Spoken document retrieval using topic models

Topic tracking language model for speech recognition

In this study, a real-time automatic speech recognition (ASR) system based on the Kaldi ASR toolkit, with low-latency and customizable models, without any internet connection, was developed. The proposed ASR system includes a voice activity detection (VAD) module and an audio transmitter as a front-end speech processing and a decoder for the received audio signals. The ASR system was evaluated in terms of ASR accuracy and speech processing speed. Consequently, the ASR system achieved high ASR accuracy on the CSJ (Corpus of Spontaneous Japanese) test set with super low-latency.

This paper presents two different approaches utilizing sta-tisticallanguagemodel(SLM)andsupportvectormachines(SVM) for sentence boundary detection of spontaneousJapanese. In the SLM-based approach, linguistic likeli-hoods and occurrence of pause are used to determine sen-tence boundaries. To suppress false alarms, heuristic pat-terns of end-of-sentence expressions are also incorporated.On the other hand, SVM is adopted to realize robust classi-fication against a wide variety of expressions and speechrecognition errors. Detection is performed by an SVM-based text chunker using lexical and pause information asfeatures. We evaluated these approaches on manual and au-tomatic transcription of spontaneous lectures and speeches,and achieved F-measures of 0.85 and 0.78, respectively. Index Terms : sentence boundary detection, spontaneousspeech, statistical language model, support vector ma-chines. 1. Introduction Recent advance of automatic speech recognition (ASR)technology, especially for spontaneous speech, enables var-ious applications such as spoken document archiving andretrieval, speech summarization and speech translation. Toorganizeaspokendocumentinastructuredformandtogiveuseful indices, transcriptions should be segmented into ap-propriate units like sentences. Moreover, these applicationsare usually built by combining an ASR system with naturallanguage processing (NLP) systems such as a parser and amachine translator, which often assume that input text is asentence. However,sentencesinspontaneousspeechareill-formed, and sentence boundaries are indistinct. Output textby ASR systems is just a sequence of words and has no ex-plicitsentenceboundaries, sothefurtherstepofsegmentingthe ASR output is required for these applications.Automatic boundary detection of spoken sentences hasbeen explored mainly on broadcast news (BN)tasks[1,2,3]and conversational telephone speech (CTS) tasks[3, 4, 5] inEnglish. As features for detection, pause, prosodic and lin-guistic information is often used. Most popular approach isa combination of prosodic and linguistic information[2, 3],which realizes high performance on BN and CTS tasks.Prosody-based approaches[1, 5] have also been investi-gated. Meanwhile, linguistic information is not used by it-self, since most of these works were performed on Englishdata, where cue words or expressions of sentence bound-aries are not easily defined.In Japanese, cue expressions are typically observed atthe end of sentences and expected to be useful for bound-ary detection. However, variety of such expressions is solarge in spoken Japanese, that it is hard to collect sufficientamountofdatafortrainingstatisticalmodelssuchasamax-imum entropy (ME) model. Moreover, many of cue expres-sions consist of particles, which are apparently difficult tobedetectedinASR.Thus,robustnessforASRerrorsshouldalso be investigated.In this paper, we address two approaches of sen-tence boundary detection for spontaneous Japanese. Asframeworks of detection, we adopt and compare statisti-cal language model (SLM) and support vector machines(SVM). The proposed approaches are evaluated with reallectures and speeches included in the Corpus of Sponta-neous Japanese (CSJ).

The accurate estimation of respiratory rate (RR) is crucial for assessing the respiratory system’s health in humans, particularly during auscultation processes. Despite the numerous automated RR estimation approaches proposed in the literature, challenges persist in accurately estimating RR in noisy environments, typical of real-life situations. This becomes especially critical when periodic noise patterns interfere with the target signal. In this study, we present a parallel driver designed to address the challenges of RR estimation in real-world environments, combining multi-core architectures with parallel and high-performance techniques. The proposed system employs a nonnegative matrix factorization (NMF) approach to mitigate the impact of noise interference in the input signal. This NMF approach is guided by pre-trained bases of respiratory sounds and incorporates an orthogonal constraint to enhance accuracy. The proposed solution is tailored for real-time processing on low-power hardware. Experimental results across various scenarios demonstrate promising outcomes in terms of accuracy and computational efficiency.

