Semantic Context Research Articles

The modulating effect of lexical predictability on perceptual learning of degraded speech

Predictive coding is considered to be an important mechanism for perceptual learning. Posterior prediction-error minimization can lead to higher rates of lasting changes in the representational hierarchy, and hence is likely to enhance the process of learning. In the field of speech processing, although considerable studies have demonstrated that a highly predictive sentence context can facilitate the perception of forthcoming word, it remains to be examined that how this type of predictability affects the perceptual learning of speech (especially degraded speech). The present study, therefore, aimed to examine whether and how the lexical predictability of spoken sentences modulates perceptual learning of speech embedded in noise, by using spoken sentences as training stimuli and strictly controlling the semantic-context constraint of these training sentences. The current study adopted a “pretest-training-posttest” procedure. Two groups of subjects participated in this perceptual learning study, with cognitive and language abilities matched across these two groups. For one group, the spoken sentences used for training all have a highly predictive semantic context; for another group, the training sentences all have a low predictive context. The results showed that both the reaction time and accuracy of the speech-in-noise intelligibility test were significantly improved in the post-training phase compared to the pre-training phase; moreover, the learning-related improvement was significantly enhanced in participants with weak-constraint sentences as training stimuli (compared to those with strong-constraint sentences as training stimuli). This enhancement effect of low lexical predictability on learning-related improvement supports a prediction-error based account of perceptual learning.

Long and short-range relevance context network for semantic segmentation

The semantic information can ensure better pixel classification, and the spatial information of the low-level feature map can ensure the detailed location of the pixels. However, this part of spatial information is often ignored in capturing semantic information, it is a huge loss for the spatial location of the image semantic category itself. To better alleviate this problem, we propose a Long and Short-Range Relevance Context Network. Specifically, we first construct a Long-Range Relevance Context Module to capture the global semantic context of the high-level feature and the ignored local spatial context information. At the same time, we build a Short-Range Relevance Context Module to capture the piecewise spatial context information in each stage of the low-level features in the form of jump connections. The whole network adopts a coding and decoding structure to better improve the segmentation results. Finally, we conduct a large number of experiments on three semantic segmentation datasets (PASCAL VOC2012, Cityscapes and ADE20K datasets) to verify the effectiveness of the network.

ESDSCNet: an enhanced shallow feature difference and semantic context network for remote sensing change detection: with building change detection as a case

ABSTRACT Change detection (CD) is a challenging task considering land surface cover changes. Recently, deep learning has been introduced into CD. However, these methods suffer from shortcomings such as insufficient utilization of shallow features and inadequate aggregation of semantic context, which result in pseudo changes and poor integrity of change objects. To address this, an enhanced shallow feature difference and semantic context network (ESDSCNet) is proposed in this study for remote sensing CD. It uses HRNet to extract multi-scale features from bi-temporal images to obtain shallow and deep features. To fully exploit the shallow features, they are input into the difference statistical texture learning (DSTL) module to extract more discriminative features. Subsequently, the features enhanced by DSTL are fed into the change object contextual representation (COCR) module along with the deep difference features extracted by HRNet to characterize the contextual information of the change object. To verify the performance of ESDSCNet in different scenes, this paper takes building change detection as a case and conducts experiments based on three datasets. The experimental results reveal that ESDSCNet outperforms six other advanced methods, regardless of the intersection over union or the harmonic mean of precision rate and recall rate, which further confirms the effectiveness of the proposed network. In addition, another speciality of this paper is that ESDSCNet can identify not only ‘Where has changed’ but also ‘What has changed what’, so it has potential application in semantic change detection besides building change detection.

Small-Scale Ship Detection for SAR Remote Sensing Images Based on Coordinate-Aware Mixed Attention and Spatial Semantic Joint Context

With the rapid development of deep learning technology in recent years, convolutional neural networks have gained remarkable progress in SAR ship detection tasks. However, noise interference of the background and inadequate appearance features of small-scale objects still pose challenges. To tackle these issues, we propose a small ship detection algorithm for SAR images by means of a coordinate-aware mixed attention mechanism and spatial semantic joint context method. First, the coordinate-aware mixed attention mechanism innovatively combines coordinate-aware channel attention and spatial attention to achieve coordinate alignment of mixed attention features. In this way, attention with finer spatial granularity is conducive to strengthening the focusing ability on small-scale objects, thereby suppressing the background clutters accurately. In addition, the spatial semantic joint context method exploits the local and global environmental information jointly. The detailed spatial cues contained in the multi-scale local context and the generalized semantic information encoded in the global context are used to enhance the feature expression and distinctiveness of small-scale ship objects. Extensive experiments are conducted on the LS-SSDD-v1.0 and the HRSID dataset. The results with an average precision of 77.23% and 90.85% on the two datasets show the effectiveness of the proposed methods.

