Language Understanding Models Research Articles

Child-Centric Robot Dialogue Systems: Fine-Tuning Large Language Models for Better Utterance Understanding and Interaction.

Dialogue systems must understand children's utterance intentions by considering their unique linguistic characteristics, such as syntactic incompleteness, pronunciation inaccuracies, and creative expressions, to enable natural conversational engagement in child-robot interactions. Even state-of-the-art large language models (LLMs) for language understanding and contextual awareness cannot comprehend children's intent as accurately as humans because of their distinctive features. An LLM-based dialogue system should acquire the manner by which humans understand children's speech to enhance its intention reasoning performance in verbal interactions with children. To this end, we propose a fine-tuning methodology that utilizes the LLM-human judgment discrepancy and interactive response data. The former data represent cases in which the LLM and human judgments of the contextual appropriateness of a child's answer to a robot's question diverge. The latter data involve robot responses suitable for children's utterance intentions, generated by the LLM. We developed a fine-tuned dialogue system using these datasets to achieve human-like interpretations of children's utterances and to respond adaptively. Our system was evaluated through human assessment using the Robotic Social Attributes Scale (RoSAS) and Sensibleness and Specificity Average (SSA) metrics. Consequently, it supports the effective interpretation of children's utterance intentions and enables natural verbal interactions, even in cases with syntactic incompleteness and mispronunciations.

A simple yet effective self-debiasing framework for transformer models

Current Transformer-based natural language understanding (NLU) models heavily rely on dataset biases, while failing to handle real-world out-of-distribution (OOD) instances. Many methods have been proposed to deal with this issue, but they ignore the fact that the features learned in different layers of Transformer-based NLU models are different. In this paper, we first conduct preliminary studies to obtain two conclusions: 1) both low- and high-layer sentence representations encode common biased features during training; 2) the low-layer sentence representations encode fewer unbiased features than the high-layer ones. Based on these conclusions, we propose a simple yet effective self-debiasing framework for Transformer-based NLU models. Concretely, we first stack a classifier on a selected low layer. Then, we introduce a residual connection that feeds the low-layer sentence representation to the top-layer classifier. In this way, the top-layer sentence representation will be trained to ignore the common biased features encoded by the low-layer sentence representation and focus on task-relevant unbiased features. During inference, we remove the residual connection and directly use the top-layer sentence representation to make predictions. Extensive experiments and in-depth analyses on NLU tasks demonstrate the superiority of our framework, achieving a new state-of-the-art (SOTA) on three OOD test sets.

HC 2L: Hybrid and Cooperative Contrastive Learning for Cross-Lingual Spoken Language Understanding.

State-of-the-art model for zero-shot cross-lingual spoken language understanding performs cross-lingual unsupervised contrastive learning to achieve the label-agnostic semantic alignment between each utterance and its code-switched data. However, it ignores the precious intent/slot labels, whose label information is promising to help capture the label-aware semantics structure and then leverage supervised contrastive learning to improve both source and target languages' semantics. In this paper, we propose Hybrid and Cooperative Contrastive Learning to address this problem. Apart from cross-lingual unsupervised contrastive learning, we design a holistic approach that exploits source language supervised contrastive learning, cross-lingual supervised contrastive learning and multilingual supervised contrastive learning to perform label-aware semantics alignments in a comprehensive manner. Each kind of supervised contrastive learning mechanism includes both single-task and joint-task scenarios. In our model, one contrastive learning mechanism's input is enhanced by others. Thus the total four contrastive learning mechanisms are cooperative to learn more consistent and discriminative representations in the virtuous cycle during the training process. Experiments show that our model obtains consistent improvements over 9 languages, achieving new state-of-the-art performance.

Cascaded cross-modal transformer for audio–textual classification

Speech classification tasks often require powerful language understanding models to grasp useful features, which becomes problematic when limited training data is available. To attain superior classification performance, we propose to harness the inherent value of multimodal representations by transcribing speech using automatic speech recognition models and translating the transcripts into different languages via pretrained translation models. We thus obtain an audio–textual (multimodal) representation for each data sample. Subsequently, we combine language-specific Bidirectional Encoder Representations from Transformers with Wav2Vec2.0 audio features via a novel cascaded cross-modal transformer (CCMT). Our model is based on two cascaded transformer blocks. The first one combines text-specific features from distinct languages, while the second one combines acoustic features with multilingual features previously learned by the first transformer block. We employed our system in the Requests Sub-Challenge of the ACM Multimedia 2023 Computational Paralinguistics Challenge. CCMT was declared the winning solution, obtaining an unweighted average recall of 65.41% and 85.87% for complaint and request detection, respectively. Moreover, we applied our framework on the Speech Commands v2 and HVB dialog data sets, surpassing previous studies reporting results on these benchmarks. Our code is freely available for download at: https://github.com/ristea/ccmt.

