networks | 60 |
query terms | 59 |
schema labels | 59 |
question answering | 51 |
sentiment analysis | 49 |
query term | 45 |
cord uid | 43 |
retrieval doi | 43 |
neural network | 43 |
doc id | 43 |
schema label | 43 |
recommender systems | 41 |
training data | 39 |
word embedding | 39 |
data sets | 38 |
retrieval models | 38 |
clef ehealth | 36 |
bantu languages | 35 |
attention layer | 34 |
deep learning | 34 |
contact recommendation | 33 |
natural language | 33 |
proposed model | 32 |
attention mechanism | 32 |
test set | 31 |
training set | 30 |
embedding space | 29 |
keyphrase extraction | 29 |
posting lists | 28 |
data set | 28 |
knowledge graph | 28 |
topic modeling | 27 |
convolutional neural | 26 |
local context | 26 |
language models | 26 |
review embedding | 25 |
sentiment lexicon | 25 |
search engines | 25 |
language model | 25 |
matrix factorization | 24 |
semantic matching | 24 |
web search | 24 |
modal retrieval | 24 |
query expansion | 24 |
future work | 24 |
start scenario | 23 |
text classification | 23 |
ir models | 23 |
previous work | 23 |
scoring function | 23 |
query reformulation | 23 |
publicly available | 22 |
cond gans | 22 |
information needs | 22 |
explicit features | 21 |
requirement set | 21 |
dataset retrieval | 21 |
results show | 20 |
network embedding | 20 |
search engine | 20 |
cosine similarity | 20 |
proposed approach | 20 |
semantic similarity | 20 |
review text | 20 |
online learning | 20 |
posting list | 19 |
machine translation | 19 |
product attention | 19 |
machine learning | 19 |
neural models | 19 |
vector space | 18 |
retrieval task | 18 |
word representation | 18 |
result list | 18 |
supporting information | 18 |
text preprocessing | 18 |
query reformulations | 18 |
document representations | 18 |
ranking model | 17 |
social networks | 17 |
attention weights | 17 |
learning rate | 17 |
dataset search | 17 |
gradient descent | 17 |
large number | 17 |
query claim | 17 |
missing views | 17 |
document ranking | 17 |
cold start | 17 |
candidate rankers | 17 |
knowledge base | 16 |
review helpfulness | 16 |
irony detection | 16 |
table shows | 16 |
representation learning | 16 |
warm start | 16 |
distributed representations | 16 |
relevance feedback | 15 |
collaborative filtering | 15 |
document clustering | 15 |
transfer learning | 15 |
recommendation task | 15 |
dynamic pruning | 15 |
knowledge graphs | 15 |
written languages | 15 |
random walks | 15 |
web pages | 15 |
readability analysis | 15 |
early risk | 15 |
long sessions | 14 |
entity alignment | 14 |
relevance score | 14 |
exploratory search | 14 |
contextual information | 14 |
local contexts | 14 |
word representations | 14 |
level attention | 14 |
retrieval model | 14 |
topic vectors | 14 |
offline performance | 14 |
seq seq | 14 |
information need | 14 |
document network | 14 |
product information | 14 |
based approach | 14 |
query evaluation | 14 |
word vec | 14 |
deep clustering | 14 |
based models | 13 |
two different | 13 |
infrastructural damage | 13 |
recurrent neural | 13 |
sequential question | 13 |
generated schema | 13 |
severity analysis | 13 |
chemical reaction | 13 |
factoid questions | 13 |
fossil fuels | 13 |
ranking models | 13 |
document classification | 13 |
objective function | 13 |
see sect | 13 |
evaluation lab | 13 |
long tasks | 13 |
retrieval systems | 13 |
neural ranking | 13 |
relevance scores | 13 |
location mentions | 13 |
latent semantic | 13 |
support scores | 12 |
data management | 12 |
ranked list | 12 |
suggestion mining | 12 |
term frequency | 12 |
test data | 12 |
implicit feedback | 12 |
data table | 12 |
deep neural | 12 |
counterfactual evaluation | 12 |
based query | 12 |
regression model | 12 |
english wikipedia | 12 |
based methods | 12 |
symptom relations | 12 |
previous works | 12 |
target user | 12 |
based sessions | 12 |
candidate users | 12 |
language modeling | 12 |
convolutional networks | 12 |
significantly better | 12 |
logistic regression | 12 |
word vectors | 11 |
social network | 11 |
stepwise recipe | 11 |
learning objective | 11 |
attention layers | 11 |
early detection | 11 |
better results | 11 |
level features | 11 |
max wand | 11 |
helpfulness prediction | 11 |
based approaches | 11 |
total number | 11 |
amazon data | 11 |
retrieval system | 11 |
inverted index | 11 |
probability distribution | 11 |
retrieval using | 11 |
relevance filtering | 11 |
online evaluation | 11 |
widely used | 11 |
modal similarity | 11 |
information extraction | 11 |
retrieval performance | 11 |
document length | 11 |
yelp data | 11 |
test collections | 11 |
ir axioms | 11 |
dimensional vector | 11 |
modal summarization | 11 |
candidate user | 11 |
entity recognition | 11 |
commonly used | 11 |
primary symptoms | 11 |
document retrieval | 11 |
words per | 11 |
standard deviation | 11 |
least one | 11 |
relevance judgments | 11 |
ab sign | 11 |
educational illustrations | 11 |
knowledge bases | 11 |
supervised learning | 11 |
ehealth evaluation | 11 |
experimental results | 11 |
relevant information | 10 |
click models | 10 |
query sequences | 10 |
optimization problem | 10 |
disjunctively written | 10 |
dmp common | 10 |
original paper | 10 |
useful information | 10 |
text pairs | 10 |
hoc retrieval | 10 |
sentiment attention | 10 |
evaluation metrics | 10 |
gating mechanism | 10 |
test time | 10 |
prior work | 10 |
bantu language | 10 |
query processing | 10 |
text retrieval | 10 |
dimensionality reduction | 10 |
attention network | 10 |
bias score | 10 |
premise clusters | 10 |
two tasks | 10 |
test sets | 10 |
image retrieval | 10 |
variational recurrent | 10 |
dataless text | 10 |
hidden state | 10 |
online performance | 10 |
ranking function | 10 |
shared task | 10 |
related work | 10 |
automatic identification | 10 |
statistically significant | 10 |
vrss model | 10 |
image search | 10 |
transfer layer | 10 |
link prediction | 10 |
label generation | 10 |
graph convolutional | 10 |
shared tasks | 10 |
hierarchical attention | 10 |
additional supporting | 10 |
length normalization | 10 |
health care | 10 |
embedding models | 10 |
results obtained | 9 |
learning models | 9 |
document terms | 9 |
position bias | 9 |
also report | 9 |
similarity score | 9 |
chemical compounds | 9 |
semantic analysis | 9 |
ground truth | 9 |
document representation | 9 |
similar claims | 9 |
similarity search | 9 |
risk prediction | 9 |
recipe dataset | 9 |
data sources | 9 |
factoid qa | 9 |
dual approach | 9 |
based session | 9 |
table retrieval | 9 |
based model | 9 |
speech tagging | 9 |
mixed ranking | 9 |
language dataset | 9 |
bias goggles | 9 |
medical articles | 9 |
relevant documents | 9 |
latent space | 9 |
existing lexicon | 9 |
relevance decision | 9 |
topic models | 9 |
search results | 9 |
spoken queries | 9 |
based retrieval | 8 |
previous studies | 8 |
trec robust | 8 |
support vector | 8 |
ad hoc | 8 |
term memory | 8 |
writing style | 8 |
interaction concept | 8 |
two datasets | 8 |
