(2)
To extract data, the full texts of the 35 selected studies were read. A data extraction form was created to collect all the information needed to address the research questions and the study quality criteria. This form can be found in Appendix B . The data extraction was done with the help of a Microsoft Excel spreadsheet which is freely available online: https://bit.ly/3Fv7qDp (accessed on 1 November 2022). Data from the included studies were extracted by two authors and verified by the third author.
When performing an SLR, the extracted data should be synthesized in a manner suitable for answering the questions that an SLR seeks to answer [ 19 ]. After the quality assessment, 35 of the 39 articles made it to the final literature selection; these articles are shown in Table 4 below.
Final articles selected.
Nr. | Reference | Title |
---|---|---|
A | [ ] | Tingog: Reading and Speech Application for Children with Repaired Cleft Palate |
B | [ ] | A serious game for speech disorder children therapy |
C | [ ] | The BioVisualSpeech Corpus of Words with Sibilants for Speech Therapy Games Development |
D | [ ] | Advanced Information Technology—Support of Improved Personalized Therapy of Speech Disorders |
E | [ ] | TERAPERS -Intelligent Solution for Personalized Therapy of Speech Disorders |
F | [ ] | An automated speech-language therapy tool with interactive virtual agent and peer-to-peer feedback |
G | [ ] | Android based Receptive Language Tracking Tool for Toddlers. |
H | [ ] | Towards a speech therapy support system based on phonological processes early detection. |
I | [ ] | Assessing comprehension of spoken language in nonspeaking children with cerebral palsy: Application of a newly developed computer-based instrument |
J | [ ] | AppVox: An Application to Assist People with Speech Impairments in Their Speech Therapy Sessions |
K | [ ] | Apraxia world: A Speech Therapy Game for Children with Speech Sound Disorders |
L | [ ] | Speak App: A Development of Mobile Application Guide for Filipino People with Motor Speech Disorder |
M | [ ] | Speech technologies in a computer-aided speech therapy system |
N | [ ] | ChilDiBu—A Mobile Application for Bulgarian Children with Special Educational Needs |
O | [ ] | Audiovisual Tools for Phonetic and Articulatory Visualization in Computer-Aided Pronunciation Training |
P | [ ] | Building on Mobile towards Better Stuttering Awareness to Improve Speech Therapy |
Q | [ ] | Pictogram Tablet: A Speech Generating Device Focused on Language Learning |
R | [ ] | Measuring performance of children with speech and language disorders using a serious game |
S | [ ] | A robotic assistant to support the development of communication skills of children with disabilities |
T | [ ] | Evaluating a multi-avatar game for speech therapy applications |
U | [ ] | Secure telemonitoring system for delivering telerehabilitation therapy to enhance children’s communication function to home |
V | [ ] | Architecture of an automated therapy tool for childhood apraxia of speech |
W | [ ] | Translation of the Speech Therapy Programs in the Logomon Assisted Therapy System. |
X | [ ] | An educational platform based on expert systems, speech recognition, and ludic activities to support the lexical and semantic development in children from 2 to 3 years. |
Y | [ ] | SPELTA: An expert system to generate therapy plans for speech and language disorders. |
Z | [ ] | SPELTA-Miner: An expert system based on data mining and multilabel classification to design therapy plans for communication disorders. |
AA | [ ] | The AppVox mobile application, a tool for speech and language training sessions |
BB | [ ] | A prelingual tool for the education of altered voices |
CC | [ ] | A Game Application to assist Speech Language Pathologists in the Assessment of Children with Speech Disorders |
DD | [ ] | End-User Recommendations on LOGOMON - a Computer Based Speech Therapy System for Romanian Language |
EE | [ ] | Multimodal Speech Capture System for Speech Rehabilitation and Learning |
FF | [ ] | Tabby Talks: An automated tool for the assessment of childhood apraxia of speech |
GG | [ ] | AACVOX: mobile application for augmentative alternative communication to help people with speech disorder and motor impairment |
HH | [ ] | An Online Expert System for Diagnostic Assessment Procedures on Young Children’s Oral Speech and Language |
II | [ ] | E-inclusion technologies for the speech handicapped |
In this section, we synthesize the SLR results and provide the answers to the identified research questions. In total, the selected papers ( Table 4 ) discussed 35 systems. Study AA was a further improvement of the system set up in study J. Study DD was a further improvement of the system set up in study W. Study AA and study DD provided further experimental validation of their previous studies. Given that both studies AA and DD also introduced new elements, we considered them in the primary studies.
