A large number of people are forced to use the Web in a language they have low literacy in due to technology asymmetries. Written text in the second language (L2) from such users often contains a large number of errors that are influenced by their native language (L1).We propose a method to mine phoneme confusions (sounds in L2 that an L1 speaker is likely to conflate) for pairs of L1 and L2.These confusions are then plugged into a generative model (Bi-Phone) for synthetically producing corrupted L2 text. Through human evaluations, we show that Bi-Phone generates plausible corruptions that differ across L1s and also have widespread coverage on the Web.We also corrupt the popular language understanding benchmark SuperGLUE with our technique (FunGLUE for Phonetically Noised GLUE) and show that SoTA language understating models perform poorly. We also introduce a new phoneme prediction pre-training task which helps byte models to recover performance close to SuperGLUE. Finally, we also release the SuperGLUE benchmark to promote further research in phonetically robust language models. To the best of our knowledge, FunGLUE is the first benchmark to introduce L1-L2 interactions in text.

Natural language processing is largely focused on written text processing. However, many computational linguists tacitly endorse myths about the nature of writing. We highlight two of these myths—the conflation of language and writing, and the notion that Chinese, Japanese, and Korean writing is ideographic—and suggest how the community can dispel them.

Word error rate (WER) and character error rate (CER) are standard metrics in Speech Recognition (ASR), but one problem has always been alternative spellings: If one’s system transcribes adviser whereas the ground truth has advisor, this will count as an error even though the two spellings really represent the same word. Japanese is notorious for “lacking orthography”: most words can be spelled in multiple ways, presenting a problem for accurate ASR evaluation. In this paper we propose a new lenient evaluation metric as a more defensible CER measure for Japanese ASR. We create a lattice of plausible respellings of the reference transcription, using a combination of lexical resources, a Japanese text-processing system, and a neural machine translation model for reconstructing kanji from hiragana or katakana. In a manual evaluation, raters rated 95.4% of the proposed spelling variants as plausible. ASR results show that our method, which does not penalize the system for choosing a valid alternate spelling of a word, affords a 2.4%–3.1% absolute reduction in CER depending on the task.

If one sees the place name Houston Mercer Dog Run in New York, how does one know how to pronounce it? Assuming one knows that Houston in New York is pronounced /ˈhaʊstən/ and not like the Texas city (/ˈhjuːstən/), then one can probably guess that /ˈhaʊstən/ is also used in the name of the dog park. We present a novel architecture that learns to use the pronunciations of neighboring names in order to guess the pronunciation of a given target feature. Applied to Japanese place names, we demonstrate the utility of the model to finding and proposing corrections for errors in Google Maps. To demonstrate the utility of this approach to structurally similar problems, we also report on an application to a totally different task: Cognate reflex prediction in comparative historical linguistics. A version of the code has been open-sourced.1

The SIGTYP 2022 shared task concerns the problem of word reflex generation in a target language, given cognate words from a subset of related languages. We present two systems to tackle this problem, covering two very different modeling approaches. The first model extends transformer-based encoder-decoder sequence-to-sequence modeling, by encoding all available input cognates in parallel, and having the decoder attend to the resulting joint representation during inference. The second approach takes inspiration from the field of image restoration, where models are tasked with recovering pixels in an image that have been masked out. For reflex generation, the missing reflexes are treated as “masked pixels” in an “image” which is a representation of an entire cognate set across a language family. As in the image restoration case, cognate restoration is performed with a convolutional network.

In a recent position paper, Turing Award Winners Yoshua Bengio, Geoffrey Hinton, and Yann LeCun make the case that symbolic methods are not needed in AI and that, while there are still many issues to be resolved, AI will be solved using purely neural methods. In this piece I issue a challenge: Demonstrate that a purely neural approach to the problem of text normalization is possible. Various groups have tried, but so far nobody has eliminated the problem of unrecoverable errors, errors where, due to insufficient training data or faulty generalization, the system substitutes some other reading for the correct one. Solutions have been proposed that involve a marriage of traditional finite-state methods with neural models, but thus far nobody has shown that the problem can be solved using neural methods alone. Though text normalization is hardly an “exciting” problem, I argue that until one can solve “boring” problems like that using purely AI methods, one cannot claim that AI is a success.

This paper presents an open-source software library that provides a set of finite-state transducer (FST) components and corresponding utilities for manipulating the writing systems of languages that use the Perso-Arabic script. The operations include various levels of script normalization, including visual invariance-preserving operations that subsume and go beyond the standard Unicode normalization forms, as well as transformations that modify the visual appearance of characters in accordance with the regional orthographies for eleven contemporary languages from diverse language families. The library also provides simple FST-based romanization and transliteration. We additionally attempt to formalize the typology of Perso-Arabic characters by providing one-to-many mappings from Unicode code points to the languages that use them. While our work focuses on the Arabic script diaspora rather than Arabic itself, this approach could be adopted for any language that uses the Arabic script, thus providing a unified framework for treating a script family used by close to a billion people.