ranking problem | 8 |
learned representations | 8 |
proposed method | 8 |
sampling technique | 8 |
latent dirichlet | 8 |
event trigger | 8 |
node features | 8 |
current state | 8 |
model outperforms | 8 |
semantic technologies | 8 |
data tables | 8 |
counterfactual learning | 8 |
document summarization | 8 |
passage retrieval | 8 |
item embeddings | 8 |
task learning | 8 |
factoid question | 8 |
dirichlet allocation | 8 |
latent topic | 8 |
document collection | 8 |
image features | 8 |
average precision | 8 |
neural information | 8 |
model based | 8 |
search stages | 8 |
multiple modalities | 8 |
i plan | 8 |
better understand | 8 |
first one | 8 |
topic model | 8 |
fake news | 8 |
pruning strategies | 8 |
adjacency matrix | 8 |
classification problem | 8 |
partial score | 8 |
bandit feedback | 8 |
named entity | 8 |
baseline system | 8 |
language processing | 8 |
modeling assumption | 8 |
neural model | 8 |
stochastic optimization | 8 |
previous queries | 8 |
verb stemming | 8 |
term context | 8 |
chemical reactions | 8 |
political debates | 8 |
star rating | 8 |
common standards | 8 |
matching model | 8 |
negation morpheme | 8 |
reinforcement learning | 8 |
text passages | 8 |
semantically related | 8 |
even though | 8 |
retrieved documents | 8 |
classification task | 8 |
rating prediction | 8 |
ranking task | 8 |
cluster representatives | 8 |
different domain | 7 |
user embeddings | 7 |
text documents | 7 |
high quality | 7 |
real world | 7 |
data rows | 7 |
multiview approaches | 7 |
visual representation | 7 |
document language | 7 |
computer vision | 7 |
space model | 7 |
attention networks | 7 |
one language | 7 |
training process | 7 |
modeling approach | 7 |
feature vectors | 7 |
specific sentiment | 7 |
deep bidirectional | 7 |
embedding methods | 7 |
based evaluation | 7 |
best results | 7 |
weighted sum | 7 |
local contextual | 7 |
term dependencies | 7 |
twitter data | 7 |
long short | 7 |
using word | 7 |
directional lstm | 7 |
relevance ranking | 7 |
product reviews | 7 |
open access | 7 |
graph embedding | 7 |
significant differences | 7 |
embedding model | 7 |
bert sw | 7 |
semantic parsing | 7 |
full text | 7 |
global vectors | 7 |
support score | 7 |
web graph | 7 |
multifield document | 7 |
see table | 7 |
significantly outperform | 7 |
important information | 7 |
run times | 7 |
previous research | 7 |
probabilistic models | 7 |
random walk | 7 |
three languages | 7 |
conjunctively written | 7 |
von mises | 7 |
small number | 7 |
standard pipeline | 7 |
bidirectional transformers | 7 |
large corpus | 7 |
biased concept | 7 |
review volume | 7 |
common neighbors | 7 |
neural machine | 7 |
review sentiment | 7 |
two main | 7 |
attention fusion | 7 |
ablation study | 7 |
fully connected | 7 |
search library | 7 |
joint embedding | 7 |
domain suggestion | 7 |
language understanding | 7 |
attention mechanisms | 7 |
label mixed | 7 |
dataset contains | 7 |
better performance | 7 |
health information | 7 |
selection bias | 7 |
neural ir | 7 |
analysis module | 7 |
retrieval based | 7 |
disaster management | 7 |
deep relevance | 7 |
relief groups | 7 |
node vec | 7 |
retrieval evaluation | 7 |
review score | 7 |
also use | 7 |
jth sentence | 7 |
premise cluster | 7 |
labeled data | 6 |
hidden states | 6 |
different ways | 6 |
axiomatic analysis | 6 |
related products | 6 |
user length | 6 |
mixed bantu | 6 |
next step | 6 |
random forest | 6 |
clustering results | 6 |
bipartite graph | 6 |
based algorithms | 6 |
clickthrough data | 6 |
local relevance | 6 |
term discrimination | 6 |
extensively used | 6 |
supervised approaches | 6 |
research questions | 6 |
english exam | 6 |
latent representations | 6 |
image classification | 6 |
baseline models | 6 |
target product | 6 |
pmc articles | 6 |
location mention | 6 |
domain dom | 6 |
theoretical analysis | 6 |
document networks | 6 |
oltr methods | 6 |
test collection | 6 |
trained using | 6 |
image sequences | 6 |
also used | 6 |
recommender system | 6 |
multiview learning | 6 |
scoring functions | 6 |
vector representation | 6 |
field document | 6 |
graph convolutions | 6 |
integer linear | 6 |
relevance annotations | 6 |
basic retrieval | 6 |
proposed system | 6 |
relatively small | 6 |
biomedical question | 6 |
bimodal embeddings | 6 |
web domains | 6 |
node types | 6 |
text passage | 6 |
inductive document | 6 |
sentiment lexicons | 6 |
fold cross | 6 |
aligned entities | 6 |
query nodes | 6 |
query length | 6 |
grammatical structure | 6 |
take advantage | 6 |
models using | 6 |
formula search | 6 |
graded disease | 6 |
incremental approach | 6 |
artificial intelligence | 6 |
evaluation forum | 6 |
hinge loss | 6 |
generative adversarial | 6 |
relevance matching | 6 |
evaluation methodology | 6 |
query logs | 6 |
mixture model | 6 |
traditional dbgd | 6 |
transformer model | 6 |
large amount | 6 |
table reports | 6 |
adversarial networks | 6 |
modeling framework | 6 |
seed word | 6 |
linear programming | 6 |
binary classification | 6 |
avg sim | 6 |
two documents | 6 |
vec model | 6 |
word attention | 6 |
first step | 6 |
graph neural | 6 |
query log | 6 |
ilp framework | 6 |
euclidean distance | 6 |
new corpus | 6 |
mean average | 6 |
biased concepts | 6 |
criminal law | 6 |
educational images | 6 |
training samples | 6 |
preliminary evaluation | 6 |
main contributions | 6 |
min sessions | 6 |
graph gauss | 6 |
media posts | 6 |
abandon fossil | 6 |
debate portals | 6 |
score upperbound | 6 |
stochastic gradient | 6 |
using attention | 6 |
evaluation results | 6 |
three tasks | 6 |
prefs vec | 6 |
seq retrieval | 6 |
word level | 6 |
damage severity | 6 |
english dataset | 6 |
max indexes | 6 |
multimodal system | 6 |
prf mechanisms | 6 |
specific representations | 6 |
text segment | 6 |
text summarization | 6 |
fast retrieval | 6 |
search tasks | 6 |
given text | 6 |
based systems | 6 |
next query | 6 |
media platforms | 6 |
score function | 6 |
training dataset | 6 |
similarity matrix | 6 |
standard image | 6 |
structural information | 6 |
approximated review | 6 |
cambridge english | 6 |
community question | 6 |
attention model | 6 |
logging ranker | 6 |
embedding techniques | 6 |
global context | 6 |
activation function | 6 |
using standard | 6 |
text categorization | 6 |
morphological analysis | 6 |
retrieval results | 6 |
biased surfer | 6 |
event extraction | 6 |
relevance labels | 6 |
var times | 6 |
dl queries | 5 |
new documents | 5 |
claim clusters | 5 |
generated views | 5 |
user interaction | 5 |
acc ari | 5 |
done using | 5 |
user embedding | 5 |
best model | 5 |
per query | 5 |
may contain | 5 |
collection contains | 5 |
golden domains | 5 |
score upperbounds | 5 |
lingual information | 5 |
datasets used | 5 |
future research | 5 |
consider two | 5 |
noun classes | 5 |
within sessions | 5 |
search logs | 5 |
ranking score | 5 |
textual data | 5 |
noun class | 5 |
learning algorithms | 5 |
deep convolutional | 5 |
query likelihood | 5 |