Figure 2 shows the distribution of the intervention goals of each study. The categorization of intervention goal for each study can be tricky, as some of the indicated intervention goals can overlap. For example, a study that uses automatic classification of speech errors can also be developed to support the SLP. The following classification is based on what each paper mentions as main intervention goal for the development of the system. The majority of the OST systems’ main goal was to support the SLP in his/her current activities (66%). The majority of the reviewed studies discussed developing a system intending to assist the SLP with their work and at the same time provide clients with the possibility to continue their practice without the constant need for an SLP present. Studies G, I, L, N, GG and II (17%) focused on building an application suitable for a specific target group, such as children with cerebral paresis. Studies C, EE and FF (8%) developed a system to automatically classify speech errors in clients’ spoken text. One could argue that this falls under the category of "Support the SLP", but in reality, diagnosing a client often happens in the first session and support is mainly needed during the follow-up sessions, which is why these three papers got their categorization "automatic classification of speech errors". Studies Q and T (6%) tried to set up a system specifically focused on maintaining the child’s attention during the therapy session. Finally, study B (3%) aimed to increase accessibility by setting up a game environment. Increasing accessibility was also mentioned in other studies, but rather as a consequence rather than as a study’s primary goal.
Intervention goal OST systems.
Four of the thirty-five studies described “people” as their target group. However, these studies (i.e., F, J, P and II) included both adults and children. Thus, these studies were also considered and not removed from the list, as the systems discussed might also be beneficial for children. The distribution of the target language of the reviewed systems is portrayed in Figure 3 . It is evident from the figure that most systems were designed with a focus on the English language. Spanish, Portuguese and Romanian were the next most common languages apart from English.
Overview of target languages of OST Systems.
To fully understand the functionality of OST systems, it is essential to know which features they have. Features can be defined as user-visible characteristics. In total, 14 common features among the 35 studies were identified. The explanations of the features are described below based on the explanations given in the primary studies.
Representation of visual feedback.
The features and their corresponding studies are presented in Table 5 . By examining the features in the table, we can observe that reporting is the most dominant feature. This was not unexpected, since two-third of the systems were made to support the SLP. A reporting feature provides the SLP a quick overview of the child’s progress and the mistakes made, saving much time for one-on-one interaction. The most repeatedly used report types are statistical data of properly practiced words and statistical data of overall activities during the therapy session. Most of the papers (25) do not discuss any reporting feature for their systems.
OST Features.
Feature | Study | Total Number |
---|---|---|
Audio feedback | C, E, J, W, AA, BB | 6 |
Emotion Screening | P | 1 |
Error Detection | V, AA, CC, EE, FF, HH, II | 7 |
Peer-to-peer feedback | F, K | 2 |
Recommendation strategy | H, S, W, Y, Z | 5 |
Reporting | D, S, V, W, X, Y, Z, AA, BB, CC, DD, FF, GG, HH | 14 |
Speech Recognition | A, H, M, O, S, V, X, BB, CC, EE, II | 11 |
Text-to-speech | A, S, GG, II | 4 |
Textual feedback | F, J, II, CC, FF | 5 |
User Data Management | S, X, Y, Z, DD, II | 6 |
User Record voice | E, Q, U, V, W, CC, EE, FF, GG, II | 10 |
Virtual 3D model | E, O, W, DD, EE | 5 |
Visual feedback | C, EE, II | 3 |
Voice commands | R, S | 2 |
An overview of the target disorders can be found in Table 6 .
Classification of communication disorders.
Classification | Study |
---|---|
Communication disorder | S, X, Z, |
Speech disorder | A, C, D, E, H, K, L, N, P, Q, V, W, BB, CC, DD, EE, FF, GG, II |
Language disorder | B, F, I, J, R, T, U, Y, AA, HH |
Hearing disorder | G, M, O |
Table 6 presents an overview of the target communication disorders of the papers. During the literature analysis, we came across many different terms. For example, study H mentioned “speech disorder” as its target disorder, study J mentioned “speech impairments” and study P mentioned “speech impediments”. As shown in Table 6 , 19 studies mentioned “speech disorder” as their target disorder, which is the majority. However, this term is rather broad. In study B, they mention that the 3D game environment was developed to meet the specific requirements of “language disordered” children, but did not clarify the type of disorder. They tested the system on a child who had a language disorder because of hearing impairment, but they do not mention if the final system was developed for this focus group. The system of study F provides tasks for both aphasia and speech disorders; however, aphasia is a non-speech disorder. Studies in the communication disorder groups did not further define the participants’ disorders (or the target disorder) regarding speech, language and hearing. This may have been due in part to the developed systems being generic SLP-assistance OSTs designed to simply support the SLP in client management.