Taxonomies of writing systems since Gelb (1952) have classified systems based on what the written symbols represent: if they represent words or morphemes, they are logographic; if syllables, syllabic; if segments, alphabetic; and so forth. Sproat (2000) and Rogers (2005) broke with tradition by splitting the logographic and phonographic aspects into two dimensions, with logography being graded rather than a categorical distinction. A system could be syllabic, and highly logographic; or alphabetic, and mostly non-logographic. This accords better with how writing systems actually work, but neither author proposed a method for measuring logography. In this article we propose a novel measure of the degree of logography that uses an attention-based sequence-to-sequence model trained to predict the spelling of a token from its pronunciation in context. In an ideal phonographic system, the model should need to attend to only the current token in order to compute how to spell it, and this would show in the attention matrix activations. In contrast, with a logographic system, where a given pronunciation might correspond to several different spellings, the model would need to attend to a broader context. The ratio of the activation outside the token and the total activation forms the basis of our measure. We compare this with a simple lexical measure, and an entropic measure, as well as several other neural models, and argue that on balance our attention-based measure accords best with intuition about how logographic various systems are. Our work provides the first quantifiable measure of the notion of logography that accords with linguistic intuition and, we argue, provides better insight into what this notion means.

Ad hoc abbreviations are commonly found in informal communication channels that favor shorter messages. We consider the task of reversing these abbreviations in context to recover normalized, expanded versions of abbreviated messages. The problem is related to, but distinct from, spelling correction, as ad hoc abbreviations are intentional and can involve more substantial differences from the original words. Ad hoc abbreviations are also productively generated on-the-fly, so they cannot be resolved solely by dictionary lookup. We generate a large, open-source data set of ad hoc abbreviations. This data is used to study abbreviation strategies and to develop two strong baselines for abbreviation expansion.

Breaking domain names such as openresearch into component words open and research is important for applications like Text-to-Speech synthesis and web search. We link this problem to the classic problem of Chinese word segmentation and show the effectiveness of a tagging model based on Recurrent Neural Networks (RNNs) using characters as input. To compensate for the lack of training data, we propose a pre-training method on concatenated entity names in a large knowledge database. Pre-training improves the model by 33% and brings the sequence accuracy to 85%.

This paper describes the NEMO submission to SIGTYP 2020 shared task (Bjerva et al., 2020) which deals with prediction of linguistic typological features for multiple languages using the data derived from World Atlas of Language Structures (WALS). We employ frequentist inference to represent correlations between typological features and use this representation to train simple multi-class estimators that predict individual features. We describe two submitted ridge regression-based configurations which ranked second and third overall in the constrained task. Our best configuration achieved the microaveraged accuracy score of 0.66 on 149 test languages.

Machine learning, including neural network techniques, have been applied to virtually every domain in natural language processing. One problem that has been somewhat resistant to effective machine learning solutions is text normalization for speech applications such as text-to-speech synthesis (TTS). In this application, one must decide, for example, that 123 is verbalized as one hundred twenty three in 123 pages but as one twenty three in 123 King Ave. For this task, state-of-the-art industrial systems depend heavily on hand-written language-specific grammars. We propose neural network models that treat text normalization for TTS as a sequence-to-sequence problem, in which the input is a text token in context, and the output is the verbalization of that token. We find that the most effective model, in accuracy and efficiency, is one where the sentential context is computed once and the results of that computation are combined with the computation of each token in sequence to compute the verbalization. This model allows for a great deal of flexibility in terms of representing the context, and also allows us to integrate tagging and segmentation into the process. These models perform very well overall, but occasionally they will predict wildly inappropriate verbalizations, such as reading 3 cm as three kilometers. Although rare, such verbalizations are a major issue for TTS applications. We thus use finite-state covering grammars to guide the neural models, either during training and decoding, or just during decoding, away from such “unrecoverable” errors. Such grammars can largely be learned from data.