overall sentiment | 5 |
different media | 5 |
manually annotated | 5 |
graded relevance | 5 |
large body | 5 |
three annotators | 5 |
answer rq | 5 |
academic research | 5 |
label classification | 5 |
different modalities | 5 |
term mismatch | 5 |
query formulation | 5 |
correlation analysis | 5 |
unsupervised cross | 5 |
original query | 5 |
baseline methods | 5 |
extraction pipeline | 5 |
multimodal deep | 5 |
canonical correlation | 5 |
disjunctive relevance | 5 |
vrss output | 5 |
learning model | 5 |
embedding vectors | 5 |
helpful reviews | 5 |
document pairs | 5 |
relevant symptoms | 5 |
complex questions | 5 |
sentence classification | 5 |
insights extraction | 5 |
image models | 5 |
obtained using | 5 |
value function | 5 |
retrieval conference | 5 |
semantic structure | 5 |
art performance | 5 |
text embedding | 5 |
docfigure dataset | 5 |
edge weight | 5 |
better understanding | 5 |
metric learning | 5 |
formula retrieval | 5 |
highly relevant | 5 |
graph attention | 5 |
document context | 5 |
class imbalance | 5 |
sessions vs | 5 |
embedding spaces | 5 |
semantic term | 5 |
risk detection | 5 |
rating scores | 5 |
using either | 5 |
contextualized word | 5 |
correct answer | 5 |
bold font | 5 |
hierarchical structure | 5 |
across different | 5 |
given query | 5 |
different languages | 5 |
perform well | 5 |
relevant passages | 5 |
standard insights | 5 |
another one | 5 |
damage assessment | 5 |
prediction time | 5 |
relief group | 5 |
simple english | 5 |
window size | 5 |
ms marco | 5 |
learning approach | 5 |
structured data | 5 |
pivoted normalization | 5 |
specific embeddings | 5 |
sampling documents | 5 |
gold standard | 5 |
expansion using | 5 |
term matching | 5 |
ranking framework | 5 |
models based | 5 |
missing view | 5 |
evaluation measures | 5 |
retrieval methods | 5 |
weight matrices | 5 |
adam optimizer | 5 |
document filtering | 5 |
text sequence | 5 |
will also | 5 |
data collected | 5 |
single document | 5 |
allows users | 5 |
cnn model | 5 |
qa pairs | 5 |
european languages | 5 |
dmp templates | 5 |
textual content | 5 |
relation types | 5 |
observed within | 5 |
table summarizes | 5 |
many different | 5 |
product embedding | 5 |
cumulative gain | 5 |
text query | 5 |
chemu lab | 5 |
base model | 5 |
symptom extraction | 5 |
media data | 5 |
per sentence | 5 |
clustering algorithms | 5 |
learning based | 5 |
dimensional representation | 5 |
will focus | 5 |
embeddings using | 5 |
partially supported | 5 |
art oltr | 5 |
classification tasks | 5 |
trained word | 5 |
latent models | 5 |
image recognition | 5 |
term repeat | 5 |
relevant images | 5 |
vector representations | 5 |
mismatch problem | 5 |
text information | 5 |
review embeddings | 5 |
chemical patent | 5 |
gram objective | 5 |
path set | 5 |
research data | 5 |
logistic model | 5 |
arabic tweets | 5 |
original pagerank | 5 |
help users | 5 |
recent work | 5 |
softmax function | 5 |
disaster categories | 5 |
context information | 5 |
negative matrix | 5 |
retrieved image | 5 |
specific bc | 5 |
category ac | 5 |
document lengths | 5 |
final representation | 5 |
interaction data | 5 |
computed using | 5 |
categorical attributes | 5 |
research papers | 5 |
classification algorithms | 5 |
tokenized path | 5 |
propagation models | 5 |
aggregation function | 5 |
user query | 5 |
image vector | 5 |
two generators | 5 |
alignment via | 5 |
visual features | 5 |
video summarization | 5 |
cooccur method | 5 |
generalized language | 5 |
input space | 5 |
learning techniques | 5 |
reproducibility study | 5 |
million pmc | 5 |
regularization term | 5 |
best performance | 5 |
user may | 5 |
logged bandit | 5 |
discounted cumulative | 5 |
search using | 5 |
connected layer | 5 |
relevance level | 5 |
information access | 5 |
wildcard queries | 5 |
domain adaptation | 5 |
benchmark datasets | 5 |
textual similarity | 5 |
sentence level | 5 |
related information | 5 |
recommendation approaches | 5 |
next section | 5 |
model without | 5 |
entity alignments | 5 |
text spans | 5 |
predictive model | 5 |
joint learning | 5 |
chemical entity | 5 |
different sources | 5 |
based features | 5 |
textual information | 5 |
based recommender | 5 |
neural architectures | 4 |
supervised ranking | 4 |
features generated | 4 |
evaluation purposes | 4 |
approaches based | 4 |
neural language | 4 |
latent variables | 4 |
structural similarity | 4 |
run time | 4 |
two participants | 4 |
transformer encoder | 4 |
weighting mechanism | 4 |
input sequence | 4 |
second level | 4 |
learning methods | 4 |
keyword search | 4 |
user clicks | 4 |
ranking functions | 4 |
context used | 4 |
using graph | 4 |
answer passage | 4 |
named entities | 4 |
different types | 4 |
input sentence | 4 |
lessons learned | 4 |
also include | 4 |
learning deep | 4 |
output label | 4 |
online reviews | 4 |
images retrieved | 4 |
label representations | 4 |
ranking algorithm | 4 |
ari acc | 4 |
guided deep | 4 |
binary matrix | 4 |
production ranker | 4 |
linear combination | 4 |
specific ab | 4 |
automatic image | 4 |
relevance framework | 4 |
long history | 4 |
ir approaches | 4 |
activated nodes | 4 |
term occurring | 4 |
gaussian mixture | 4 |
complex grammatical | 4 |
reuters rcv | 4 |
similarity measures | 4 |
previous section | 4 |
hierarchical bi | 4 |
image embedding | 4 |
feature representations | 4 |
task will | 4 |
dual queries | 4 |
manual seed | 4 |
method proposed | 4 |
relevant articles | 4 |
manually selected | 4 |
key information | 4 |
label features | 4 |
biomedical literature | 4 |
ranking algorithms | 4 |
sampling distribution | 4 |
different approaches | 4 |
decision making | 4 |
candidate phrases | 4 |
threshold values | 4 |
text content | 4 |
human evaluation | 4 |
natural disasters | 4 |
two categories | 4 |
text input | 4 |
edge weights | 4 |
end neural | 4 |
risk minimization | 4 |
evaluation protocol | 4 |
candidate ranker | 4 |
medical diagnosis | 4 |
make sure | 4 |
story picturing | 4 |
ranking results | 4 |
computational cost | 4 |
streaming module | 4 |
statistical language | 4 |
low dimensional | 4 |
three categories | 4 |
semantic meanings | 4 |
specific information | 4 |
document embeddings | 4 |
contextual features | 4 |
based method | 4 |
user queries | 4 |
query sentence | 4 |
punctuation marks | 4 |
attention scores | 4 |
query time | 4 |
optimal value | 4 |
document passages | 4 |
first layer | 4 |
art model | 4 |
automatically selected | 4 |
short length | 4 |
average number | 4 |
single modality | 4 |
participants stated | 4 |
entity types | 4 |
ranking based | 4 |
severe damage | 4 |
case study | 4 |
cooking recipes | 4 |
category topics | 4 |
candidate documents | 4 |
search applications | 4 |
previous state | 4 |
graph structure | 4 |
network schema | 4 |
exact matching | 4 |
also show | 4 |
chemical patents | 4 |
ir prototyping | 4 |
oltr method | 4 |
counterfactual risk | 4 |
retrieval tasks | 4 |
reformulated queries | 4 |