In this analysis, we aimed to examine the studies that presented and discussed the adopted architectural patterns used in developing the OST systems. Fifteen of the thirty-five studies did not mention or discuss any architecture approach at all. The rest of the studies did, but the extent of details discussed differed a lot per study. Table 7 provides an overview of the adopted architectural patterns of those 20 studies that did discuss them. In the subsections below, we discuss each architecture found in the studies.
Overview of architectures of OST systems.
Adopted Architecture Approach | Study |
---|---|
client–server system | D, F, H, L, P, U, V, DD, HH, II |
Repository pattern | T, CC |
Layered approach | S, X, Y, Z |
Standalone system | A |
Pipe-and-Filter Architecture | E, W, FF |
From Table 7 , the most commonly adopted architecture approach was a client–server approach. Study D used a two-tier client–server architecture in which data mining was used to derive knowledge from the data. Study H described a software architecture containing a capture module and a service module. The first collects the client’s audio data and then revises the data, which the SLP eventually double-checks after it is checked by the service module. In study F, they created a virtual therapist in the application that provides the sound/word task with audiovisual cues and articulation. The client pronounces the sound/word, and the virtual therapist then sends the client’s audio response to the system server. The system server analyses the speech and detects mispronunciation. The detected errors are sent back to the application, which sends feedback to the client. The system in study L also has one module for the therapist, namely, the assessment module. In the assessment module, the SLP can view the clients and their scores, and the practice module is used by the client to do exercises. The system described in study P has a mobile application used by the client and a server that manages all the applications’ requests. All the data are stored in a central database. An Internet connection is needed from time to time to synchronize the client’s progress in the application with the central database. The server module of study U was threefold, as it comprises a management application program, a database server and a data transfer application. When the child is playing the game, its interactions are captured by the desktop client software and stored on a local open-source database. The therapist can review these interactions and send the data to the database server with the data transfer application’s help. Through an Internet connection, the therapist then has remote access to the data of each child. The application of study V has a multi-tier client–server architecture and provides remote administration of speech therapy. Through the user interface, the SLP can remotely manage therapy for his or her clients, create exercises and organize speech recordings, as this application is also where the speech analysis can be done. The client has access to the therapy session through a mobile application. Study DD showed a detailed architecture of the speech therapy system, but no additional explanation was provided in the other studies that presented a picture of the RA. The provided pictures showed the interaction between the SLP’s computer and the client’s mobile device. The expert system on the computer of the therapist selects exercises, and if the SLP agrees to them, they are sent to the child. The mobile device collects the child’s vocal production, and the results are sent to the monitor program installed on the SLP’s computer. The design of the online expert system in study HH is based on a system that is built from smaller subsystems, including the conditions for articulation disorders, phonological disorders, fluency disorders and language disorders. Study II discussed their architecture in the style of a block diagram. The application works with only audio input and lets the user know whether the word/sentence was pronounced correctly or not.
The system’s interface in study T tests a client’s performance with the help of a database with correctly pronounced sounds. The game developed in study CC stores the speech input in an SQL database; the study mainly discusses the pre-processing, feature extraction and automatic speech recognition, but does not go into deep detail regarding the RA.
The system proposed in study S relies on a user interface, an expert system and a domain knowledge layer. Each layer interacts with the other and provides different services. The UI and services layer provide several functionalities for helping SLPs. The expert-system layer relies on two modules for performing the processes to interact with users and support the decision making of SLPs. The domain knowledge is managed in the last layer through ontologies, databases for monitoring and activities, standardized vocabularies and a clinical data repository. The interaction layer of the system in study X has two applications, one that the therapist can use to manage the exercises of the client and one to perform administrative tasks. The information captured in the interaction layer is sent to the expert system in the service layer, which contains the following modules: a web services module, a user management module, a report generation module and a module for speech recognition. The SPELTA system from study Y consists of different systems that all work with a knowledge base. The SPELTA-miner system in study Z is an expert system responsible for conducting machine learning, analyzing and generating therapy plans and educational content.
The architecture design in study A was only textually described, and no figure/image was provided. The client practices pronunciation with the help of the training module.
In the systems from studies E and W, the modules are mainly designed for assisting the SLP. The two main components of the system are an intelligent system that is installed on the computer of the SLP and a mobile system used by the client. Between the SLP and the client itself, there is a personal relationship, and they can see each other through the home monitor program. The intelligent system contains a 3D model that analyses the client’s words, and these can be reviewed by the SLP. The data are transferred to the expert system, and it can offer suggestions regarding which exercises are most suitable for the child. It is a fuzzy expert system that is rule-based, which makes it easier for the SLP to understand. The proposed system in study FF consists of a tablet-based mobile application that records the child’s speech when he is talking during the exercises. The spoken words are assessed by the voice activity detector (VAD) in the speech recognition engine, which provides the assessment results to the SLP at the interface. At this interface, the SLP can also create and assign new exercises to clients and obtain an overview of each child’s progress.
