arXiv subjects

Language models

Find arXiv papers on language models and computational linguistics, including cs.CL metadata. Use abstracts to explore model training, evaluation and natural-language processing.

Follow this topic with RSS

At least 199 records · Page 11Linked to original sources

arXiv2018

hep-th

We apply techniques in natural language processing, computational linguistics, and machine-learning to investigate papers in hep-th and four related sections of the arXiv: hep-ph, hep-lat, gr-qc, and math-ph. All of the titles of papers in each of these sections, from the inception of the arXiv until the end of 2017, are extracted and treated as a corpus which we use to train the neural network Word2Vec. A comparative study of common n-grams, linear syntactical identities, word cloud and word similarities is carried out. We find notable scientific and sociological differences between the fields. In conjunction with support vector machines, we also show that the syntactic structure of the titles in different sub-fields of high energy and mathematical physics are sufficiently different that a neural network can perform a binary classification of formal versus phenomenological sections with 87.1% accuracy, and can perform a finer five-fold classification across all sections with 65.1% accuracy.

cs.CL↗

arXiv2018

Waveform to Single Sinusoid Regression to Estimate the F0 Contour from Noisy Speech Using Recurrent Deep Neural Networks

The fundamental frequency (F0) represents pitch in speech that determines prosodic characteristics of speech and is needed in various tasks for speech analysis and synthesis. Despite decades of research on this topic, F0 estimation at low signal-to-noise ratios (SNRs) in unexpected noise conditions remains difficult. This work proposes a new approach to noise robust F0 estimation using a recurrent neural network (RNN) trained in a supervised manner. Recent studies employ deep neural networks (DNNs) for F0 tracking as a frame-by-frame classification task into quantised frequency states but we propose waveform-to-sinusoid regression instead to achieve both noise robustness and accurate estimation with increased frequency resolution. Experimental results with PTDB-TUG corpus contaminated by additive noise (NOISEX-92) demonstrate that the proposed method improves gross pitch error (GPE) rate and fine pitch error (FPE) by more than 35 % at SNRs between -10 dB and +10 dB compared with well-known noise robust F0 tracker, PEFAC. Furthermore, the proposed method also outperforms state-of-the-art DNN-based approaches by more than 15 % in terms of both FPE and GPE rate over the preceding SNR range.

eess.AS↗

arXiv2018

Representation Mapping: A Novel Approach to Generate High-Quality Multi-Lingual Emotion Lexicons

In the past years, sentiment analysis has increasingly shifted attention to representational frameworks more expressive than semantic polarity (being positive, negative or neutral). However, these richer formats (like Basic Emotions or Valence-Arousal-Dominance, and variants therefrom), rooted in psychological research, tend to proliferate the number of representation schemes for emotion encoding. Thus, a large amount of representationally incompatible emotion lexicons has been developed by various research groups adopting one or the other emotion representation format. As a consequence, the reusability of these resources decreases as does the comparability of systems using them. In this paper, we propose to solve this dilemma by methods and tools which map different representation formats onto each other for the sake of mutual compatibility and interoperability of language resources. We present the first large-scale investigation of such representation mappings for four typologically diverse languages and find evidence that our approach produces (near-)gold quality emotion lexicons, even in cross-lingual settings. Finally, we use our models to create new lexicons for eight typologically diverse languages.

cs.CL↗

arXiv2018

Pragmatic approach to structured data querying via natural language interface

As the use of technology increases and data analysis becomes integral in many businesses, the ability to quickly access and interpret data has become more important than ever. Information retrieval technologies are being utilized by organizations and companies to manage their information systems and processes. Despite information retrieval of a large amount of data being efficient organized in relational databases, a user still needs to master the DB language/schema to completely formulate the queries. This puts a burden on organizations and companies to hire employees that are proficient in DB languages/schemas to formulate queries. To reduce some of the burden on already overstretched data teams, many organizations are looking for tools that allow non-developers to query their databases. Unfortunately, writing a valid SQL query that answers the question a user is trying to ask isn't always easy. Even seemingly simple questions, like "Which start-up companies received more than $200M in funding?" can actually be very hard to answer, let alone convert into a SQL query. How do you define start-up companies? By size, location, duration of time they have been incorporated? This may be fine if a user is working with a database they're already familiar with, but what if users are not familiar with the database. What is needed is a centralized system that can effectively translate natural language queries into specific database queries for different customer database types. There is a number of factors that can dramatically affect the system architecture and the set of algorithms used to translate NL queries into a structured query representation.