incomplete judgments | 4 |
prediction tasks | 4 |
chemical compound | 4 |
bm model | 4 |
one way | 4 |
result claim | 4 |
prediction models | 4 |
makes relevance | 4 |
statistical classification | 4 |
relevance model | 4 |
food images | 4 |
document i | 4 |
source code | 4 |
tf component | 4 |
simulating human | 4 |
input feature | 4 |
text embeddings | 4 |
path length | 4 |
conditional random | 4 |
mining suggestions | 4 |
clinical case | 4 |
readability formula | 4 |
supporting premises | 4 |
ir task | 4 |
stepwise illustration | 4 |
feature space | 4 |
computation graph | 4 |
hit path | 4 |
evaluation metric | 4 |
research fund | 4 |
idf component | 4 |
long queries | 4 |
recommendation methods | 4 |
bias terms | 4 |
search stage | 4 |
rank features | 4 |
english text | 4 |
users tend | 4 |
interesting observation | 4 |
based neural | 4 |
architecture models | 4 |
use word | 4 |
click model | 4 |
ir replicability | 4 |
every day | 4 |
symptom relation | 4 |
sequence learning | 4 |
premises supporting | 4 |
learning approaches | 4 |
story recall | 4 |
slightly better | 4 |
learning architecture | 4 |
recent advances | 4 |
retrieval research | 4 |
task asks | 4 |
results presented | 4 |
national research | 4 |
methods used | 4 |
open domain | 4 |
naive bayes | 4 |
contextual representations | 4 |
multimodal summarization | 4 |
previous approaches | 4 |
dimensional feature | 4 |
propensity model | 4 |
current ranker | 4 |
heterogeneous information | 4 |
web table | 4 |
existing approaches | 4 |
yet effective | 4 |
preprocessing steps | 4 |
extraction methods | 4 |
participants said | 4 |
two nodes | 4 |
random fields | 4 |
electronic health | 4 |
set contains | 4 |
significant improvements | 4 |
embedding point | 4 |
significantly outperforms | 4 |
ceur workshop | 4 |
result lists | 4 |
complex information | 4 |
include queries | 4 |
relation collection | 4 |
bound morphemes | 4 |
node embeddings | 4 |
patent documents | 4 |
query prediction | 4 |
vector machine | 4 |
web tables | 4 |
detect irony | 4 |
query claims | 4 |
media forms | 4 |
workshop proceedings | 4 |
topical relevance | 4 |
significantly different | 4 |
bidirectional lstm | 4 |
online setting | 4 |
nlp tasks | 4 |
second language | 4 |
two previous | 4 |
subword information | 4 |
text data | 4 |
probabilistic relevance | 4 |
participating systems | 4 |
management plan | 4 |
empirical results | 4 |
view generation | 4 |
scientific literature | 4 |
training graph | 4 |
verb stem | 4 |
annotator agreement | 4 |
computational argumentation | 4 |
results reported | 4 |
final score | 4 |
message passing | 4 |
available dataset | 4 |
multimodal graphs | 4 |
digital libraries | 4 |
base completion | 4 |
image summarization | 4 |
qatar national | 4 |
evaluated using | 4 |
scale reproducibility | 4 |
model using | 4 |
reformulation patterns | 4 |
two modalities | 4 |
truth cluster | 4 |
node type | 4 |
bringing order | 4 |
semantic relations | 4 |
topic representations | 4 |
cosine distance | 4 |
convolution weights | 4 |
emergency management | 4 |
hoc ranking | 4 |
two attention | 4 |
label interaction | 4 |
anchor query | 4 |
starting point | 4 |
graph structures | 4 |
user interactions | 4 |
bm variants | 4 |
softmax layer | 4 |
july th | 4 |
goggles model | 4 |
possibly correct | 4 |
video summary | 4 |
step size | 4 |
academic ir | 4 |
corresponding generator | 4 |
standards model | 4 |
centric context | 4 |
many researchers | 4 |
qa datasets | 4 |
search task | 4 |
working notes | 4 |
query language | 4 |
clustering views | 4 |
computational linguistics | 4 |
performs best | 4 |
digital images | 4 |
exploratory nature | 4 |
edge types | 4 |
heterogeneous skip | 4 |
semantic features | 4 |
large datasets | 4 |
session vs | 4 |
mutual information | 4 |
six bantu | 4 |
attention weight | 4 |
randomly select | 4 |
research council | 4 |
sentence encoder | 4 |
batch size | 4 |
traditional classification | 4 |
attacking premises | 4 |
original labels | 4 |
image modalities | 4 |
hoc information | 4 |
table presents | 4 |
curated features | 4 |
worthy claims | 4 |
different domains | 4 |
query suggestion | 4 |
feature learning | 4 |
political parties | 4 |
methods using | 4 |
ranking methods | 4 |
contextual representation | 4 |
analysis task | 4 |
dependencies among | 4 |
bias characteristics | 4 |
may also | 4 |
twitter stream | 4 |
recent years | 4 |
masked version | 4 |
metadata information | 4 |
weight vector | 4 |
retrieval method | 4 |
stop words | 4 |
textual transcripts | 4 |
trained models | 4 |
newly introduced | 4 |
analysis methods | 4 |
times add | 4 |
prior knowledge | 4 |
new interaction | 4 |
network architecture | 4 |
target query | 4 |
large graphs | 4 |
deep document | 4 |
recent works | 4 |
listwise context | 4 |
reformulation strategies | 4 |
parallel corpus | 4 |
lab overview | 4 |
based search | 4 |
feature weights | 4 |
containing text | 4 |
early stopping | 4 |
weight constraints | 4 |
recent approaches | 4 |
product recommendation | 4 |
information systems | 4 |
image embeddings | 4 |
math formula | 4 |
annotated data | 4 |
trigger word | 4 |
downstream task | 4 |
convolution layer | 4 |
attention score | 4 |
also find | 4 |
judgment process | 4 |
next question | 4 |
heterogeneous graph | 4 |
similar approach | 4 |
probabilistic ranking | 4 |
enterprise architecture | 4 |
common subtrees | 4 |
semantic representations | 4 |
multilayer perceptron | 4 |
ir researchers | 4 |
saliency recall | 4 |
vanilla bert | 4 |
context based | 4 |
text extraction | 4 |
bias vector | 4 |
fisher distributions | 4 |
preferential attachment | 4 |
network graphing | 4 |
also shown | 4 |
visual saliency | 4 |
original version | 4 |
kernel pooling | 4 |
real users | 4 |
single verb | 4 |
capture important | 4 |
relevance decisions | 4 |
simple way | 4 |
matching signals | 4 |
simple yet | 4 |
using bert | 4 |
positive reviews | 4 |
aggregate relevance | 4 |
three different | 4 |
visual feature | 4 |
semantic indexing | 4 |
constrained clustering | 4 |
rcv rcv | 4 |
statistical significance | 4 |
user votes | 4 |
structural features | 4 |
using bm | 4 |
common friends | 4 |
related terms | 4 |
extracting insights | 4 |
medium sessions | 4 |
educational image | 4 |
i sum | 4 |
result claims | 4 |
simple concatenation | 4 |
domain knowledge | 4 |
exploratory tasks | 4 |
semantic technology | 4 |
counterfactual online | 4 |
widest common | 4 |
query run | 3 |
similarity measure | 3 |
single task | 3 |
model allows | 3 |
produce recommendations | 3 |
new york | 3 |
extraction tasks | 3 |
maximum matching | 3 |
explainable artificial | 3 |
normalization vsm | 3 |
uniform distribution | 3 |
initialized randomly | 3 |
representation space | 3 |
keeping track | 3 |
much larger | 3 |
probabilistic multileaving | 3 |
brief analysis | 3 |
additional context | 3 |
document pair | 3 |
two key | 3 |
document analysis | 3 |
label dependencies | 3 |
top retrieved | 3 |
language use | 3 |