Biomedical knowledge graph learning for drug repurposing by extending guilt-by-association to multiple layers

Computational drug repurposing aims to identify new indications for existing drugs by utilizing high-throughput data, often in the form of biomedical knowledge graphs. However, learning on biomedical knowledge graphs can be challenging due to the dominance of genes and a small number of drug and disease entities, resulting in less effective representations. To overcome this challenge, we propose a “semantic multi-layer guilt-by-association" approach that leverages the principle of guilt-by-association - “similar genes share similar functions", at the drug-gene-disease level. Using this approach, our model DREAMwalk: Drug Repurposing through Exploring Associations using Multi-layer random walk uses our semantic information-guided random walk to generate drug and disease-populated node sequences, allowing for effective mapping of both drugs and diseases in a unified embedding space. Compared to state-of-the-art link prediction models, our approach improves drug-disease association prediction accuracy by up to 16.8%. Moreover, exploration of the embedding space reveals a well-aligned harmony between biological and semantic contexts. We demonstrate the effectiveness of our approach through repurposing case studies for breast carcinoma and Alzheimer’s disease, highlighting the potential of multi-layer guilt-by-association perspective for drug repurposing on biomedical knowledge graphs.

FTransCNN: Fusing Transformer and a CNN based on fuzzy logic for uncertain medical image segmentation

The accurate segmentation of medical images plays a crucial role in diagnosing and treating diseases. Although many methods now use multimodal joint segmentation, the joint use of segmentation features extracted by multiple models can lead to heterogeneity and uncertainty. Unreasonable fusion methods cannot exploit the advantages of multiple models and still lack good performance in segmentation. Therefore, this study proposes the FTransCNN model, which is composed of a convolutional neural network (CNN) and Transformer and is based on a fuzzy fusion strategy that jointly utilizes the features extracted by a CNN and Transformer through a new fuzzy fusion module. First, the CNN and Transformer act as the backbone network for parallel feature extraction. Second, channel attention is used to promote the global key information of Transformer to improve the feature representation ability, and spatial attention is used to enhance the local details of CNN features and suppress irrelevant regions. Third, the proposed model applies the Hadamard product to model fine-grained interactions between the two branches and uses the Choquet fuzzy integral to suppress heterogeneity and uncertainty in fused features. Fourth, FTransCNN employs fuzzy attention fusion module (FAFM) hierarchical upsampling to effectively capture both low-level spatial features and high-level semantic context. Finally, the new model obtains the final segmentation result by using the deconvolution and results in an improvement in segmentation. The experimental results on Chest X-ray and Kvasir-SEG dataset show that FTransCNN has better performance on segmentation tasks than the-state-of-the-art deep segmentation models.

Conversation Modeling to Predict Derailment

Conversations among online users sometimes derail, i.e., break down into personal attacks. Derailment interferes with the healthy growth of communities in cyberspace. The ability to predict whether an ongoing conversation will derail could provide valuable advance, even real-time, insight to both interlocutors and moderators. Prior approaches predict conversation derailment retrospectively without the ability to forestall the derailment proactively. Some existing works attempt to make dynamic predictions as the conversation develops, but fail to incorporate multisource information, such as conversational structure and distance to derailment. We propose a hierarchical transformer-based framework that combines utterance-level and conversation-level information to capture fine-grained contextual semantics. We propose a domain-adaptive pretraining objective to unite conversational structure information and a multitask learning scheme to leverage the distance from each utterance to derailment. An evaluation of our framework on two conversation derailment datasets shows an improvement in F1 score for the prediction of derailment. These results demonstrate the effectiveness of incorporating multisource information for predicting the derailment of a conversation.