Virtual Assistant by using Large Language Model

Abstract— This paper presents the development of a virtual assistant named “Pragyan,” utilizing a Large Language Model (LLM) for natural language understanding and response generation. The virtual assistant is designed to interact with users through text and speech inputs, providing informative and conversational responses. Pragyan is built with a frontend user interface, a backend server, and a database for efficient communication and data management. The performance evaluation and result analysis demonstrate the effectiveness and user-friendliness of the virtual assistant Keywords: Virtual Assistant, Large Language Model, Natural Language Processing, Frontend, Backend, Database, Performance Evaluation, Result Analysis.

FastHPOCR: pragmatic, fast, and accurate concept recognition using the human phenotype ontology.

Human Phenotype Ontology (HPO)-based phenotype concept recognition (CR) underpins a faster and more effective mechanism to create patient phenotype profiles or to document novel phenotype-centred knowledge statements. While the increasing adoption of large language models (LLMs) for natural language understanding has led to several LLM-based solutions, we argue that their intrinsic resource-intensive nature is not suitable for realistic management of the phenotype CR lifecycle. Consequently, we propose to go back to the basics and adopt a dictionary-based approach that enables both an immediate refresh of the ontological concepts as well as efficient re-analysis of past data. We developed a dictionary-based approach using a pre-built large collection of clusters of morphologically equivalent tokens-to address lexical variability and a more effective CR step by reducing the entity boundary detection strictly to candidates consisting of tokens belonging to ontology concepts. Our method achieves state-of-the-art results (0.76 F1 on the GSC+ corpus) and a processing efficiency of 10 000 publication abstracts in 5 s. FastHPOCR is available as a Python package installable via pip. The source code is available at https://github.com/tudorgroza/fast_hpo_cr. A Java implementation of FastHPOCR will be made available as part of the Fenominal Java library available at https://github.com/monarch-initiative/fenominal. The up-to-date GCS-2024 corpus is available at https://github.com/tudorgroza/code-for-papers/tree/main/gsc-2024.

Artificial Intelligence in Banking Internal Demand Management Systems: The Example of Vakıf Participation Bank

The development of artificial intelligence and technology has accelerated the transformation of internal processes in the banking sector. In particular, Natural Language Processing (NLP) technology provides time and cost savings by automating processes such as data entry, querying, and reporting. While NLP-based systems increase customer satisfaction by understanding customer demands and providing appropriate responses quickly, they also increase operational efficiency. Classification algorithms, which are frequently used together with NLP technology, analyze text data and assign them to certain categories or classes, creating a powerful combination for the processing and analysis of text-based data. Vakıf Participation Demand Management System R&D Project has developed an NLP and classification model to be used in its internal processes. With the developed model, it was aimed to eliminate the problems encountered in workflow processes and increase efficiency by developing a language understanding model using the records of requests (demand management system) kept within Vakıf Participation and frequently used in operational processes. During this study, existing data containing in-house requests were subjected to pre-processing, and model training studies were carried out with these data. As a result of the developments, a model with 75% accuracy was developed and improvement efforts on the model continue. Thanks to the developed model, aims to shorten the response time for requests in the demand management system, reduce operational burdens, and increase internal customer satisfaction. It is planned to use the developed model in other banking internal processes as well.

