Multiple word alignment with profile hidden Markov models

Bhargava, Aditya; Kondrak, Grzegorz

doi:10.3115/1620932.1620940

Cited by 20 publications

(11 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since strings in biology and linguistics are substantially different, with the former being long and comprising a small alphabet and the latter being very short but containing a large number of different phones, algorithms used in biology cannot be directly applied to linguistic data. In the past decade, several approaches to automatic multiple string alignment have been developed for linguistic data (Bhargava and Kondrak, 2009;Prokić, 2010;List, 2012).…”

Section: Automatic Alignment Of Phonetic Datamentioning

confidence: 99%

Combining Regular Sound Correspondences and Geographic Spread

Prokić¹,

Cysouw²

2013

Lang Dyn Change

View full text Add to dashboard Cite

In this paper we combine the geographic variation of closely related language variants (‘dialects’) with the distribution of sound correspondences through the lexicon. One of the central problems with sound correspondences at the dialect level is that they are not very regular, especially when they are investigated in sufficient detail. Sound changes spread both through a language (e.g., from one word to another) and through the population of speakers (in our case through a population of villages with different dialects). Both processes happen at the same time, and the challenge is to reconstruct what has happened from a snapshot of synchronic data. The method described in this paper allows us to track the geographic spread of sound changes and the underlying patterns of linguistic diversity simultaneously. By combining the two, it is possible to detect areas of intensive linguistic contact and gain better insight into the mechanisms of language change.

show abstract

Section: Automatic Alignment Of Phonetic Datamentioning

confidence: 99%

Combining Regular Sound Correspondences and Geographic Spread

Prokić¹,

Cysouw²

2013

Lang Dyn Change

View full text Add to dashboard Cite

show abstract

“…For the purpose of joint generation, we need to align triples S, P and T prior to training. The alignment of multiple strings is a challenging problem (Bhargava and Kondrak, 2009). In general, there is no obvious way of merging three pairwise alignments.…”

Section: Multi-alignmentmentioning

confidence: 99%

Joint Generation of Transliterations from Multiple Representations

Yao

Kondrak

2015

Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua

Self Cite

View full text Add to dashboard Cite

Machine transliteration is often referred to as phonetic translation.We show that transliterations incorporate information from both spelling and pronunciation, and propose an effective model for joint transliteration generation from both representations. We further generalize this model to include transliterations from other languages, and enhance it with reranking and lexicon features. We demonstrate significant improvements in transliteration accuracy on several datasets.

show abstract

“…When visiting the Google search page, we may get 50 bursts, while we may obtain 200 bursts when viewing the CNN front page. If a profile HMM is to be constructed from a set of unaligned sequences, there are many local optima, and it is therefore easy for the constructing algorithm to get stuck around one of these [30]. So, for each application type, we should align its training samples to the same length.…”

Section: G He Et Almentioning

confidence: 99%

A novel application classification attack against Tor

Luo

2015

Concurrency and Computation

View full text Add to dashboard Cite

SUMMARYTor is a famous anonymous communication system for preserving users' online privacy. It supports TCP applications and packs upper-layer application data into encrypted equal-sized cells with onion routing to hide private information of users. However, we note that the current Tor design cannot conceal certain application behaviors. For example, P2P applications usually upload and download files simultaneously, and this behavioral feature is also kept in Tor traffic. Motivated by this observation, we investigate a new attack against Tor, application classification attack, which can recognize application types from Tor traffic. An attacker first carefully selects some flow features such as burst volumes and directions to represent the application behaviors and takes advantage of some efficient machine-learning algorithm (e.g., Profile Hidden Markov Model) to model different types of applications. Then he or she can use these established models to classify target's Tor traffic and infer its application type. We have implemented the application classification attack on Tor using parallel computing, and our experiments validate the feasibility and effectiveness of the attack. We argue that the disclosure of application type information is a serious threat to Tor users' anonymity because it can be used to reduce the anonymity set and facilitate other attacks. We also present guidelines to defend against application classification attack.

show abstract

Multiple word alignment with profile hidden Markov models

Cited by 20 publications

References 6 publications

Combining Regular Sound Correspondences and Geographic Spread

Combining Regular Sound Correspondences and Geographic Spread

Joint Generation of Transliterations from Multiple Representations

A novel application classification attack against Tor

Contact Info

Product

Resources

About