CoWNet: A correlation weighted network for geological hazard detection

The geological hazard of landslide or debris flow is one of the most widespread causes of mortality every year. Segmentation for them has been an effective tool to reduce the number of casualties since its critical contribution to rescue planning. However, existing segmentation methodologies are mainly designed for natural imagery in which most objects have similar shapes and sizes. In contrast to them, landslides and debris flow usually take irregular contour and arbitrary scope, causing most existing methodologies to suffer from low performance. Therefore, it is imperative to design a novel segmentation method that is specific to geological hazard auto-detection. To address these issues, we propose a correlation weighted network (CoWNet) to precisely segment the hazard by concentrating on the negative impact of irregular contour and arbitrary scope. The CoWNet gains profits from two elaborately designed modules, named correlation weighted module (CoW) and feature fusion module (FFM), respectively. Firstly, in contrast to the common preference to blindly expand the receptive field of existing methods, the CoW module chooses to adjust the receptive field based on disaster scope adaptively. The CoW module enables the receptive field of the CoWNet to encompass any scope of hazards. Meanwhile, another module, FFM, is used to relieve the adverse effect brought by the irregular contour problem. The core idea of FFM is to utilize sophisticated semantic context to preliminarily locate the overall disaster and detailed local cues to refine disaster edges. Through the above strategies, the CoWNet could precisely segment any shape and scope of landslides or debris flow. To verify the superiority of the CoWNet, we compare our network with several state-of-art segmentation networks in a high-resolution geological hazard dataset, including landslide and debris flow disasters. The results demonstrate our model outperforms other state-of-the-art models.

Cultural unit blue in the Old Testament

Abstract The paper describes and analyzes the full presence of blue in the Old Testament – in Hebrew and translations. The interdisciplinary approach includes the treatment of color as a cultural unit according to of Eco’s idea, lexical and contextual semantics, distinguishing visual and verbal color languages. The interface between verbal and visual color language is the prototype. Prototypes are universal natural visual objects – sky, sea, fire, blood, the sun at noon, all plants, light, milk, snow, darkness, and coal and have evolved into cultural units (Eco 1996 [1985]) in all cultures and languages. Basic Color Terms (BCT – blue), Prototype Terms (PT – sky and sea), Rivals Terms of prototypes (RT – sapphire, blue skins), Terms for Basic Features of the Prototypes (TBFP – breadth, infinite, boundless) are examined. Translation is a criterion for semiotic value, cultural and linguistic context. Norm of Test of Free Word-Associations is a source of non-color (secondary) meanings of verbal colors. Test results allow us to flesh out the hidden links ‘prototype color─most typical feature of the prototype─secondary meanings of color’ in text and visual culture. The secondary cultural meanings of BCTs and RTs are specified in color compounds. The color blue tehèlet is used in most cases in synergy (not in opposition) with the other three colors in a tetrad: purple argamàn─scarlet tolàat shanì─linen shesh. Analysis on the synergic relations is performed. The same applies to duads blue-linen, blue-purple, blue skins-scarlet, blue skins-red skins. Color compounds are elements of the Priestly Code, therefore a hypothesis on compounds at semiotic axes is developed. The PT sky is involved in religious heritage. Themes as Hebrew substitute shamàim for the Tetragrammaton, the Hierarchy of heavens complete the areas of cultural unit blue in the Old Testament.

A Lightweight Dual-Branch Swin Transformer for Remote Sensing Scene Classification

The main challenge of scene classification is to understand the semantic context information of high-resolution remote sensing images. Although vision transformer (ViT)-based methods have been explored to boost the long-range dependencies of high-resolution remote sensing images, the connectivity between neighboring windows is still limited. Meanwhile, ViT-based methods commonly contain a large number of parameters, resulting in a huge computational consumption. In this paper, a novel lightweight dual-branch swin transformer (LDBST) method for remote sensing scene classification is proposed, and the discriminative ability of scene features is increased through combining a ViT branch and convolutional neural network (CNN) branch. First, based on the hierarchical swin transformer model, LDBST divides the input features of each stage into two parts, which are then separately fed into the two branches. For the ViT branch, a dual multilayer perceptron structure with a depthwise convolutional layer, termed Conv-MLP, is integrated into the branch to boost the connections with neighboring windows. Then, a simple-structured CNN branch with maximum pooling preserves the strong features of the scene feature map. Specifically, the CNN branch lightens the LDBST, by avoiding complex multi-head attention and multilayer perceptron computations. To obtain better feature representation, LDBST was pretrained on the large-scale remote scene classification images of the MLRSN and RSD46-WHU datasets. These two pretrained weights were fine-tuned on target scene classification datasets. The experimental results showed that the proposed LDBST method was more effective than some other advanced remote sensing scene classification methods.