cs.CL↗

arXiv2021

DRAFT-What you always wanted to know but could not find about block-based environments

Block-based environments are visual programming environments, which are becoming more and more popular because of their ease of use. The ease of use comes thanks to their intuitive graphical representation and structural metaphors (jigsaw-like puzzles) to display valid combinations of language constructs to the users. Part of the current popularity of block-based environments is thanks to Scratch. As a result they are often associated with tools for children or young learners. However, it is unclear how these types of programming environments are developed and used in general. So we conducted a systematic literature review on block-based environments by studying 152 papers published between 2014 and 2020, and a non-systematic tool review of 32 block-based environments. In particular, we provide a helpful inventory of block-based editors for end-users on different topics and domains. Likewise, we focused on identifying the main components of block-based environments, how they are engineered, and how they are used. This survey should be equally helpful for language engineering researchers and language engineers alike.

cs.SE↗

arXiv2021

Introspective Distillation for Robust Question Answering

Question answering (QA) models are well-known to exploit data bias, e.g., the language prior in visual QA and the position bias in reading comprehension. Recent debiasing methods achieve good out-of-distribution (OOD) generalizability with a considerable sacrifice of the in-distribution (ID) performance. Therefore, they are only applicable in domains where the test distribution is known in advance. In this paper, we present a novel debiasing method called Introspective Distillation (IntroD) to make the best of both worlds for QA. Our key technical contribution is to blend the inductive bias of OOD and ID by introspecting whether a training sample fits in the factual ID world or the counterfactual OOD one. Experiments on visual QA datasets VQA v2, VQA-CP, and reading comprehension dataset SQuAD demonstrate that our proposed IntroD maintains the competitive OOD performance compared to other debiasing methods, while sacrificing little or even achieving better ID performance compared to the non-debiasing ones.

cs.CV↗

arXiv2021

GPU-Accelerated Forward-Backward algorithm with Application to Lattice-Free MMI

We propose to express the forward-backward algorithm in terms of operations between sparse matrices in a specific semiring. This new perspective naturally leads to a GPU-friendly algorithm which is easy to implement in Julia or any programming languages with native support of semiring algebra. We use this new implementation to train a TDNN with the LF-MMI objective function and we compare the training time of our system with PyChain - a recently introduced C++/CUDA implementation of the LF-MMI loss. Our implementation is about two times faster while not having to use any approximation such as the "leaky-HMM".

cs.DC↗

arXiv2022

Gender and Racial Stereotype Detection in Legal Opinion Word Embeddings

Studies have shown that some Natural Language Processing (NLP) systems encode and replicate harmful biases with potential adverse ethical effects in our society. In this article, we propose an approach for identifying gender and racial stereotypes in word embeddings trained on judicial opinions from U.S. case law. Embeddings containing stereotype information may cause harm when used by downstream systems for classification, information extraction, question answering, or other machine learning systems used to build legal research tools. We first explain how previously proposed methods for identifying these biases are not well suited for use with word embeddings trained on legal opinion text. We then propose a domain adapted method for identifying gender and racial biases in the legal domain. Our analyses using these methods suggest that racial and gender biases are encoded into word embeddings trained on legal opinions. These biases are not mitigated by exclusion of historical data, and appear across multiple large topical areas of the law. Implications for downstream systems that use legal opinion word embeddings and suggestions for potential mitigation strategies based on our observations are also discussed.

cs.CL↗

arXiv2022

Data-augmented cross-lingual synthesis in a teacher-student framework

Cross-lingual synthesis can be defined as the task of letting a speaker generate fluent synthetic speech in another language. This is a challenging task, and resulting speech can suffer from reduced naturalness, accented speech, and/or loss of essential voice characteristics. Previous research shows that many models appear to have insufficient generalization capabilities to perform well on every of these cross-lingual aspects. To overcome these generalization problems, we propose to apply the teacher-student paradigm to cross-lingual synthesis. While a teacher model is commonly used to produce teacher forced data, we propose to also use it to produce augmented data of unseen speaker-language pairs, where the aim is to retain essential speaker characteristics. Both sets of data are then used for student model training, which is trained to retain the naturalness and prosodic variation present in the teacher forced data, while learning the speaker identity from the augmented data. Some modifications to the student model are proposed to make the separation of teacher forced and augmented data more straightforward. Results show that the proposed approach improves the retention of speaker characteristics in the speech, while managing to retain high levels of naturalness and prosodic variation.