interaction mechanism | 3 |
training collection | 3 |
relational data | 3 |
story recipe | 3 |
high level | 3 |
data used | 3 |
media users | 3 |
changes trends | 3 |
benchmark results | 3 |
human relevance | 3 |
unsupervised approaches | 3 |
relational databases | 3 |
worthy claim | 3 |
define two | 3 |
ranking premises | 3 |
claim cluster | 3 |
feature similarity | 3 |
successful approaches | 3 |
existing sentiment | 3 |
ehealth information | 3 |
temperature parameter | 3 |
neural story | 3 |
click logs | 3 |
combine information | 3 |
ehealth tasks | 3 |
use attention | 3 |
heterogeneous sources | 3 |
ranked documents | 3 |
automatic verification | 3 |
make use | 3 |
bm ranking | 3 |
significantly larger | 3 |
neural learning | 3 |
feature nodes | 3 |
longer documents | 3 |
multilingual setting | 3 |
views complete | 3 |
adam optimization | 3 |
strong baselines | 3 |
different settings | 3 |
embedding approaches | 3 |
lda model | 3 |
different information | 3 |
sentence components | 3 |
mention prediction | 3 |
include performance | 3 |
documents retrieved | 3 |
task features | 3 |
cnn architecture | 3 |
field weights | 3 |
reading comprehension | 3 |
frequently used | 3 |
traditional oltr | 3 |
metapath vec | 3 |
remain sharp | 3 |
system needs | 3 |
performs well | 3 |
stl lmote | 3 |
consider semantic | 3 |
input image | 3 |
best result | 3 |
modal objects | 3 |
clustering algorithm | 3 |
following sections | 3 |
umls vocabulary | 3 |
auc scores | 3 |
running example | 3 |
might include | 3 |
model consists | 3 |
large collection | 3 |
will explore | 3 |
computationally expensive | 3 |
informative tweet | 3 |
modeling approaches | 3 |
different aspects | 3 |
source search | 3 |
system returns | 3 |
text using | 3 |
promising results | 3 |
binary programming | 3 |
views observed | 3 |
document frequency | 3 |
three trials | 3 |
different methods | 3 |
widest matched | 3 |
distance function | 3 |
shannon divergence | 3 |
task characteristics | 3 |
neural word | 3 |
mutual attention | 3 |
performs better | 3 |
image annotation | 3 |
using lucene | 3 |
entire document | 3 |
deep structured | 3 |
gram model | 3 |
context selection | 3 |
text matching | 3 |
class assignments | 3 |
several experiments | 3 |
using stochastic | 3 |
generalize well | 3 |
benchmark data | 3 |
english training | 3 |
single field | 3 |
text search | 3 |
term similarities | 3 |
complete model | 3 |
shot setting | 3 |
african languages | 3 |
attention vector | 3 |
web images | 3 |
combining two | 3 |
first publicly | 3 |
current approaches | 3 |
news articles | 3 |
corpus used | 3 |
graph matching | 3 |
auc improvement | 3 |
input text | 3 |
freely available | 3 |
documents containing | 3 |
lucene open | 3 |
task neural | 3 |
test graph | 3 |
interpretable document | 3 |
sigmoid function | 3 |
many approaches | 3 |
approach called | 3 |
document collections | 3 |
online ltr | 3 |
specific domain | 3 |
image modality | 3 |
effective ir | 3 |
ranking tasks | 3 |
damage identification | 3 |
highest score | 3 |
embeddings trained | 3 |
data collection | 3 |
two classes | 3 |
gb ram | 3 |
dataset content | 3 |
previous models | 3 |
rank factorization | 3 |
review representations | 3 |
iterative refinement | 3 |
ir techniques | 3 |
main components | 3 |
english languages | 3 |
important ir | 3 |
different variants | 3 |
part comprises | 3 |
semantic model | 3 |
existing models | 3 |
recurrent seq | 3 |
classification model | 3 |
different search | 3 |
volume prediction | 3 |
matched common | 3 |
graph completion | 3 |
penultimate layer | 3 |
following three | 3 |
table content | 3 |
context features | 3 |
art approaches | 3 |
dynamic tree | 3 |
application scenarios | 3 |
standard information | 3 |
lexicon may | 3 |
proposed framework | 3 |
item similarity | 3 |
unsupervised baselines | 3 |
text descriptions | 3 |
dataset ranking | 3 |
external resource | 3 |
baseline model | 3 |
science foundation | 3 |
snippet retrieval | 3 |
downstream tasks | 3 |
relevant answers | 3 |
node representations | 3 |
domain shift | 3 |
three datasets | 3 |
embedding dimension | 3 |
via deep | 3 |
general topics | 3 |
tree cut | 3 |
retrieval heuristics | 3 |
longer tasks | 3 |
content generated | 3 |
classification accuracy | 3 |
image descriptions | 3 |
best way | 3 |
queries within | 3 |
neural architecture | 3 |
personalized recommendation | 3 |
search sessions | 3 |
data using | 3 |
using cross | 3 |
language dependent | 3 |
evaluation corpus | 3 |
ranker weights | 3 |
unbiased estimator | 3 |
former case | 3 |
will now | 3 |
supporting premise | 3 |
model learns | 3 |
crossmodal retrieval | 3 |
annotation scores | 3 |
simpler ones | 3 |
initial experiments | 3 |
item i | 3 |
semantic space | 3 |
based relevance | 3 |
online retailers | 3 |
recurrent convolutional | 3 |
multimedia retrieval | 3 |
parameter space | 3 |
parallel data | 3 |
two vectors | 3 |
variational autoencoder | 3 |
filtering modules | 3 |
efficient query | 3 |
deep residual | 3 |
one answer | 3 |
related document | 3 |
negative values | 3 |
view learning | 3 |
datasets show | 3 |
clinical cases | 3 |
different modules | 3 |
recently proposed | 3 |
every document | 3 |
previous editions | 3 |
centric contexts | 3 |
dake outperforms | 3 |
joint embeddings | 3 |
based loss | 3 |
visual analytics | 3 |
fusion model | 3 |
various methods | 3 |
research problem | 3 |
text snippets | 3 |
interdisciplinary actors | 3 |
patents task | 3 |
synthetic minority | 3 |
antonym detection | 3 |
available web | 3 |
noisy gazetteer | 3 |
efficient estimation | 3 |
text fields | 3 |
given document | 3 |
textual summary | 3 |
document scores | 3 |
spanish version | 3 |
text illustration | 3 |
corresponding image | 3 |
questions start | 3 |
scientific publications | 3 |
various data | 3 |
expansion methods | 3 |
latent feature | 3 |
semantic embeddings | 3 |
similar trends | 3 |
using monolingual | 3 |
stemming lemmatizing | 3 |
fair comparison | 3 |
maximum value | 3 |
many domains | 3 |
use three | 3 |
successive queries | 3 |
score based | 3 |
click settings | 3 |
descent algorithms | 3 |
informative tweets | 3 |
rate decay | 3 |
topics underlying | 3 |
irony devices | 3 |
baseline approaches | 3 |
sentiment classification | 3 |
novel neural | 3 |
use early | 3 |
learning semantic | 3 |
rank algorithm | 3 |
opinion words | 3 |
computationally efficient | 3 |
query capitalism | 3 |
substantial community | 3 |
unsupervised fashion | 3 |
new benchmark | 3 |
network based | 3 |
structure information | 3 |
rich text | 3 |
flow graph | 3 |
york times | 3 |
research community | 3 |
mixture models | 3 |
two models | 3 |
multiclass classification | 3 |
product review | 3 |
include adaptations | 3 |
supervised model | 3 |
claim frequency | 3 |
users may | 3 |
complex structures | 3 |
latent factor | 3 |
opposite stance | 3 |
nbest answers | 3 |
test stage | 3 |
grant agreement | 3 |
task labels | 3 |
research prototype | 3 |