What foreign language learners make of grammatical descriptions depends on description type, proficiency, and context

Abstract Most usage-based research emphasizes the importance of implicit, input-driven learning in naturalistic environments, but recent studies have adopted usage-based grammatical descriptions for instructed learning in classrooms. These descriptions are intended to draw learners’ deliberate attention to relevant usage patterns in the input and thereby support intake. Most of these studies compare usage-based descriptions to other types of descriptions for their efficiency, while little attention has been paid to the ways in which learners understand and apply such descriptions. This study examines what foreign language learners understand of usage-based grammatical descriptions of target structures. In an experimental forced choice task, Chinese learners of German received usage-based descriptions of case structures and then classified target instances in variable contexts. A multivariate regression analysis indicated that choices were influenced by interactions of the type of description with participants’ target-language proficiency and the semantic and lexical target contexts. This is discussed in terms of noticing and category formation. This study argues that learners are able to use grammatical descriptions as some kind of auxiliary model for recognizing and categorizing target patterns. The descriptions thus make learners aware of the mechanisms underlying implicit learning and help them exploit these mechanisms for explicit learning.

Semantic Annotation Based Effective and Quality Oriented Web Service Discovery

The main objective of this work is to design an effective web service discovery engine that retrieves the best matching results based on the measure of syntactic cum semantic similarity between the user query and the web service to be fetched. The idea is to draw bridges across the broad spectrum of lexical entities based on their relativeness. The essence of this work could be extended to support a wide range of applications from ‘being inculcated in search engines to fetch user-relevant data’ to ‘being used for training robots and AI based devices to respond/adhere appropriately to the different phrases of human commands’.With the accelerated revolution of internet, enterprises and organizations highly rely on Service oriented computing. Web services support inter-operation of distributed applications. Humongous amount of web services present on the internet the user is searching for. The paper focuses ondesigninga practical means of fast and relevant retrieval of web services.The phrase used for searching any given web service differs from one person to another. This work deals with the syntactic cum semantic mechanism proposed for retrieving web services based on the measure of similarity between the name of the web service and the search query entered by the user to discover that service. A list of the Web Service Description Language (WSDL) files is taken as the dataset. Protégé is the tool used for semantic annotation of WSDL files for converting them to Semantically Annotated Web Service Description Language (SAWSDL) files. ‘Wordnet’ is used as the lexical dictionary and Java language has been used to build the rest of the package for implementing the search discovery mechanism.Netbeans is used as IDE.Wamp server with PHPMyAdmin was used for managing the database of SAWSDL files. The degree of similarity is measured by evaluating performance of the engine using metrics such as ‘Precision’, ‘Recall’, ‘Accuracy’ and ‘F-measure’.While the syntactic approach is easier to implement, it suffers from keyword polysemy issues. The proposed search discovery mechanism is based on semantically annotating WSDL files and retrieving the files based on a novel syntactic cum semantic discovery algorithm which usesLeocockChordorow function for computing the similarity. The effectiveness of the proposed algorithm is tested experimentally by building a desktop application using Java.The WSDL files from various domains were semantically annotated by tagging related concepts using ontology. Themachine learning algorithm that best classifies the web services on the basis of their performance metrics is identified. The related services are retrieved by the application of the proposed LCH based Syntactic cum Semantic discovery algorithm (LCH based SSDA).A ranking system is proposed to rank the results by evaluating various QoS attributes. The results of the experiment showed that the proposed system yielded high precision and recall value.The solution has been found to be effective in minimizing the execution time and in improving the degree of relevancy. With an impending need for constructing a semantic context based secure structure,the proposed solution will- help in refining the search results and in minimizing users’ cognitive load during search formulation and execution.

DMCT-Net: dual modules convolution transformer network for head and neck tumor segmentation in PET/CT

Objective. Accurate segmentation of head and neck (H&N) tumors is critical in radiotherapy. However, the existing methods lack effective strategies to integrate local and global information, strong semantic information and context information, and spatial and channel features, which are effective clues to improve the accuracy of tumor segmentation. In this paper, we propose a novel method called dual modules convolution transformer network (DMCT-Net) for H&N tumor segmentation in the fluorodeoxyglucose positron emission tomography/computed tomography (FDG-PET/CT) images. Approach. The DMCT-Net consists of the convolution transformer block (CTB), the squeeze and excitation (SE) pool module, and the multi-attention fusion (MAF) module. First, the CTB is designed to capture the remote dependency and local multi-scale receptive field information by using the standard convolution, the dilated convolution, and the transformer operation. Second, to extract feature information from different angles, we construct the SE pool module, which not only extracts strong semantic features and context features simultaneously but also uses the SE normalization to adaptively fuse features and adjust feature distribution. Third, the MAF module is proposed to combine the global context information, channel information, and voxel-wise local spatial information. Besides, we adopt the up-sampling auxiliary paths to supplement the multi-scale information. Main results. The experimental results show that the method has better or more competitive segmentation performance than several advanced methods on three datasets. The best segmentation metric scores are as follows: DSC of 0.781, HD95 of 3.044, precision of 0.798, and sensitivity of 0.857. Comparative experiments based on bimodal and single modal indicate that bimodal input provides more sufficient and effective information for improving tumor segmentation performance. Ablation experiments verify the effectiveness and significance of each module. Significance. We propose a new network for 3D H&N tumor segmentation in FDG-PET/CT images, which achieves high accuracy.

