Please enable JavaScript.
Coggle requires JavaScript to display documents.
answer extraction (text summarization (algorithms (TextRank (https://rare…
answer extraction
text summarization
extraction get part of original text
abstraction - generates new one
supervised
unsupervised
algorithms
TextRank
https://rare-technologies.com/text-summarization-with-gensim/
LexRank
https://github.com/miso-belica/sumy
Google TextSum
https://github.com/tensorflow/models/tree/master/textsum
http://delivery.acm.org/10.1145/980000/974188/p296-abney.pdf?ip=185.121.116.2&id=974188&acc=OPEN&key=4D4702B0C3E38B35%2E4D4702B0C3E38B35%2E4D4702B0C3E38B35%2E6D218144511F3437&CFID=995569141&CFTOKEN=68780300&__acm__=1508151028_575362bc3e74e617fc0f3a8a353854ec
1 stage: get most promising passages of most promissing documents
SMART information retrieval system
2 Linguistic stage: answer extracion/detection top 50 passages
Entity Extraction
Entity Classification
Entity Ranking
Query Classification
score
250 B length = 0.54, 68% in top5
50B 0.35 46% top 5
datasets
TREC
TREC information-retrieval conference
Kwok et al. 2001, Clarke et al. 2001
http://trec.nist.gov/pubs/trec25/trec2016.html
passage retrieval
https://groups.csail.mit.edu/infolab/publications/Tellex-etal-SIGIR03.pdf
https://aclweb.org/anthology/I/I11/I11-1117.pdf
lemur
http://trec.nist.gov/pubs/trec25/papers/MayoNLPTeam-CL.pdf
:
Part-of-Speech based Query Term Weighting (POS-BoW)
W eight{NN, NNS, V BN, V BD, JJ, RB, NNP} =
{0.5970, 0.2265, 0.3065, 0.2260, 0.3730, 0.1040, 0.8930}
Clinical Information Extraction Enhanced Markov Random
Field Model (IE-MRF)
kind of ngrams builder
http://trec.nist.gov/pubs/trec25/papers/prna-QA.pdf
LDA
нормування по піддаленості слів - спочатку нормуємо від 0-1 потим множимо на коефіцієнт
http://ccis2k.org/iajit/PDF/%20Vol%2013,%20No.%201A%20(Special%20Issue)/348.pdf
pattern transformation
http://trec.nist.gov/pubs/trec15/papers/fudan-zheng.geo.final.pdf
sentence retival score
match score
mismatched score
other, sims like tf-idf
http://www.aclweb.org/anthology/D13-1103
QANUS - open source framework
YourQA (Quarteroni
and Manandhar, 2006
The system is able
to provide both factoid and complex answers such
as definitions and descriptions