Computational linguistics at the crossroads: A comprehensive review of NLP advancements

New NLP breakthroughs have put Computational Linguistics at a crossroads. NLP's past, present, and future are covered. This review explains computational linguistics' creation with a brief history of linguistics and computer science. Early solutions processed and understood natural language using rule-based systems using manually constructed linguistic rules. Over time, these tactics became increasingly problematic as language became more complex and obscure. Statistical approaches transformed operations. Neural network-based machine learning methods are leading the area because they can learn complicated patterns and representations from large text collections. A data-driven model revolution in natural language processing enhanced language modelling, machine translation, and sentiment analysis. Next, NLP improvements for several tasks and applications are evaluated. Language understanding models that capture semantic nuances and contextual relationships use deep learning frameworks. Word embeddings and transformer-based architectures like GPT and BERT perform well on benchmark datasets for text classification, question answering, and named item identification. The paper also shows how NLP interacts with computer vision, voice processing, and other domains to show the merits and cons of cross-disciplinary research. Multimodal techniques that combine text, graphics, and audio may increase natural language processing and interpretation. The review discusses NLP's effects on prejudice, justice, and privacy. Responsible development and implementation are needed when NLP technology becomes widespread due to algorithmic bias and data privacy concerns. NLP research directions and concerns are reviewed. Existing models may meet standards but fail in practice.

Cross-Model Comparative Loss for Enhancing Neuronal Utility in Language Understanding

Current natural language understanding (NLU) models have been continuously scaling up, both in terms of model size and input context, introducing more hidden and input neurons. While this generally improves performance on average, the extra neurons do not yield a consistent improvement for all instances. This is because some hidden neurons are redundant, and the noise mixed in input neurons tends to distract the model. Previous work mainly focuses on extrinsically reducing low-utility neurons by additional post- or pre-processing, such as network pruning and context selection, to avoid this problem. Beyond that, can we make the model reduce redundant parameters and suppress input noise by intrinsically enhancing the utility of each neuron? If a model can efficiently utilize neurons, no matter which neurons are ablated (disabled), the ablated submodel should perform no better than the original full model. Based on such a comparison principle between models, we propose a cross-model comparative loss for a broad range of tasks. Comparative loss is essentially a ranking loss on top of the task-specific losses of the full and ablated models, with the expectation that the task-specific loss of the full model is minimal. We demonstrate the universal effectiveness of comparative loss through extensive experiments on 14 datasets from three distinct NLU tasks based on five widely used pre-trained language models and find it particularly superior for models with few parameters or long input.

Structured Probabilistic Coding

This paper presents a new supervised representation learning framework, namely structured probabilistic coding (SPC), to learn compact and informative representations from input related to the target task. SPC is an encoder-only probabilistic coding technology with a structured regularization from the target space. It can enhance the generalization ability of pre-trained language models for better language understanding. Specifically, our probabilistic coding simultaneously performs information encoding and task prediction in one module to more fully utilize the effective information from input data. It uses variational inference in the output space to reduce randomness and uncertainty. Besides, to better control the learning process of probabilistic representations, a structured regularization is proposed to promote uniformity across classes in the latent space. With the regularization term, SPC can preserve the Gaussian structure of the latent code and achieve better coverage of the hidden space with class uniformly. Experimental results on 12 natural language understanding tasks demonstrate that our SPC effectively improves the performance of pre-trained language models for classification and regression. Extensive experiments show that SPC can enhance the generalization capability, robustness to label noise, and clustering quality of output representations.

Artificial Intelligence Methods in Natural Language Processing: A Comprehensive Review

The rapid evolution of Artificial Intelligence (AI) since its inception in the mid-20th century has significantly influenced the field of Natural Language Processing (NLP), transforming it from a rule-based system to a dynamic and adaptive model capable of understanding the complexities of human language. This paper aims to offer a comprehensive review of the various applications and methodologies of AI in NLP, serving as a detailed guide for future research and practical applications. In the early sections, the paper elucidates the indispensable role of AI in NLP, highlighting its transition from symbolic reasoning to a focus on machine learning and deep learning, and its extensive applications in sectors such as healthcare, transportation, and finance. It emphasizes the symbiotic relationship between AI and NLP, facilitated by platforms like AllenNLP, which aid in the development of advanced language understanding models. Further, the paper explores specific AI techniques employed in NLP, including machine learning, Naive Bayes, and Support Vector Machines, and identifies pressing challenges and avenues for future research. It delves into the applications of AI in NLP, showcasing its transformative potential in tasks such as machine translation, facilitated by deep learning methods, and the development of chatbots and virtual assistants that have revolutionized human-technology interaction. The paper also highlights other fields impacted by AI techniques, including text summarization, sentiment analysis, and named entity recognition, emphasizing the efficiency and accuracy brought about by the integration of AI in these areas. In conclusion, the paper summarizes the remarkable advancements and persistent challenges in NLP, such as language ambiguity and contextual understanding, and underscores the need for diverse and representative labeled data for training. Looking forward, it identifies promising research avenues including Explainable AI, Few-shot and Zero-shot Learning, and the integration of NLP with other data modalities, aiming for a holistic understanding of multimodal data. The paper calls for enhanced robustness and security in NLP systems, especially in sensitive applications like content moderation and fake news detection, to foster trust and reliability in AI technologies. It advocates for continual learning in NLP models to adapt over time without losing previously acquired knowledge, paving the way for a future where AI and NLP work synergistically to understand and generate human language more effectively and efficiently.