Safety helmet detection method based on semantic guidance and feature selection fusion

Safety helmet detection is a hot topic of research in the field of industrial safety for object detection technology. Existing object detection methods still face great challenges for the detection of small-scale safety helmet object. In this paper, we propose a safety helmet detection method based on the fusion of semantic guidance and feature selection. The method is able to consider the balance between detection performance and efficiency. First, a multi-scale non-local module is proposed to establish internal correlations between different scales of deep image features as well as to aggregate semantic context information to guide the information recovery of decoder network features. Then the feature selection fusion structure is proposed to adaptively select deep features and underlying key features for fusion to make up for the missing semantic and spatial detail information of the decoding network and improve the spatial location expression capability of the decoding network. Experimental analysis shows that the method in this paper has good detection performance on the expanded safety helmet wearing dataset with 5.12% improvement in mAP compared to the baseline method CenterNet, and 6.11% improvement in AP for the safety helmet object.

Unveiling the marginal role of energy storage solutions in Italy: Insights from semantic network analysis of online news

Adopting energy storage systems is crucial in the transition to sustainable energy sources. However, significant obstacles hinder their widespread use, often attributed to socio-cultural factors such as low user awareness. In this study, we examine the societal awareness of energy storage in Italy by analyzing online news, through innovative methods and tools of text mining and network analysis. Our analysis of over 143,000 energy-related articles published on major Italian news outlets identifies prominent discourse topics, specifically focusing on energy storage, and assesses their semantic context. Our results demonstrate that media coverage can play a vital role in increasing awareness and societal acceptance of energy storage solutions, thus facilitating their potential market uptake. Leveraging the visibility of these themes and connecting the concept of storage with specific arguments in the energy discourse could help engage a broader audience in implementing this technology. Our approach provides valuable insights into media narratives and can inform future communications strategies. We present metrics that are useful as marketing tools and can help policymakers and organizations assess their communication strategies.

Semantic Context Can Mask Intelligibility Declines at Above-Conversational Speech Levels in Normal-Hearing Listeners.

While speech audibility generally improves with increasing level, declines in intelligibility are inconsistently observed at above-conversational levels, even in listeners with normal audiograms (NAs). The inconsistent findings could be due to different types of speech materials, ranging from monosyllabic words to everyday sentences, used across studies. Here, we hypothesized that semantic context can "mask" intelligibility declines at high levels by constraining plausible response options. Intelligibility was assessed in speech-shaped noise with monosyllabic words, sentences without semantic context, and sentences with semantic context. Two presentation levels were used: 80 and 95 dB SPL broadband. Bandpass filtering was applied to minimize upward spread of masking. Twenty-two young adults with NAs were tested. Poorer performance at the higher level was found with the monosyllabic words and context-free sentences but not with the context-rich sentences. The scores obtained at the higher level with the two context-free materials were strongly correlated. The correlation was independent of the lower-level scores, suggesting that the high-level performance declines reflect "normal" auditory functioning. Young adults with NAs show intelligibility decreases at above-conversational levels when tested with speech materials without semantic context. Top-down processing as facilitated by context information can mask such declines.

BERT-PG: a two-branch associative feature gated filtering network for aspect sentiment classification

Aspect sentiment classification is an important branch of sentiment classification that has gained increasing attention recently. Existing aspect sentiment classification methods typically use different network branches to encode context and aspect words separately, and then use an attention mechanism to capture their associations. This attention-based approach cannot completely ignore the contexts unrelated to the current aspect words, which brings noise interference. In this paper, a gated filtering network based on BERT is suggested as a solution to this issue. We employ BERT to encode the text semantics of contexts and sentence pairs consisting of context and aspect words respectively, and to extract lexical features as well as associative features of context and aspect words. Based on this, we designed a gating module that, unlike the attention mechanism, uses association features to precisely filter irrelevant contexts. Additionally, because the BERT network parameters are so big, there is a tendency to over-fitting during training. To effectively combat this problem, we developed a loss function with a threshold. We carried out extensive experiments using three benchmark datasets to verify the performance of our proposed model. The experimental results show that the method improves the accuracy by 0.5%, 1.39% and 2.57% on the Laptop, Restaurant and Twitter datasets respectively, and 1.564%, 2.36% and 4.144% on Macro-F1 respectively, compared to the recent RA-CNN (BERT), proving that our method is effective in improving the presentation of aspect sentiment classification in comparison to other cutting-edge sentiment classification methods.