sets respectively | 3 |
approximate review | 3 |
including medical | 3 |
wiki dataset | 3 |
external resources | 3 |
critical component | 3 |
propose memis | 3 |
per cluster | 3 |
central vector | 3 |
label embedding | 3 |
i aim | 3 |
link dom | 3 |
traditional ranking | 3 |
semantic text | 3 |
unimodal frameworks | 3 |
location extraction | 3 |
text understanding | 3 |
medium tasks | 3 |
cee task | 3 |
feature map | 3 |
ymptom relation | 3 |
source ir | 3 |
corresponding posting | 3 |
root paths | 3 |
following previous | 3 |
graph querying | 3 |
hierarchical transformer | 3 |
indigenous african | 3 |
house annotators | 3 |
generated machine | 3 |
bias scores | 3 |
disaster response | 3 |
document set | 3 |
query paths | 3 |
thousand articles | 3 |
least squares | 3 |
find terms | 3 |
maintaining similar | 3 |
community interest | 3 |
incorporate sentiment | 3 |
clustering problem | 3 |
domain regarding | 3 |
ranking scores | 3 |
determine whether | 3 |
performance compared | 3 |
text queries | 3 |
comprehension dataset | 3 |
linguistic resources | 3 |
learns word | 3 |
chemical named | 3 |
one participant | 3 |
retrieved images | 3 |
dot product | 3 |
national funds | 3 |
disambiguation task | 3 |
hierarchical clustering | 3 |
relative importance | 3 |
traditional approaches | 3 |
search systems | 3 |
smoothing methods | 3 |
data analysis | 3 |
two data | 3 |
generating functions | 3 |
one class | 3 |
recommend people | 3 |
supporting contextual | 3 |
judged documents | 3 |
baseline ranker | 3 |
multimodal tweets | 3 |
newsgroups dataset | 3 |
automatic speech | 3 |
wide range | 3 |
math expressions | 3 |
medical tasks | 3 |
psychological disorders | 3 |
across datasets | 3 |
time complexity | 3 |
document vectors | 3 |
sentences construct | 3 |
clustering approaches | 3 |
multimodal damage | 3 |
two annotators | 3 |
embedding layer | 3 |
raw co | 3 |
symptom relationships | 3 |
based data | 3 |
extracted features | 3 |
sequential semantic | 3 |
players game | 3 |
codiesp corpus | 3 |
bootstrapping approach | 3 |
topic tweets | 3 |
five years | 3 |
textual description | 3 |
init seeds | 3 |
word list | 3 |
annotated corpus | 3 |
hierarchical learning | 3 |
second term | 3 |
users often | 3 |
human evaluators | 3 |
markov random | 3 |
digital music | 3 |
official measure | 3 |
ratings matrix | 3 |
clustering approach | 3 |
syntactic dependencies | 3 |
false positives | 3 |
noisy clicks | 3 |
test split | 3 |
support flow | 3 |
index format | 3 |
review rating | 3 |
interesting results | 3 |
sequence labeling | 3 |
better overview | 3 |
lingual knowledge | 3 |
automatic generation | 3 |
biomedical text | 3 |
walk based | 3 |
node i | 3 |
th sentence | 3 |
images relevant | 3 |
social platforms | 3 |
experimental setup | 3 |
one possible | 3 |
web site | 3 |
human generated | 3 |
improve performance | 3 |
term representations | 3 |
novel approach | 3 |
model significantly | 3 |
linear transformation | 3 |
hierarchical self | 3 |
sentence representation | 3 |
query history | 3 |
speech recognition | 3 |
new stepwise | 3 |
transformation matrix | 3 |
retrieved results | 3 |
classification models | 3 |
national science | 3 |
three main | 3 |
retrieval problem | 3 |
propagation model | 3 |
specialized inverted | 3 |
damage present | 3 |
two approaches | 3 |
i anticipate | 3 |
recommendation problem | 3 |
verb forms | 3 |
var add | 3 |
two aspects | 3 |
semantically meaningful | 3 |
semantic meaning | 3 |
mathir task | 3 |
prediction problem | 3 |
attribute modalities | 3 |
deep architecture | 3 |
greek web | 3 |
first time | 3 |
false positive | 3 |
using clickthrough | 3 |
see eq | 3 |
result premises | 3 |
argument retrieval | 3 |
existing works | 3 |
results list | 3 |
dense layer | 3 |
grant gsra | 3 |
based syst | 3 |
much lower | 3 |
complete views | 3 |
differentiating factor | 3 |
vector machines | 3 |
world impact | 3 |
risk estimator | 3 |
ltr model | 3 |
exact document | 3 |
approach described | 3 |
clinical codes | 3 |
unsuccessful tasks | 3 |
comparative study | 3 |
embedding method | 3 |
paper describes | 3 |
interaction network | 3 |
retrieval functions | 3 |
sql query | 3 |
parameter setting | 3 |
empirical analysis | 3 |
story illustration | 3 |
image analysis | 3 |
new field | 3 |
different values | 3 |
funding agencies | 3 |
item co | 3 |
reconstruction loss | 3 |
various sources | 3 |
outperforms previous | 3 |
lingual word | 3 |
error analysis | 3 |
open source | 3 |
overall objective | 3 |
analysis systems | 3 |
different claims | 3 |
summary generation | 3 |
relevance assessments | 3 |
guided topic | 3 |
graph alignment | 3 |
parameter scan | 3 |
hierarchical recurrent | 3 |
working group | 3 |
analysis using | 3 |
work will | 3 |
class membership | 3 |
single image | 3 |
data structure | 3 |
binarized version | 3 |
political party | 3 |
degree users | 3 |
prediction based | 3 |
data mining | 3 |
commercial search | 3 |
different sizes | 3 |
scores obtained | 3 |
classification using | 3 |
based ranking | 3 |
citation network | 3 |
explainable recommendation | 3 |
prediction task | 3 |
loss function | 3 |
image representation | 3 |
bert model | 3 |
multimodal network | 3 |
reading difficulty | 3 |
empirically analyze | 3 |
analysis approaches | 3 |
parameter tuning | 3 |
downstream modules | 3 |
papers show | 3 |
uniform sampling | 3 |
preference matrix | 3 |
given organisation | 3 |
rank documents | 3 |
multilingual word | 3 |
abdominal pain | 3 |
many challenges | 3 |
models used | 3 |
completed view | 3 |
two parts | 3 |
recommendation algorithms | 3 |
new chemical | 3 |
predict relevance | 3 |
antique consists | 3 |
crf layer | 3 |
estimated using | 3 |
high levels | 3 |
high support | 3 |
across modalities | 3 |
will examine | 3 |
evaluation using | 3 |
corpus using | 3 |
effective way | 3 |
using local | 3 |
proposed solutions | 3 |
unimodal models | 3 |
track overview | 3 |
factorization techniques | 3 |
query refinement | 3 |
negation morphemes | 3 |
reproduce prf | 3 |
ranking performance | 3 |
content analysis | 3 |
similarity threshold | 3 |
terms based | 3 |
machine reading | 3 |
unique product | 3 |
report results | 3 |
multiple rankers | 3 |
estimated ratings | 3 |
majority voting | 3 |
meaningful word | 3 |
outperform stm | 3 |
image database | 3 |
time step | 3 |
recent methods | 3 |
existing recommendation | 3 |
selected seed | 3 |
two stages | 3 |
qualitative analysis | 3 |
achieve better | 3 |
scale hierarchical | 3 |
word i | 3 |
continuous word | 3 |
coltr uses | 3 |
information like | 3 |
iters i | 3 |
good quality | 3 |
different scoring | 3 |
missing versions | 3 |
graph embeddings | 3 |
unsupervised manner | 3 |
global attention | 3 |
html parser | 3 |
recent studies | 3 |
math retrieval | 3 |
will use | 3 |
summarization using | 3 |
chemical entities | 3 |
term frequencies | 3 |
larger dataset | 3 |
relevance principle | 3 |
recipe instruction | 3 |
normalized discounted | 3 |