Attention-based sequence-to-sequence neural network models learn to jointly align and translate. The quadratic-time attention mechanism is powerful as it is capable of handling arbitrary long-distance reordering, but computationally expensive. In this paper, towards making neural translation both accurate and efficient, we follow the traditional pre-reordering approach to decouple reordering from translation. We add a reordering RNN that shares the input encoder with the decoder. The RNNs are trained jointly with a multi-task loss function and applied sequentially at inference time. The task of the reordering model is to predict the permutation of the input words following the target language word order. After reordering, the attention in the decoder becomes more peaked and monotonic. For reordering, we adopt the Inversion Transduction Grammars (ITG) and propose a transition system to parse input to trees for reordering. We harness the ITG transition system with RNN. With the modeling power of RNN, we achieve superior reordering accuracy without any feature engineering. In experiments, we apply the model to the task of text normalization. Compared to a strong baseline of attention-based RNN, our ITG RNN re-ordering model can reach the same reordering accuracy with only 1/10 of the training data and is 2.5x faster in decoding.

We propose two models for verbalizing numbers, a key component in speech recognition and synthesis systems. The first model uses an end-to-end recurrent neural network. The second model, drawing inspiration from the linguistics literature, uses finite-state transducers constructed with a minimal amount of training data. While both models achieve near-perfect performance, the latter model can be trained using several orders of magnitude less data than the former, making it particularly useful for low-resource languages.

We present a text-to-speech (TTS) system designed for the dialect of Bengali spoken in Bangladesh. This work is part of an ongoing effort to address the needs of under-resourced languages. We propose a process for streamlining the bootstrapping of TTS systems for under-resourced languages. First, we use crowdsourcing to collect the data from multiple ordinary speakers, each speaker recording small amount of sentences. Second, we leverage an existing text normalization system for a related language (Hindi) to bootstrap a linguistic front-end for Bangla. Third, we employ statistical techniques to construct multi-speaker acoustic models using Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) and Hidden Markov Model (HMM) approaches. We then describe our experiments that show that the resulting TTS voices score well in terms of their perceived quality as measured by Mean Opinion Score (MOS) evaluations.

Which languages convey the most information in a given amount of space? This is a question often asked of linguists, especially by engineers who often have some information theoretic measure of “information” in mind, but rarely define exactly how they would measure that information. The question is, in fact remarkably hard to answer, and many linguists consider it unanswerable. But it is a question that seems as if it ought to have an answer. If one had a database of close translations between a set of typologically diverse languages, with detailed marking of morphosyntactic and morphosemantic features, one could hope to quantify the differences between how these different languages convey information. Since no appropriate database exists we decided to construct one. The purpose of this paper is to present our work on the database, along with some preliminary results. We plan to release the dataset once complete.

We describe ScriptTranscriber, an open source toolkit for extracting transliterations in comparable corpora from languages written in different scripts. The system includes various methods for extracting potential terms of interest from raw text, for providing guesses on the pronunciations of terms, and for comparing two strings as possible transliterations using both phonetic and temporal measures. The system works with any script in the Unicode Basic Multilingual Plane and is easily extended to include new modules. Given comparable corpora, such as newswire text, in a pair of languages that use different scripts, ScriptTranscriber provides an easy way to mine transliterations from the comparable texts. This is particularly useful for underresourced languages, where training data for transliteration may be lacking, and where it is thus hard to train good transliterators. ScriptTranscriber provides an open source package that allows for ready incorporation of more sophisticated modules ― e.g. a trained transliteration model for a particular language pair. ScriptTranscriber is available as part of the nltk contrib source tree at http://code.google.com/p/nltk/.

Processing of Colloquial Arabic is a relatively new area of research, and a number of interesting challenges pertaining to spoken Arabic dialects arise. On the one hand, a whole continuum of Arabic dialects exists, with linguistic differences on phonological, morphological, syntactic, and lexical levels. On the other hand, there are inter-dialectal similarities that need be explored. Furthermore, due to scarcity of dialect-specific linguistic resources and availability of a wide range of resources for Modern Standard Arabic (MSA), it is desirable to explore the possibility of exploiting MSA tools when working on dialects. This paper describes challenges in processing of Colloquial Arabic in the context of language modeling for Automatic Speech Recognition. Using data from Egyptian Colloquial Arabic and MSA, we investigate the question of improving language modeling of Egyptian Arabic with MSA data and resources. As part of the project, we address the problem of linguistic variation between Egyptian Arabic and MSA. To account for differences between MSA and Colloquial Arabic, we experiment with the following techniques of data transformation: morphological simplification (stemming), lexical transductions, and syntactic transformations. While the best performing model remains the one built using only dialectal data, these techniques allow us to obtain an improvement over the baseline MSA model. More specifically, while the effect on perplexity of syntactic transformations is not very significant, stemming of the training and testing data improves the baseline perplexity of the MSA model trained on words by 51%, and lexical transductions yield an 82% perplexity reduction. Although the focus of the present work is on language modeling, we believe the findings of the study will be useful for researchers involved in other areas of processing Arabic dialects, such as parsing and machine translation.