The effects of speech masking on neural tracking of acoustic and semantic features of natural speech

Listening environments contain background sounds that mask speech and lead to communication challenges. Sensitivity to slow acoustic fluctuations in speech can help segregate speech from background noise. Semantic context can also facilitate speech perception in noise, for example, by enabling prediction of upcoming words. However, not much is known about how different degrees of background masking affect the neural processing of acoustic and semantic features during naturalistic speech listening. In the current electroencephalography (EEG) study, participants listened to engaging, spoken stories masked at different levels of multi-talker babble to investigate how neural activity in response to acoustic and semantic features changes with acoustic challenges, and how such effects relate to speech intelligibility. The pattern of neural response amplitudes associated with both acoustic and semantic speech features across masking levels was U-shaped, such that amplitudes were largest for moderate masking levels. This U-shape may be due to increased attentional focus when speech comprehension is challenging, but manageable. The latency of the neural responses increased linearly with increasing background masking, and neural latency change associated with acoustic processing most closely mirrored the changes in speech intelligibility. Finally, tracking responses related to semantic dissimilarity remained robust until severe speech masking (−3 dB SNR). The current study reveals that neural responses to acoustic features are highly sensitive to background masking and decreasing speech intelligibility, whereas neural responses to semantic features are relatively robust, suggesting that individuals track the meaning of the story well even in moderate background sound.

Multi-Hop Question Generation with Knowledge Graph-Enhanced Language Model

The task of multi-hop question generation (QG) seeks to generate questions that require a complex reasoning process that spans multiple sentences and answers. Beyond the conventional challenges of what to ask and how to ask, multi-hop QG necessitates sophisticated reasoning from dispersed evidence across multiple sentences. To address these challenges, a knowledge graph-enhanced language model (KGEL) has been developed to imitate human reasoning for multi-hop questions.The initial step in KGEL involves encoding the input sentence with a pre-trained GPT-2 language model to obtain a comprehensive semantic context representation. Next, a knowledge graph is constructed using the entities identified within the context. The critical information in the graph that is related to the answer is then utilized to update the context representations through an answer-aware graph attention network (GAT). Finally, the multi-head attention generation module (MHAG) is performed over the updated latent representations of the context to generate coherent questions. Human evaluations demonstrate that KGEL generates more logical and fluent multi-hop questions compared to GPT-2. Furthermore, KGEL outperforms five prominent baselines in automatic evaluations, with a BLEU-4 score that is 27% higher than that of GPT-2.

Acoustic and Semantic Processing of Auditory Scenes in Children with Autism Spectrum Disorders

Purpose: Processing real-world sounds requires acoustic and higher-order semantic information. We tested the theory that individuals with autism spectrum disorder (ASD) show enhanced processing of acoustic features and impaired processing of semantic information. Methods: We used a change deafness task that required detection of speech and non-speech auditory objects being replaced and a speech-in-noise task using spoken sentences that must be comprehended in the presence of background speech to examine the extent to which 7–15 year old children with ASD (n = 27) rely on acoustic and semantic information, compared to age-matched (n = 27) and IQ-matched (n = 27) groups of typically developing (TD) children. Within a larger group of 7–15 year old TD children (n = 105) we correlated IQ, ASD symptoms, and the use of acoustic and semantic information. Results: Children with ASD performed worse overall at the change deafness task relative to the age-matched TD controls, but they did not differ from IQ-matched controls. All groups utilized acoustic and semantic information similarly and displayed an attentional bias towards changes that involved the human voice. Similarly, for the speech-in-noise task, age-matched–but not IQ-matched–TD controls performed better overall than the ASD group. However, all groups used semantic context to a similar degree. Among TD children, neither IQ nor the presence of ASD symptoms predict the use of acoustic or semantic information. Conclusion: Children with and without ASD used acoustic and semantic information similarly during auditory change deafness and speech-in-noise tasks.