Neural machine translation of clinical text: an empirical investigation into multilingual pre-trained language models and transfer-learning.

Clinical text and documents contain very rich information and knowledge in healthcare, and their processing using state-of-the-art language technology becomes very important for building intelligent systems for supporting healthcare and social good. This processing includes creating language understanding models and translating resources into other natural languages to share domain-specific cross-lingual knowledge. In this work, we conduct investigations on clinical text machine translation by examining multilingual neural network models using deep learning such as Transformer based structures. Furthermore, to address the language resource imbalance issue, we also carry out experiments using a transfer learning methodology based on massive multilingual pre-trained language models (MMPLMs). The experimental results on three sub-tasks including (1) clinical case (CC), (2) clinical terminology (CT), and (3) ontological concept (OC) show that our models achieved top-level performances in the ClinSpEn-2022 shared task on English-Spanish clinical domain data. Furthermore, our expert-based human evaluations demonstrate that the small-sized pre-trained language model (PLM) outperformed the other two extra-large language models by a large margin in the clinical domain fine-tuning, which finding was never reported in the field. Finally, the transfer learning method works well in our experimental setting using the WMT21fb model to accommodate a new language space Spanish that was not seen at the pre-training stage within WMT21fb itself, which deserves more exploitation for clinical knowledge transformation, e.g. to investigate into more languages. These research findings can shed some light on domain-specific machine translation development, especially in clinical and healthcare fields. Further research projects can be carried out based on our work to improve healthcare text analytics and knowledge transformation. Our data is openly available for research purposes at: https://github.com/HECTA-UoM/ClinicalNMT.

NER Based Law Entity Privacy Protection

Within the field of legal AI, named entity recognition, also known as NER, is an essential step that must be completed before moving on to subsequent processing stages. In this paper, we present the creation of a dataset for the purpose of training natural language understanding models in the legal domain. The dataset is produced by locating and establishing a complete set of legal entities, which goes beyond traditionally employed entities such as person, organization, and location. These are examples of commonly used entities. Annotators are now provided with the means to effectively tag a wide variety of legal documents thanks to these additional entities. The authors tried out several different text annotation tools before settling on the one that proved to be the most effective for this study. The completed annotations are saved in the JavaScript Object Notation (JSON) format, which makes the data more readable and makes it easier to manipulate the data. The dataset that was produced as a result includes approximately thirty documents and five thousand sentences. Following that, these data are use in order to train a pre-trained SpaCy pipeline for accurate legal named entity prediction. There is a possibility that the accuracy of legal named entity recognition can be improved by performing additional fine-tuning on pre-trained models using legal texts.

Models and Approaches for Comprehension of Dysarthric Speech Using Natural Language Processing: Systematic Review.