Topic modeling extracts useful potential topics that reflect market information from massive financial news and is widely used in data mining and economic research. Traditional topic modeling approaches such as Latent Dirichlet Allocation (LDA) and Non-Negative Matrix Factorization (NMF) lack semantic information, and short texts have feature sparse problems. We develop a topic clustering model based on BERT-LDA joint embedding that takes both contextual semantics and thematic narrative into account. We cluster document embeddings with the HDBSCAN algorithm and utilize a class-based TF-IDF (c-TF-IDF) method to create topic representations. Empirical results show that the BERT-LDA model is competitive compared with traditional and single topic models. It generates coherent topic words that are dissimilar to each other.

This paper proposes to use Web information for confidence measure and to extract keywords for speech recognition results. Spoken document processing has been attracting attention particularly for information retrieval and video (audiovisual) content systems. For example, measuring a confidence score which indicates how likely a document or a segmented document includes recognition errors has been studied. It is well known keyword extraction from recognition results is also an important issue. For these purposes, in this paper, pointwise mutual information (PMI) between two words is employed. PMI has been used to calculate a confidence measure of speech recognition, as a coherence measure by co-occurrence of words. We propose to further improve the method by using a Web query expansion technique with term triplets which consist of nouns in the same document. We also apply PMI to keyword estimation by summing a co-occurrence score (sumPMI) between a targeting keyword candidate and each term. The proposed methods were tested with 10 lectures in Corpus of Spontaneous Japanese (CSJ) and 2 simulated movie dialogues. In the experiments it is shown that the estimated confidence score has high relationship with recognition accuracy, indicating the effectiveness of our method. And sumPMI scores for keywords have higher values in the subjective tests.

In an effort to develop topic modeling methods that can be quickly applied to large data sets, we revisit the problem of maximum-likelihood estimation in topic models. It is known, at least informally, that maximum-likelihood estimation in topic models is closely related to non-negative matrix factorization (NMF). Yet, to our knowledge, this relationship has not been exploited previously to fit topic models. We show that recent advances in NMF optimization methods can be leveraged to fit topic models very efficiently, often resulting in much better fits and in less time than existing algorithms for topic models. We also formally make the connection between the NMF optimization problem and maximum-likelihood estimation for the topic model, and using this result we show that the expectation maximization (EM) algorithm for the topic model is essentially the same as the classic multiplicative updates for NMF. Our methods are implemented in the R package "fastTopics".

Text embedding plays a crucial role in natural language processing (NLP). Among various approaches, nonnegative matrix factorization (NMF) is an effective method for this purpose. However, the standard NMF approach, fundamentally based on the bag-of-words model, fails to utilize the contextual information of documents and may result in a significant loss of semantics. To address this issue, we propose a new NMF scheme incorporating a regularization term based on the Wasserstein metric, which quantifies an approximation of a word-context matrix to leverage semantic information. Since the word-context matrix is typically symmetric and positive definite (SPD), the Wasserstein metric used in the regularization is related to a natural SPD manifold structure. We then build upon this manifold structure to integrate an explicit expression of the Wasserstein metric into the NMF framework. Additionally, we introduce modifications to the gradient computation to adapt three fundamental classes of numerical algorithms from ordinary NMF to tackle the Wasserstein-regularized NMF (NMF-WR) problem. To demonstrate the effectiveness of the NMF-WR model and the proposed algorithms, we conduct experiments on popular datasets, focusing on topic modeling and document clustering. The results indicate that the NMF-WR model exhibits superior performance compared to other conventional NMF models in processing text embeddings. These findings suggest that this novel NMF-WR framework not only enhances semantic representation but also underscores a commitment to the model’s reliability and interpretability.