output images | 3 |
level contextual | 3 |
using external | 3 |
input claim | 3 |
queries containing | 3 |
ss variation | 3 |
image generation | 3 |
hierarchical image | 3 |
achieving similar | 3 |
data portals | 3 |
results showed | 3 |
standard classification | 3 |
random field | 3 |
based collaborative | 3 |
obtain good | 3 |
two people | 3 |
introduced biased | 3 |
term reuse | 3 |
making use | 3 |
bayes classifier | 3 |
goggles system | 3 |
estimated based | 3 |
english collections | 3 |
discriminative power | 3 |
web page | 3 |
neural ad | 3 |
top documents | 3 |
completion methods | 3 |
depression level | 3 |
level relevance | 3 |
single neighbor | 3 |
performance across | 3 |
sentiment values | 3 |
model training | 3 |
expected since | 3 |
semantically similar | 3 |
residual learning | 3 |
score text | 3 |
two application | 3 |
system variants | 3 |
randomly selected | 3 |
final label | 3 |
sentiment rating | 3 |
exact matches | 3 |
damage indicated | 3 |
word sentence | 3 |
good performance | 3 |
label generator | 3 |
two terms | 3 |
first attempt | 3 |
available datasets | 3 |
information source | 3 |
tokenized paths | 3 |
candidate images | 3 |
observed view | 3 |
duplicate results | 3 |
features based | 3 |
clef lab | 3 |
interaction matrix | 3 |
much better | 3 |
art results | 3 |
discrimination constraint | 3 |
gsra grant | 3 |
graph visualisation | 3 |
clicks may | 3 |
research purposes | 3 |
independence cascade | 3 |
scientific papers | 3 |
starting material | 3 |
chs task | 3 |
models applied | 3 |
top results | 3 |
shown improvements | 3 |
average improvement | 3 |
multiple clustering | 3 |
aloe vera | 3 |
trained model | 3 |
outperforms hft | 3 |
appropriate illustrations | 3 |
text pair | 3 |
interactions within | 3 |
simple model | 3 |
per word | 3 |
text features | 3 |
structured semantic | 3 |
clustering loss | 3 |
embedding matrix | 3 |
randomly sampled | 3 |
batch learning | 3 |
enriching word | 3 |
actionable data | 3 |
retrieval metrics | 3 |
negative sampling | 3 |
robust reading | 3 |
different techniques | 3 |
new batch | 3 |
sequential data | 3 |
english queries | 3 |
linear svm | 3 |
user searching | 3 |
using different | 3 |
point scale | 3 |
network model | 3 |
following four | 3 |
sequence models | 3 |
solved using | 3 |
adjusted rand | 3 |
document image | 3 |
encoding layer | 3 |
embedding via | 3 |
growing interest | 3 |
also provide | 3 |
different query | 3 |
learning representations | 3 |
individual modules | 3 |
takes advantage | 3 |
represented using | 3 |
crf model | 3 |
different axioms | 3 |
small set | 3 |
opinion mining | 3 |
verification pipeline | 3 |
memory networks | 3 |
noisy click | 3 |
best answer | 3 |
query characteristics | 3 |
retrieval effectiveness | 3 |
fusion models | 3 |
syntactic changes | 3 |
understand relationships | 3 |
biased surfers | 3 |
may share | 3 |
traditional retrieval | 3 |
greek political | 3 |
query node | 3 |
based learning | 3 |
later used | 3 |
upper bound | 3 |
large corpora | 3 |
text readability | 3 |
model pre | 3 |
graphs using | 3 |
also define | 3 |
rand index | 3 |
also found | 3 |
linear threshold | 3 |
relevant document | 3 |
document according | 3 |
retrieve reviews | 3 |
user behaviour | 3 |
early prediction | 3 |
parameter complexity | 3 |
international joint | 3 |
inverse document | 3 |
models provide | 3 |
ironic tweets | 3 |
user might | 3 |
grid search | 3 |
suggestion class | 3 |
media entries | 3 |
original document | 3 |
large scale | 3 |
customers may | 3 |
query result | 3 |
root nodes | 3 |
instruction steps | 3 |
use different | 3 |
may vary | 3 |
show empirically | 3 |
attacking premise | 3 |
larger set | 3 |
mercer smoothing | 3 |
optimization technique | 3 |
latent representation | 3 |
within tasks | 3 |
i indicates | 3 |
component analysis | 3 |
visual cues | 3 |
new task | 3 |
common subtree | 3 |
data generated | 3 |
modal correlation | 3 |
information gain | 3 |
means algorithm | 3 |
data model | 3 |
language training | 3 |
text mining | 3 |
information space | 3 |
related product | 3 |
search corpus | 3 |
damage analysis | 3 |
argument search | 3 |
sufficient information | 3 |
relevance judgment | 3 |
small dataset | 3 |
item representations | 3 |
multiple domains | 3 |
novel multimodal | 3 |
context embedding | 3 |
joint conference | 3 |
recursive layer | 3 |
standards working | 3 |
chemdner patents | 3 |
stemming verbs | 3 |
rank i | 3 |
new disaster | 3 |
significant amount | 3 |
linear model | 3 |
automatic keyphrase | 3 |
also called | 3 |
singleview approach | 2 |
semantic representation | 2 |
different word | 2 |
learning cross | 2 |
prevent neural | 2 |
tabular form | 2 |
primarily designed | 2 |
document level | 2 |
hybrid approaches | 2 |
state output | 2 |
without explicit | 2 |
two scenarios | 2 |
view conditionally | 2 |
attention aggregation | 2 |
first method | 2 |
models rely | 2 |
attentive recurrent | 2 |
product representation | 2 |
results using | 2 |
wpso jug | 2 |
ask questions | 2 |
third edition | 2 |
input representations | 2 |
unsupervised counterpart | 2 |
multiple information | 2 |
federal ministry | 2 |
human judgment | 2 |
formula browsing | 2 |
matching signal | 2 |
classification recurrent | 2 |
matching size | 2 |
million query | 2 |
different levels | 2 |
three baselines | 2 |
designing task | 2 |
unexpected associations | 2 |
table show | 2 |
found tweets | 2 |
polarity classification | 2 |
model includes | 2 |
fusion approach | 2 |
fully supporting | 2 |
features using | 2 |
average cosine | 2 |
tag pair | 2 |
recurrent sequence | 2 |
next word | 2 |
factor vector | 2 |
five publicly | 2 |
random selection | 2 |
proximity measures | 2 |
effectiveness differences | 2 |
encoding techniques | 2 |
transrev outperforms | 2 |
qa dataset | 2 |
novel recommendations | 2 |
evaluation tasks | 2 |
model satisfies | 2 |
words based | 2 |
previous step | 2 |
item recommendation | 2 |
unsupervised visual | 2 |
engines using | 2 |
per block | 2 |
adding ra | 2 |
second snapshot | 2 |
complete bipartite | 2 |
ad placement | 2 |
automatically identify | 2 |
sigir conference | 2 |
th term | 2 |
researchers using | 2 |
common users | 2 |
pivotal role | 2 |
first describe | 2 |
visualizing data | 2 |
wikipedia pages | 2 |
disease network | 2 |
linear methods | 2 |
english datasets | 2 |
ironic hashtags | 2 |
filtering methods | 2 |
support obtained | 2 |
wrong predictions | 2 |
using insights | 2 |
oltr approaches | 2 |
data statistics | 2 |
english documents | 2 |
give us | 2 |
agreement using | 2 |
formula similarity | 2 |
information society | 2 |
additional relevant | 2 |
surface features | 2 |
diseases symptoms | 2 |
semantic relationships | 2 |
one used | 2 |
first work | 2 |
leveraging domain | 2 |
social representations | 2 |
per topic | 2 |
argumentative discourse | 2 |
dmp creation | 2 |
shared representations | 2 |
implicit matrix | 2 |
whole collection | 2 |
support flows | 2 |