The majority of the papers did not discuss the usage of machine learning (ML). Only 13 of the 35 primary studies discussed which ML approach they used to develop the system. In all cases, the ML types could be broken down into either supervised or unsupervised. Semi-supervised and reinforcement learning were not mentioned in the papers. Figure 5 shows the distribution of the ML types.
The proportion of ML types in the literature.
Supervised learning is more commonly used than unsupervised learning because more clustering tasks are applied in OST Systems. The proportions of ML types are shown in Figure 5 .
Figure 6 shows the number of times each algorithm appeared in the final literature. The algorithms that each paper used are presented in Table 8 . Study A used automatic speech recognition (ASR) as a tool to transfer the speech signal into a string of words. The words spoken by the client in the microphone are processed by a computer program that first extracts the spoken text features. These extracted features are then compared with the trained patterns. Natural language processing was used for the text-to-speech system. The text that is given as input comes out as synthetic voice output.
The number of times each algorithm appeared in the final literature.
Overview of adopted ML approaches.
Nr. | ML Types | ML Tasks | Algorithms | Application | Adopted Dataset |
---|---|---|---|---|---|
A | Unsupervised | Clustering | Not mentioned | Speech Recognition | Not mentioned |
C | Supervised | Classification | Convolutional Neural Networks (CNN) Hidden-Markov Model | Speech Recognition | The database contains reading aloud recordings of 284 children. The corpus contains reading aloud recordings from 510 children. |
E | Unsupervised | Clustering | Not mentioned | Generate a therapy plan | Not mentioned |
F | Unsupervised | Clustering | Hidden Markov Model | Time prediction | Not mentioned |
H | Supervised | Classification | Decision Tree Neural Network Support Vector Machine k-Nearest Neighbor Random Forest | Speech classification | A Phonological Knowledge Base containing speech samples collected from 1114 evaluations performed with 84 Portuguese words. |
M | Supervised | Classification | Artificial Neural Networks (ANN) | Speech recognition | The authors refer to a large speech database, but no further details are given. |
W | Unsupervised | Clustering | Not mentioned | Generate a therapy plan | Not mentioned |
Y | Supervised | Classification | Decision Tree Artificial Neural networks | Generate a therapy plan | Not mentioned |
Z | Supervised | Classification | Artificial Neural Networks | Generate a therapy plan | Database of thousands of therapy strategies. |
CC | Supervised | Classification | Convolutional Neural Networks (CNN) | Speech to Text | TORGO Dataset that contains audio data of people with dysarthria and people without dysarthria. |
DD | Unsupervised | Clustering | Not Mentioned | Emotion recognition | Not applicable |
FF | Supervised | Classification | Artificial Neural Network (ANN) Logistic regression Support Vector Machine | Speech recognition | A dataset with correctly-pronounced utterances from 670 speakers. |
HH | Supervised | Classification | Neural Networks | Detect disorder | Not mentioned |
In study C, a convolutional neural network (CNN) was used to recognize the child’s words during the gameplay. The authors also tried other classification models for speech recognition, such as support vector machines and artificial neural networks. However, using the CNN model gave the lowest number of false negatives compared to the other models. In comparison with the other studies, the explanation was very extensive. The study also showed a representation of the 1D CNN architecture. To evaluate the children’s pronunciations, study E discussed the use of a fuzzy logic algorithm that assigns weight to the level of the speech disorder. The tool discussed in study F detects errors in a voice with the help of a hidden Markov model (HMM). The spoken phonemes by the client are compared with the target phoneme voice. The model can detect insertion, deletion and substitution. They also described a face tracker used to analyze the client’s nonverbal behavior by coordinating their eyes. Study H used a decision tree (DT) classifier for evaluating the correctness of speech samples. With this method, they managed to reach correct classification of the pronunciation of almost 93%. The authors then used this information to extend their phonological database. Other classifiers, such as KNN, random forest classifier and Adam’s neural network reached lower accuracy than 93%. Study M applied a three-layer ANN for speech recognition. Study W created a system that can suggest a helpful therapeutic plan for each client with fuzzy logic. The fuzzy expert system can recommend the needed follow-up actions, such as some exercises needed for a client based on several parameters. Study Y and Z used an ANN to generate a therapy plan, for which they used a multilayer perceptron. Study CC used CNN, which breaks the words into pieces and then analyses them. Study DD used a fuzzy logic algorithm to answer how frequent the therapy sessions should be, how long they should take and what types of exercises should be included. Study FF used three types of classifiers to enable speech recognition, namely, a multi-layer perceptron, an SVM and a logistic regression model with the help of MATLAB Toolboxes. Study HH used neural networks to detect the severity of the disorder.