cs.SD↗

arXiv2022

Position-wise optimizer: A nature-inspired optimization algorithm

The human nervous system utilizes synaptic plasticity to solve optimization problems. Previous studies have tried to add the plasticity factor to the training process of artificial neural networks, but most of those models require complex external control over the network or complex novel rules. In this manuscript, a novel nature-inspired optimization algorithm is introduced that imitates biological neural plasticity. Furthermore, the model is tested on three datasets and the results are compared with gradient descent optimization.

cs.NE↗

arXiv2022

Unified Speech-Text Pre-training for Speech Translation and Recognition

We describe a method to jointly pre-train speech and text in an encoder-decoder modeling framework for speech translation and recognition. The proposed method incorporates four self-supervised and supervised subtasks for cross modality learning. A self-supervised speech subtask leverages unlabelled speech data, and a (self-)supervised text to text subtask makes use of abundant text training data. Two auxiliary supervised speech tasks are included to unify speech and text modeling space. Our contribution lies in integrating linguistic information from the text corpus into the speech pre-training. Detailed analysis reveals learning interference among subtasks. Two pre-training configurations for speech translation and recognition, respectively, are presented to alleviate subtask interference. Our experiments show the proposed method can effectively fuse speech and text information into one model. It achieves between 1.7 and 2.3 BLEU improvement above the state of the art on the MuST-C speech translation dataset and comparable WERs to wav2vec 2.0 on the Librispeech speech recognition task.

cs.CL↗

arXiv2022

Easy Adaptation to Mitigate Gender Bias in Multilingual Text Classification

Existing approaches to mitigate demographic biases evaluate on monolingual data, however, multilingual data has not been examined. In this work, we treat the gender as domains (e.g., male vs. female) and present a standard domain adaptation model to reduce the gender bias and improve performance of text classifiers under multilingual settings. We evaluate our approach on two text classification tasks, hate speech detection and rating prediction, and demonstrate the effectiveness of our approach with three fair-aware baselines.

cs.CL↗

arXiv2022

Not always about you: Prioritizing community needs when developing endangered language technology

Languages are classified as low-resource when they lack the quantity of data necessary for training statistical and machine learning tools and models. Causes of resource scarcity vary but can include poor access to technology for developing these resources, a relatively small population of speakers, or a lack of urgency for collecting such resources in bilingual populations where the second language is high-resource. As a result, the languages described as low-resource in the literature are as different as Finnish on the one hand, with millions of speakers using it in every imaginable domain, and Seneca, with only a small-handful of fluent speakers using the language primarily in a restricted domain. While issues stemming from the lack of resources necessary to train models unite this disparate group of languages, many other issues cut across the divide between widely-spoken low resource languages and endangered languages. In this position paper, we discuss the unique technological, cultural, practical, and ethical challenges that researchers and indigenous speech community members face when working together to develop language technology to support endangered language documentation and revitalization. We report the perspectives of language teachers, Master Speakers and elders from indigenous communities, as well as the point of view of academics. We describe an ongoing fruitful collaboration and make recommendations for future partnerships between academic researchers and language community stakeholders.

cs.CL↗

arXiv2022

Solving Price Per Unit Problem Around the World: Formulating Fact Extraction as Question Answering

Price Per Unit (PPU) is an essential information for consumers shopping on e-commerce websites when comparing products. Finding total quantity in a product is required for computing PPU, which is not always provided by the sellers. To predict total quantity, all relevant quantities given in a product attributes such as title, description and image need to be inferred correctly. We formulate this problem as a question-answering (QA) task rather than named entity recognition (NER) task for fact extraction. In our QA approach, we first predict the unit of measure (UoM) type (e.g., volume, weight or count), that formulates the desired question (e.g., "What is the total volume?") and then use this question to find all the relevant answers. Our model architecture consists of two subnetworks for the two subtasks: a classifier to predict UoM type (or the question) and an extractor to extract the relevant quantities. We use a deep character-level CNN architecture for both subtasks, which enables (1) easy expansion to new stores with similar alphabets, (2) multi-span answering due to its span-image architecture and (3) easy deployment by keeping model-inference latency low. Our QA approach outperforms rule-based methods by 34.4% in precision and also BERT-based fact extraction approach in all stores globally, with largest precision lift of 10.6% in the US store.

