Sravana Reddy

---

Research

Publications

• Sravana Reddy and Kevin Knight (2012). Decoding Running Key Ciphers. In Proceedings of ACL.
  [paper] [abstract] [bib]

  Abstract

  There has been recent interest in the problem of decoding letter substitution ciphers using techniques inspired by natural language processing. We consider a different type of classical encoding scheme known as the running key cipher, and propose a search solution using Gibbs sampling with a word language model. We evaluate our method on synthetic ciphertexts of different lengths, and find that it outperforms previous work that employs Viterbi decoding with character-based models.

  .bib

  @inproceedings{Reddy:2012,
    title = {Decoding Running Key Ciphers},
    author = {Sravana Reddy and Kevin Knight},
    booktitle = {ACL},
    year = {2012},
  }
  

• Sonjia Waxmonsky and Sravana Reddy (2012). G2P Conversion of Proper Names Using Word Origin Information. In Proceedings of NAACL.
  [paper] [abstract] [bib] [word origin data] [poster]

  Abstract

  Motivated by the fact that the pronunciation of a name may be influenced by its language of origin, we present methods to improve pronunciation prediction of proper names using word origin information. We train grapheme-to-phoneme (G2P) models on language-specific data sets and interpolate the outputs. We perform experiments on US personal surnames, a data set where word origin variation occurs naturally. Our methods can be used with any G2P algorithm that outputs posterior probabilities of phoneme sequences for a given word.

  .bib

  @inproceedings{Waxmonsky:2012,
    title = {G2{P} Conversion of Proper Names Using Word Origin Information},
    author = {Sonjia Waxmonsky and Sravana Reddy},
    booktitle = {NAACL},
    year = {2012},
  }
  

• Sravana Reddy and Evandro Gouvêa (2011). Learning from Mistakes: Expanding Pronunciation Lexicons Using Word Recognition Errors. In Proceedings of Interspeech.
  [paper] [abstract] [bib] [slides]

  Abstract

  We introduce the problem of learning pronunciations of out-of-vocabulary words from word recognition mistakes made by an automatic speech recognition (ASR) system. This question is especially relevant in cases where the ASR engine is a black box -- meaning that the only acoustic cues about the speech data come from the word recognition outputs. This paper presents an expectation maximization approach to inferring pronunciations from ASR word recognition hypotheses, which outperforms pronunciation estimates of a state of the art grapheme-to-phoneme system.

  .bib

  @inproceedings{Reddy:2011c,
    title = {Learning from Mistakes: Expanding Pronunciation Lexicons Using Word Recognition Errors},
    author = {Sravana Reddy and Evandro Gouv\^{e}a},
    booktitle = {Interspeech},
    year = {2011},
  }
  

• Sravana Reddy and Kevin Knight (2011). Unsupervised Discovery of Rhyme Schemes. In Proceedings of ACL.
  [paper] [abstract] [bib] [slides] [rhyming corpus] [code]

  Abstract

  This paper describes an unsupervised, language-independent model for finding rhyme schemes in poetry, using no prior knowledge about rhyme or pronunciation.

  .bib

  @inproceedings{Reddy:2011a,
    title = {Unsupervised Discovery of Rhyme Schemes},
    author = {Sravana Reddy and Kevin Knight},
    booktitle = {ACL},
    year = {2011},
  }
  

• Sravana Reddy and Kevin Knight (2011). What We Know About The Voynich Manuscript. In Proceedings of the ACL Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities.
  [paper] [abstract] [bib] [slides] [note]

  Abstract

  The Voynich Manuscript is an undeciphered document from medieval Europe. We present current knowledge about the manuscript's text through a series of questions about its linguistic properties.