BackgroundTopic models are a class of unsupervised machine learning models, which facilitate summarization, browsing and retrieval from large unstructured document collections. This study reviews several methods for assessing the quality of unsupervised topic models estimated using non-negative matrix factorization. Techniques for topic model validation have been developed across disparate fields. We synthesize this literature, discuss the advantages and disadvantages of different techniques for topic model validation, and illustrate their usefulness for guiding model selection on a large clinical text corpus.Design, setting and dataUsing a retrospective cohort design, we curated a text corpus containing 382,666 clinical notes collected between 01/01/2017 through 12/31/2020 from primary care electronic medical records in Toronto Canada.MethodsSeveral topic model quality metrics have been proposed to assess different aspects of model fit. We explored the following metrics: reconstruction error, topic coherence, rank biased overlap, Kendall’s weighted tau, partition coefficient, partition entropy and the Xie-Beni statistic. Depending on context, cross-validation and/or bootstrap stability analysis were used to estimate these metrics on our corpus.ResultsCross-validated reconstruction error favored large topic models (K ≥ 100 topics) on our corpus. Stability analysis using topic coherence and the Xie-Beni statistic also favored large models (K = 100 topics). Rank biased overlap and Kendall’s weighted tau favored small models (K = 5 topics). Few model evaluation metrics suggested mid-sized topic models (25 ≤ K ≤ 75) as being optimal. However, human judgement suggested that mid-sized topic models produced expressive low-dimensional summarizations of the corpus.ConclusionsTopic model quality indices are transparent quantitative tools for guiding model selection and evaluation. Our empirical illustration demonstrated that different topic model quality indices favor models of different complexity; and may not select models aligning with human judgment. This suggests that different metrics capture different aspects of model goodness of fit. A combination of topic model quality indices, coupled with human validation, may be useful in appraising unsupervised topic models.

Background/Objectives: Unsupervised topic models are often used to facilitate improved understanding of large unstructured clinical text datasets. In this study we investigated how ICD-9 diagnostic codes, collected alongside clinical text data, could be used to establish concurrent-, convergent- and discriminant-validity of learned topic models. Design/Setting: Retrospective open cohort design. Data were collected from primary care clinics located in Toronto, Canada between 01/01/2017 through 12/31/2020. Methods: We fit a non-negative matrix factorization topic model, with K = 50 latent topics/themes, to our input document term matrix (DTM). We estimated the magnitude of association between each Boolean-valued ICD-9 diagnostic code and each continuous latent topical vector. We identified ICD-9 diagnostic codes most strongly associated with each latent topical vector; and qualitatively interpreted how these codes could be used for external validation of the learned topic model. Results: The DTM consisted of 382,666 documents and 2210 words/tokens. We correlated concurrently assigned ICD-9 diagnostic codes with learned topical vectors, and observed semantic agreement for a subset of latent constructs (e.g. conditions of the breast, disorders of the female genital tract, respiratory disease, viral infection, eye/ear/nose/throat conditions, conditions of the urinary system, and dermatological conditions, etc.). Conclusions: When fitting topic models to clinical text corpora, researchers can leverage contemporaneously collected electronic medical record data to investigate the external validity of fitted latent variable models.

Japanese dependency structure is usually represented by relationships between phrasal units called bunsetsus. One of the biggest problems with dependency structure analysis in spontaneous speech is that clause boundaries are ambiguous. This paper describes a method for detecting the boundaries of quotations and inserted clauses and that for improving the dependency accuracy by applying the detected boundaries to dependency structure analysis. The quotations and inserted clauses are determined by using an SVM-based text chunking method that considers information on morphemes, pauses, fillers, etc. The information on automatically analyzed dependency structure is also used to detect the beginning of the clauses. Our evaluation experiment using Corpus of Spontaneous Japanese (CSJ) showed that the automatically estimated boundaries of quotations and inserted clauses helped to improve the accuracy of dependency structure analysis.