cs.CL↗

arXiv2022

Stylized Knowledge-Grounded Dialogue Generation via Disentangled Template Rewriting

Current Knowledge-Grounded Dialogue Generation (KDG) models specialize in producing rational and factual responses. However, to establish long-term relationships with users, the KDG model needs the capability to generate responses in a desired style or attribute. Thus, we study a new problem: Stylized Knowledge-Grounded Dialogue Generation (SKDG). It presents two challenges: (1) How to train a SKDG model where no triples are available. (2) How to cohere with context and preserve the knowledge when generating a stylized response. In this paper, we propose a novel disentangled template rewriting (DTR) method which generates responses via combing disentangled style templates (from monolingual stylized corpus) and content templates (from KDG corpus). The entire framework is end-to-end differentiable and learned without supervision. Extensive experiments on two benchmarks indicate that DTR achieves a significant improvement on all evaluation metrics compared with previous state-of-the-art stylized dialogue generation methods. Besides, DTR achieves comparable performance with the state-of-the-art KDG methods in standard KDG evaluation setting.

cs.CL↗

arXiv2022

ASR in German: A Detailed Error Analysis

The amount of freely available systems for automatic speech recognition (ASR) based on neural networks is growing steadily, with equally increasingly reliable predictions. However, the evaluation of trained models is typically exclusively based on statistical metrics such as WER or CER, which do not provide any insight into the nature or impact of the errors produced when predicting transcripts from speech input. This work presents a selection of ASR model architectures that are pretrained on the German language and evaluates them on a benchmark of diverse test datasets. It identifies cross-architectural prediction errors, classifies those into categories and traces the sources of errors per category back into training data as well as other sources. Finally, it discusses solutions in order to create qualitatively better training datasets and more robust ASR systems.

cs.CL↗

arXiv2022

X-DETR: A Versatile Architecture for Instance-wise Vision-Language Tasks

In this paper, we study the challenging instance-wise vision-language tasks, where the free-form language is required to align with the objects instead of the whole image. To address these tasks, we propose X-DETR, whose architecture has three major components: an object detector, a language encoder, and vision-language alignment. The vision and language streams are independent until the end and they are aligned using an efficient dot-product operation. The whole network is trained end-to-end, such that the detector is optimized for the vision-language tasks instead of an off-the-shelf component. To overcome the limited size of paired object-language annotations, we leverage other weak types of supervision to expand the knowledge coverage. This simple yet effective architecture of X-DETR shows good accuracy and fast speeds for multiple instance-wise vision-language tasks, e.g., 16.4 AP on LVIS detection of 1.2K categories at ~20 frames per second without using any LVIS annotation during training.

cs.CV↗

arXiv2022

Idiomify -- Building a Collocation-supplemented Reverse Dictionary of English Idioms with Word2Vec for non-native learners

The aim of idiomify is to build a collocation-supplemented reverse dictionary of idioms for the non-native learners of English. We aim to do so because the reverse dictionary could help the non-natives explore idioms on demand, and the collocations could also guide them on using idioms more adequately. The cornerstone of the project is a reliable way of mining idioms from corpora, which is however a challenge because idioms extensively vary in forms. We tackle this by automatically deriving matching rules from their base forms. We use Point-wise Mutual Inclusion (PMI), Term Frequency - Inverse Document Frequency (TF-IDF) to model collocations, since both of them are popular metric for pairwise significance. We also try Term Frequency (TF) as the baseline model. As for implementing the reverse-dictionary, three approaches could be taken: inverted index, graphs and distributional semantics. We choose to take the last approach and implement the reverse dictionary with Word2Vec, because it is the most flexible approach of all and Word2Vec is a simple yet strong baseline. Evaluating the methods has revealed rooms for improvement. We learn that we can better identify idioms with the help of slop, wildcard and reordering techniques. We also learn that we can get the best of both PMI and TF-IDF if we use machine learning to find the sweet spot. Lastly, We learn that Idiomify could be further improved with a mixture of inverted index and distributional semantics approach. The limits aside, the proposed methods are feasible, and their benefits to the non-natives are apparent, which therefore can be used to aid the non-natives in acquiring English idioms.

cs.CL↗