  .bib

  @inproceedings{Reddy:2011b,
    title = {What We Know About The {V}oynich {M}anuscript},
    author = {Sravana Reddy and Kevin Knight},
    booktitle = {ACL Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities},
    year = {2011},
  }
  

  A note on the conclusions of our work

  We have presented a statistical overview of the text that covers a number of questions; however, this is obviously not comprehensive, and there are many more avenues to explore. One point that has been raised about this paper is that we claim that the characters represent alphabets in some writing system. This is not the case. We do focus our analysis around this possibility (since it's the most obvious one on a first look at the manuscript) and present some interpretations accordingly, but we are careful to highlight the statistics, which are plain truths, and remain agnostic about what they imply. In particular, we do not arrive at a conclusion about the nature of the writing system or the underlying text.

• Sravana Reddy and John Goldsmith. (2010). An MDL-based Approach to Extracting Subword Units for Grapheme-to-Phoneme Conversion. In Proceedings of NAACL.
  [paper] [abstract] [bib]

  Abstract

  We address a key problem in grapheme-to-phoneme conversion: the ambiguity in mapping grapheme units to phonemes. Rather than using single letters and phonemes as units, we propose learning chunks, or subwords, to reduce ambiguity. This can be interpreted as learning a lexicon of subwords that has minimum description length. We implement an algorithm to build such a lexicon, as well as a simple decoder that uses these subwords.

  .bib

  @inproceedings{Reddy:2010,
    title = {An {MDL}-based Approach to Extracting Subword Units for Grapheme-to-Phoneme Conversion},
    author = {Sravana Reddy and John Goldsmith},
    booktitle = {NAACL},
    year = {2010},
  }
  

• Sravana Reddy and Sonjia Waxmonsky (2009). Substring-based Transliteration with Conditional Random Fields. In Proceedings of the ACL Named Entities Workshop (Shared Task).
  [paper] [abstract] [bib]

  Abstract

  Motivated by phrase-based translation research, we present a transliteration system where characters are grouped into substrings to be mapped atomically into the target language. We show how this substring representation can be incorporated into a Conditional Random Field model that uses local context and phonemic information. Our training and test data consists of three sets: English to Hindi, English to Kannada, and English to Tamil (Kumaran and Kellner, 2007) – from the NEWS 2009 Machine Transliteration Shared Task (Li et al., 2009).

  .bib

  @inproceedings{Reddy:2009b,
    title = {Substring-based Transliteration with Conditional Random Fields},
    author = {Sravana Reddy and Sonjia Waxmonsky},
    booktitle = {ACL Named Entities Workshop},
    year = {2009},
  }
  

• Sravana Reddy (2009). Understanding Eggcorns. In Proceedings of the NAACL Workshop on Computational Approaches to Linguistic Creativity.
  [paper] [abstract] [bib]

  Abstract

  An eggcorn is a type of linguistic error where a word is substituted with one that is semantically plausible – that is, the substitution is a semantic reanalysis of what may be a rare, archaic, or otherwise opaque term. We build a system that, given the original word and its eggcorn form, finds a semantic path between the two. Based on these paths, we derive a typology that reflects the different classes of semantic reinterpretation underlying eggcorns.

  .bib

  @inproceedings{Reddy:2009a,
    title = {Understanding Eggcorns},
    author = {Sravana Reddy},
    booktitle = {NAACL Workshop on Computational Approaches to Linguistic Creativity},
    year = {2009},
  }
  

Theses

• Learning Pronunciations from Unlabeled Evidence. Doctoral Dissertation, The University of Chicago, 2012.
  [e-mail for copy] [abstract] [bib]

  Abstract

  The pronunciation of a word represented in an alphabetic writing system (such as this one) is relatively transparent -- but a language's sounds change over time and vary across space, while its spellings remain relatively static, resulting in some amount of divergence between the written and spoken forms. The introduction of loanwords and proper names from other languages with different phonologies and scripts further complicates the relationship between orthography and pronunciation.

  However, there are sources of information about the sound of a word in addition to spelling. This dissertation presents methods for learning pronunciations using three forms of evidence: speech, word origins, and rhymes.