Speech intelligibility and speech comprehension for dysarthric speech has attracted much attention recently. Dysarthria is characterized by irregularities in the speed, strength, pitch, breath control, range, steadiness, and accuracy of muscle movements required for articulatory aspects of speech production. This study examined the contributions made by other studies involved in dysarthric speech comprehension. We focused on the modes of meaning extraction used in generalizing speaker-listener underpinnings in light of semantic ontology extraction as a desired technique, applied method types, speech representations used, and databases sourced from. This study involved a systematic literature review using 7 electronic databases: Cochrane Database of Systematic Reviews, Web of Science Core Collection, Scopus, PubMed, ACM, IEEE Xplore, and Google Scholar. The main eligibility criterion was the extraction of meaning from dysarthric speech using natural language processing or understanding approaches to improve on dysarthric speech comprehension. In total, out of 834 search results, 30 studies that matched the eligibility requirements were acquired following screening by 2 independent reviewers, with a lack of consensus being resolved through joint discussion or consultation with a third party. In order to evaluate the studies' methodological quality, the risk of bias assessment was based on the Cochrane risk-of-bias tool version 2 (RoB2) with 23 of the studies (77%) registering low risk of bias and 7 studies (33%) raising some concern over the risk of bias. The overall quality assessment of the study was done using TRIPOD (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis). Following a review of 30 primary studies, this study revealed that the reviewed studies focused on natural language understanding or clinical approaches, with an increase in proposed solutions from 2020 onwards. Most studies relied on speaker-dependent speech features, while others used speech patterns, semantic knowledge, or hybrid approaches. The prevalent use of vector representation aligned with natural language understanding models, while Mel-frequency cepstral coefficient representation and no representation approaches were applied in neural networks. Hybrid representation studies aimed to reconstruct dysarthric speech or improve comprehension. Comprehensive databases, like TORGO and UA-Speech, were commonly used in combination with other curated databases, while primary data was preferred for specific or unique research objectives. We found significant gaps in dysarthric speech comprehension characterized by the lack of inclusion of important listener or speech-independent features in the speech representations, mode of extraction, and data sources used. Further research is therefore proposed regarding the formulation of models that accommodate listener and speech-independent features through semantic ontologies that will be useful in the inclusion of key features of listener and speech-independent features for meaning extraction of dysarthric speech.

USING LONG SHORT-TERM MEMORY NETWORKS FOR NATURAL LANGUAGE PROCESSING

The problem of emotion classification is a complex and non-trivial task of language interpretation due to the natural language structure and its dynamic nature. The significance of the study is in covering the important issue of automatic processing of client feedbacks, collecting opinions and trend-catching. In this work, a number of existing solutions for emotion classification problem were considered, having their shortcomings and advantages illustrated. The evaluation of performance of the considered models was conducted on emotion classification on four emotion classes, namely Happy, Sad, Angry and Others. The model for emotion classification in three-sentence conversations was proposed in this work. The model is based on smileys and word embeddings with domain specificity in state of art conversations on the Internet. The importance of taking into account the information extracted from smileys as an additional data source of emotional coloring is investigated. The model performance is evaluated and compared with language processing model BERT (Bidirectional Encoder Representations from Transformers). The proposed model achieved better performance at classifying emotions comparing to BERT (having F1 score as 78 versus 75). It should be noted, that further study should be performed to enhance the processing by the model of mixed reviews represented by emotion class Others. However, modern performance of models for language representation and understanding did not achieve the human performance. There is a variety of factors to consider when choosing the word embeddings and training methods to design the model architecture.

NL2LTL – a Python Package for Converting Natural Language (NL) Instructions to Linear Temporal Logic (LTL) Formulas

This is a demonstration of our newly released Python package NL2LTL which leverages the latest in natural language understanding (NLU) and large language models (LLMs) to translate natural language instructions to linear temporal logic (LTL) formulas. This allows direct translation to formal languages that a reasoning system can use, while at the same time, allowing the end-user to provide inputs in natural language without having to understand any details of an underlying formal language. The package comes with support for a set of default LTL patterns, corresponding to popular DECLARE templates, but is also fully extensible to new formulas and user inputs. The package is open-source and is free to use for the AI community under the MIT license. Open Source: https://github.com/IBM/nl2ltl. Video Link: https://bit.ly/3dHW5b1

IndicSUPERB: A Speech Processing Universal Performance Benchmark for Indian Languages