Objective: We analyzed nanoparticle regulation research to examine the evolution of regulatory frameworks, identify major thematic structures, and evaluate current challenges in the governance of rapidly advancing nanotechnologies. By drawing parallels with the historical development of radiation regulation, the study aimed to contextualize emerging regulatory strategies and derive lessons for future governance. Methods: A total of 9095 PubMed-indexed articles published between January 2015 and October 2025 were analyzed using text mining, keyword frequency analysis, and topic modeling. Preprocessed titles and abstracts were transformed into a TF-IDF (Term Frequency-Inverse Document Frequency) document-term matrix, and NMF (Non-negative Matrix Factorization) was applied to extract semantically coherent topics. Candidate topic numbers (K = 1-12) were evaluated using UMass coherence scores and qualitative interpretability criteria to determine the optimal topic structure. Results: Six major research topics were identified, spanning energy and sensor applications, metal oxide toxicity, antibacterial silver nanoparticles, cancer nano-therapy, and nanoparticle-enabled drug and mRNA delivery. Publication output increased markedly after 2019 with interdisciplinary journals driving much of the growth. Regulatory considerations were increasingly embedded within experimental and biomedical research, particularly in safety assessment and environmental impact analyses. Conclusions: Nanoparticle regulation matured into a dynamic multidisciplinary field. Regulatory efforts should prioritize adaptive, data-informed, and internationally harmonized frameworks that support innovation while ensuring human and environmental safety. These findings provide a data-driven overview of how regulatory thinking was evolved alongside scientific development and highlight areas where future governance efforts were most urgently needed.

In this paper, we address the problem of spectral unmixing in urban hyperspectral images using a Maximum A Posteriori (MAP)-based Non-negative Matrix Factorization (NMF) approach. Considering a Linear-Quadratic (LQ) mixing model, we seek to decompose the spectrum observed in each pixel of the image into a set of pure material spectra, as well as their abundance fractions and the mixing coefficients associated with products of these pure material spectra. The main idea of the proposed method is to take into account the available prior information about the unknown parameters for a better estimation of them. To this end, we first derive a MAP-based cost function, then minimize it using a projected gradient algorithm by modifying a recently proposed NMF method adapted to LQ mixtures. Simulation results confirm the relevance of our approach.

Since spontaneous utterances include many variations, speakerand task-independent general models do not work well. This paper proposes combining cluster-based language and acoustic models based on the framework of Massively Parallel Decoder (MPD). The MPD is a parallel decoder that has a large number of decoding units, in which each unit is assigned to each combination of element models. It runs efficiently on a parallel computer, and thus the turnaround time is comparable to conventional decoders using a single model and a processor. In the experiments conducted using lecture speeches from the Corpus of Spontaneous Japanese, two types of cluster models have been investigated: lecture-based cluster models and utterancebased cluster models. It has been confirmed that utterancebased cluster models give significantly lower recognition error rate than lecture-based cluster models in both language and acoustic modeling. It has also been shown that roughly 100 decoding units are enough in terms of recognition rate, and in the best setting, 12% reduction in word error rate was obtained in comparison with the conventional decoder.

Matrix factorization plays an important role in scientific computation. The widely used one is singular value decomposition (SVD) which approximates the original data matrix with three lower rank matrices with orthogonality constraints. Recently nonnegative matrix factorization (NMF) considering the nonnegativity of data makes the results more interpretable than those of SVD. However NMF finds only two factor matrices and there is no significant index as singular values of SVD which can be used for sorting learned basis vectors. In this paper we take into account the nonnegativity for SVD and propose nonnegative SVD (NNSVD). The preliminary results on the microarray data of spermatogenesis show that NNSVD has advantages of both SVD and NMF.