  Speech is the most natural source of evidence: a word's pronunciation is usually clarified upon hearing it in a spoken utterance. In the case of proper names, knowing the linguistic or ethnic origin of the name is often instrumental in determining how it should be pronounced. Rhymes in poems or songs also provide a cue to pronunciation, particularly for inferring the sound of a word at an earlier point in history.

  Using extra-orthographic evidence in a computational model necessitates access to data that provides this information in sufficiently large quantities. Collecting annotated data -- speech labeled with the words that it contains, names with their ethnic origins, or poetry with rhyming patterns -- can be extremely difficult and expensive. On the other hand, raw data -- speech recordings, lists of names, archives of poetry -- is available in plenty. The focus of this dissertation is, therefore, on using unlabeled data for pronunciation learning.

  There are two facets of 'pronunciation learning' in this work. One is that of converting the written form of a word to its phonemic representation, known as grapheme-to-phoneme conversion. The other is a less fine-grained objective: rather than learning the exact phonemic forms, we aim to discover clusters of words with similar pronunciations.

  The first contribution of this dissertation is a model to incorporate untranscribed spoken data, in addition to an existing lexicon of words and their pronunciations, into a grapheme-to-phoneme learner. This approach is found to improve over models that are trained only on a lexicon. The second contribution is an algorithm that models the latent linguistic origins of names as part of grapheme-to-phoneme conversion. Not only does this method do better than a model without word origin awareness, it also outperforms existing approaches that first classify orthographic words by origin in a supervised manner, and then train language-specific pronunciation models. The final contribution is an unsupervised learner that discovers the rhyme schemes of stanzas of poetry, as well as rhyming relations between words and clusters of words with a common rhyming sound.

  .bib

  @phdthesis{Reddy:2012,
    title = {Learning Pronunciations from Unlabeled Evidence},
    author = {Sravana Reddy},
    school = {The University of Chicago},
    year = {2012},
  }
  

• Part of Speech Induction Using Non-Negative Matrix Factorization. Masters' Thesis, The University of Chicago, 2009.
  [e-mail for copy] [abstract] [bib]

  Abstract

  Unsupervised part-of-speech induction involves the discovery of syntactic categories in a text, given no additional information other than the text itself. One requirement of an induction system is the ability to handle multiple categories for each word, in order to deal with word sense ambiguity. We construct an algorithm for unsupervised part-of-speech induction, treating the problem as one of soft clustering. The key technical component of the algorithm is the application of the recently developed technique of non-negative matrix factorization to the task of category discovery, using word contexts and morphology as syntactic cues.

  .bib

  @mastersthesis{Reddy:2009,
    title = {Part of Speech Induction Using Non-Negative Matrix Factorization},
    author = {Sravana Reddy},
    school = {The University of Chicago},
    year = {2009},
  }
  

Teaching

Instructor (Dartmouth)

• Computational Linguistics (Winter 2013)

Lab Instructor (UChicago)

• Fundamentals of Programming (Fall 2011)
• Intro to Computer Science (Fall 2009, Fall 2010)
  
• Intro to Programming for the World Wide Web (Spring 2010)
• Distributed Objects (Spring 2009)
  

Teaching Assistant (UChicago)

• AI: Computational Linguistics, Artifical Intelligence, Computational Biology
• Programming: Intro to Computer Science, Fundamentals of Programming, Intro to WWW Programming
• Other: Foundations of Software

Education

I received a Ph.D. in Computer Science from The University of Chicago in 2012, and an M.S. from the same department along the way in '09. My undergraduate education ('02 - '06) was at Brandeis University under the Wien Scholarship, where I majored in Computer Science, Math, and Creative Writing.

Many of my summers were spent interning: 2010 and 2011 with the natural language group at the USC Information Sciences Institute, where I worked mainly with Kevin Knight, and 2009 with the speech group at Mitsubishi Electric Research Labs (MERL) under Evandro Gouvêa. In the more distant past, I spent the summer of 2006 at the Perseus Digital Library at Tufts, working on Latin OCR, and 2005 at the CMU Robotics Institute. I grew up in Bangalore.