A cornerstone in AI research has been the creation and adoption of standardized training and test datasets to earmark the progress of state-of-the-art models. A particularly successful example is the GLUE dataset for training and evaluating Natural Language Understanding (NLU) models for English. The large body of research around self-supervised BERT-based language models revolved around performance improvements on NLU tasks in GLUE. To evaluate language models in other languages, several language-specific GLUE datasets were created. The area of speech language understanding (SLU) has followed a similar trajectory. The success of large self-supervised models such as wav2vec2 enable creation of speech models with relatively easy to access unlabelled data. These models can then be evaluated on SLU tasks, such as the SUPERB benchmark. In this work, we extend this to Indic languages by releasing the IndicSUPERB benchmark. Specifically, we make the following three contributions. (i) We collect Kathbath containing 1,684 hours of labelled speech data across 12 Indian languages from 1,218 contributors located in 203 districts in India. (ii) Using Kathbath, we create benchmarks across 6 speech tasks: Automatic Speech Recognition, Speaker Verification, Speaker Identification (mono/multi), Language Identification, Query By Example, and Keyword Spotting for 12 languages. (iii) On the released benchmarks, we train and evaluate different self-supervised models alongside the a commonly used baseline FBANK. We show that language-specific fine-tuned models are more accurate than baseline on most of the tasks, including a large gap of 76% for Language Identification task. However, for speaker identification, self-supervised models trained on large datasets demonstrate an advantage. We hope IndicSUPERB contributes to the progress of developing speech language understanding models for Indian languages.

Supervised Text Classification System Detects Fontan Patients in Electronic Records With Higher Accuracy Than ICD Codes.

Background The Fontan operation is associated with significant morbidity and premature mortality. Fontan cases cannot always be identified by International Classification of Diseases (ICD) codes, making it challenging to create large Fontan patient cohorts. We sought to develop natural language processing-based machine learning models to automatically detect Fontan cases from free texts in electronic health records, and compare their performances with ICD code-based classification. Methods and Results We included free-text notes of 10 935 manually validated patients, 778 (7.1%) Fontan and 10 157 (92.9%) non-Fontan, from 2 health care systems. Using 80% of the patient data, we trained and optimized multiple machine learning models, support vector machines and 2 versions of RoBERTa (a robustly optimized transformer-based model for language understanding), for automatically identifying Fontan cases based on notes. For RoBERTa, we implemented a novel sliding window strategy to overcome its length limit. We evaluated the machine learning models and ICD code-based classification on 20% of the held-out patient data using the F1 score metric. The ICD classification model, support vector machine, and RoBERTa achieved F1 scores of 0.81 (95% CI, 0.79-0.83), 0.95 (95% CI, 0.92-0.97), and 0.89 (95% CI, 0.88-0.85) for the positive (Fontan) class, respectively. Support vector machines obtained the best performance (P<0.05), and both natural language processing models outperformed ICD code-based classification (P<0.05). The sliding window strategy improved performance over the base model (P<0.05) but did not outperform support vector machines. ICD code-based classification produced more false positives. Conclusions Natural language processing models can automatically detect Fontan patients based on clinical notes with higher accuracy than ICD codes, and the former demonstrated the possibility of further improvement.

Symbols and grounding in large language models.

Large language models (LLMs) are one of the most impressive achievements of artificial intelligence in recent years. However, their relevance to the studyof language more broadly remains unclear. This article considers the potential of LLMs to serve as models of language understanding in humans. While debate on this question typically centres around models' performance on challenging language understanding tasks, this article argues that the answer depends on models' underlying competence, and thus that the focus of the debate should be on empirical work which seeks to characterize the representations and processing algorithms that underlie model behaviour. From this perspective, the article offers counterarguments to two commonly cited reasons why LLMs cannot serve as plausible models of language in humans: their lack of symbolic structure and their lack of grounding. For each, a case is made that recent empirical trends undermine the common assumptions about LLMs, and thus that it is premature to draw conclusions about LLMs' ability (or lack thereof) to offer insights on human language representation and understanding. This article is part of a discussion meeting issue 'Cognitive artificial intelligence'.

Catch Me If You Can: Deceiving Stance Detection and Geotagging Models to Protect Privacy of Individuals on Twitter

The recent advances in natural language processing have yielded many exciting developments in text analysis and language understanding models; however, these models can also be used to track people, bringing severe privacy concerns. In this work, we investigate what individuals can do to avoid being detected by those models while using social media platforms. We ground our investigation in two exposure-risky tasks, stance detection and geotagging. We explore a variety of simple techniques for modifying text, such as inserting typos in salient words, paraphrasing, and adding dummy social media posts. Our experiments show that the performance of BERT-based models fine-tuned for stance detection decreases significantly due to typos, but it is not affected by paraphrasing. Moreover, we find that typos have minimal impact on state-of-the-art geotagging models due to their increased reliance on social networks; however, we show that users can deceive those models by interacting with different users, reducing their performance by almost 50%.