analysis modules | 2 |
symbol similarity | 2 |
clustering process | 2 |
original information | 2 |
new user | 2 |
monte carlo | 2 |
singular value | 2 |
findings based | 2 |
long documents | 2 |
stop word | 2 |
automatic construction | 2 |
search context | 2 |
trained separately | 2 |
modern information | 2 |
image encoder | 2 |
highly connected | 2 |
significantly lower | 2 |
token paths | 2 |
two domains | 2 |
typical setting | 2 |
three shared | 2 |
different users | 2 |
reformulation behavior | 2 |
using block | 2 |
first part | 2 |
patent claims | 2 |
customer reviews | 2 |
improve results | 2 |
text quality | 2 |
semantic textual | 2 |
involves two | 2 |
vrss outperforms | 2 |
available data | 2 |
feedback methods | 2 |
triple alignments | 2 |
plugging back | 2 |
chronological order | 2 |
kincaid index | 2 |
class samples | 2 |
allows transrev | 2 |
competitive baselines | 2 |
horizon programme | 2 |
best supporting | 2 |
majority consensus | 2 |
languages beyond | 2 |
social context | 2 |
irrelevant documents | 2 |
english speaking | 2 |
gazetteer using | 2 |
information processing | 2 |
include articles | 2 |
query sessions | 2 |
allocation model | 2 |
hierarchical tasks | 2 |
using lexical | 2 |
multimodal neural | 2 |
extract important | 2 |
includes queries | 2 |
three test | 2 |
health search | 2 |
query opt | 2 |
idne performs | 2 |
must identify | 2 |
writing styles | 2 |
separate embeddings | 2 |
systems usually | 2 |
outperforms several | 2 |
model demonstrates | 2 |
semantic models | 2 |
passages aggregated | 2 |
task uses | 2 |
using twitter | 2 |
specific challenges | 2 |
image types | 2 |
semantic concepts | 2 |
possible ways | 2 |
two hidden | 2 |
late fusion | 2 |
pseudo relevance | 2 |
different clustering | 2 |
customer expectations | 2 |
corp eval | 2 |
multilingual tasks | 2 |
indexed paths | 2 |
graph convolution | 2 |
online review | 2 |
crowdsourcing procedure | 2 |
optimization algorithm | 2 |
different platforms | 2 |
high performance | 2 |
readability index | 2 |
results include | 2 |
bimodal low | 2 |
training deep | 2 |
optimal performance | 2 |
source corpus | 2 |
academic researchers | 2 |
generating schema | 2 |
model uses | 2 |
yet another | 2 |
relevance information | 2 |
jointly learn | 2 |
personal search | 2 |
best matched | 2 |
task model | 2 |
shows examples | 2 |
comprehensive relevance | 2 |
generation networks | 2 |
short text | 2 |
english test | 2 |
relevant snippets | 2 |
steps described | 2 |
faculty award | 2 |
learning objectives | 2 |
kgje qy | 2 |
sent dataset | 2 |
based modeling | 2 |
two clusters | 2 |
model produces | 2 |
experimental foundations | 2 |
scholarly documents | 2 |
multiple attribute | 2 |
subtree rooted | 2 |
browsing task | 2 |
prf models | 2 |
primarily focused | 2 |
makes sense | 2 |
plays within | 2 |
previous two | 2 |
counterfactual ltr | 2 |
different solutions | 2 |
faster top | 2 |
load weebit | 2 |
representations learned | 2 |
time comparative | 2 |
tivs task | 2 |
overall meanings | 2 |
match using | 2 |
experiments show | 2 |
approaches used | 2 |
reviews online | 2 |
common data | 2 |
based matrix | 2 |
information must | 2 |
product identifier | 2 |
traditional readability | 2 |
calculated based | 2 |
ratn shah | 2 |
mechanical turk | 2 |
observed differences | 2 |
produces interpretable | 2 |
hit rate | 2 |
normal distribution | 2 |
recipe text | 2 |
speaking users | 2 |
aware sequential | 2 |
structure matching | 2 |
better result | 2 |
approach achieves | 2 |
overall retrieval | 2 |
given user | 2 |
popular wse | 2 |
auxiliary function | 2 |
available collection | 2 |
worthiness task | 2 |
joint ilp | 2 |
yet received | 2 |
label enhanced | 2 |
two sides | 2 |
like cuisines | 2 |
genomept project | 2 |
learning multi | 2 |
learning algorithm | 2 |
still significantly | 2 |
efficient data | 2 |
top ranked | 2 |
useful nodes | 2 |
following research | 2 |
gated neural | 2 |
dimensional fasttext | 2 |
language technologies | 2 |
making process | 2 |
reproducibility challenge | 2 |
similar questions | 2 |
lmp systems | 2 |
pet food | 2 |
spanish variants | 2 |
performance improvements | 2 |
ranked search | 2 |
research achieving | 2 |
th row | 2 |
diagnostic evaluation | 2 |
two editions | 2 |
rank challenge | 2 |
process web | 2 |
deep feedforward | 2 |
rank information | 2 |
dataset without | 2 |
embedding algorithm | 2 |
query text | 2 |
related queries | 2 |
linear classifier | 2 |
lexicon information | 2 |
labels obtained | 2 |
assessment systems | 2 |
versus disjunctively | 2 |
heterogeneous networks | 2 |
movie summarization | 2 |
special attention | 2 |
majority vote | 2 |
ranking baselines | 2 |
generation method | 2 |
required participants | 2 |
model configuration | 2 |
responsible humans | 2 |
monolingual architectures | 2 |
valuable insights | 2 |
feature vector | 2 |
relevance evidence | 2 |
modal similarities | 2 |
achieve greater | 2 |
will show | 2 |
languages typically | 2 |
retrieved using | 2 |
ltr datasets | 2 |
monolingual data | 2 |
effective systems | 2 |
respectively second | 2 |
structured diagrams | 2 |
categorizes articles | 2 |
new evaluation | 2 |
lower star | 2 |
different task | 2 |
principal component | 2 |
via graph | 2 |
source information | 2 |
value decomposition | 2 |
first relevant | 2 |
world stats | 2 |
simple example | 2 |
maximum distance | 2 |
straightforward way | 2 |
network information | 2 |
abstract concepts | 2 |
inferior performance | 2 |
emerging field | 2 |
less nbest | 2 |
meeting recordings | 2 |
every node | 2 |
effective retrieval | 2 |
propensity overfitting | 2 |
provide textual | 2 |
illustrative images | 2 |
performance significantly | 2 |
trainable weight | 2 |
vice versa | 2 |
long time | 2 |
researcher susana | 2 |
outperforms existing | 2 |
less affected | 2 |
single word | 2 |
hard negatives | 2 |
unsupervised semantic | 2 |
standard alone | 2 |
search trails | 2 |
intelligent information | 2 |
several techniques | 2 |
bilingual documents | 2 |
effective ranker | 2 |
wikimedia commons | 2 |
stl smote | 2 |
upc xu | 2 |
joint integer | 2 |
target tweet | 2 |
supervised classification | 2 |
conducted using | 2 |
word translation | 2 |
representations based | 2 |
text references | 2 |
deep structure | 2 |
summarization task | 2 |
weighting query | 2 |
relevant signals | 2 |
north american | 2 |
associated images | 2 |
slideimages contains | 2 |
automatic helpfulness | 2 |
next layer | 2 |
results indicate | 2 |
clustering performance | 2 |
fuses information | 2 |
laplacian seed | 2 |
sentiment value | 2 |
special queries | 2 |
dependency graphs | 2 |
media images | 2 |
dataset pair | 2 |
argumentation quality | 2 |
differentiable formulations | 2 |
query pair | 2 |
tweet streaming | 2 |
words like | 2 |
evolutionary principle | 2 |
significant impact | 2 |
work well | 2 |
terms used | 2 |
fundamental ir | 2 |
positive examples | 2 |
first two | 2 |
many languages | 2 |