Figure 7 shows the relationships between ML approaches and the OST goals. The figure shows that ML was only used for two of the five intervention goals, namely, “support the SLP” and “automatic classification of speech errors”. Not surprisingly, the two studies C and FF that had automatic classification as a goal only had classification as a ML task. For the intervention goal of supporting the SLP, both classification and clustering were used.
Bubblechart of ML approaches and OST goals.
We describe the delivery type and indicate whether it is designed for desktops, mobile phones, browsers or multiple formats. The supported platform for each piece of software can affect the number of users and its accessibility. If the software supports different platforms, the range of users might be many and more exhaustive. Our review shows that most of the systems are desktop-based, followed by mobile-based. Figure 8 shows the distribution of delivery types in numbers. Desktop only is the most popular one, with 34%, followed by mobile-only, with 29%.
OST delivery types of primary studies.
Table 9 shows the evaluation approach that each of the studies used to evaluate its OST system. Overall, five main evaluation approaches were observed across all studies; the case study approach was more prevalent than the others. None of the studies adopted two or more approaches. Each evaluation approach is discussed below.
Approaches to evaluation of the literature.
Evaluation Approach | Study |
---|---|
Case Study | C, G, K, L, M, O, P, R, S, U, V, X, Y, Z, AA, GG |
Experimental | E, I, Q, T, BB, DD |
Not evaluated | F, N, W, HH |
Observational | B, J |
Simulation-based | A, D, H, CC, EE, FF, II |
Sixteen studies tested the discussed OST system with a case study. All sixteen stated the sample size (see Appendix C ), apart from study U. Study U mentioned that pilot tests with children with disabilities and typically developing children were done, but not the sample size. Study C first did a screening activity for six months with 356 5- to 9-year-old children. During this screening period, SLPs were asked to fill in individual reports for every child about the screening results. Children were assessed individually in a quiet room in their school setting by an SLP or an SLP graduate student. Each child had two different screening moments. Study G’s system evaluation was done throughout therapy sessions in seven weeks, with five 2- to 6-year-old children. Throughout the sessions, the SLP made notes about the improved receptive vocabulary of the children. The authors of study K tested their system with a within-subject study where children played two versions of the game. During the game, the children were asked to fill in surveys. The authors analyzed meta-data to identify differences in versions in the amount of speech practice completed. Study L consulted a technical evaluation by asking 30 IT professionals to fill in an evaluation form and rate the developed mobile application. In study M, the software was used for five months in the therapy at a school for the hearing impaired to improve the number of correctly pronounced vowels before and after five months of therapy. The tool from study O was evaluated in the therapy of speech disorders by performing a sensitivity test. The SLP noted that when the target sound was reached or not during the therapy, and there was a statistically significant improvement after therapy. A preliminary user study was done in study P to evaluate whether people would understand the concept of the OST application. As can be seen in Appendix C , only one of the five participants could be considered as a minor. However, the authors indicated that their OST system can also be used by young people; thus, we did not exclude it from our analysis. Study R tested their game with a 4-year-old and a 6-year-old and checked their test results. The drawback of this study was that apart from age, no additional information about the two children was given. After using the application, the participants were asked to fill in a questionnaire anonymously. The authors of study S conducted a pilot experiment consisting of two stages: a first one consisting of laboratory tests to determine the robot’s performance and a second stage to analyze the client’s response to the robot’s appearance. Study U evaluated the system in four families with one or more children with disabilities who were currently receiving speech therapy. The system in study V was validated through a pilot study with children diagnosed with apraxia of speech, together with their parents and SLPs. After a session of ten minutes, the children were asked several questions, and the SLP and parents were asked to fill in questionnaires. A similar procedure was conducted for the system in study X, where a pilot experiment was conducted. However, in this study, only the SLPs were asked to fill in the survey. Similarly, for study Y and study Z, a pilot experiment was conducted, during which SLPs were asked to execute evaluations with the help of an online tool. Study AA used three different methods of assessment. First, they asked SLPs to provide feedback on the application. Secondly, they asked three usability experts, such as a professor in computer engineering, to perform a heuristic evaluation. Finally, they performed user tests with a group of children with speech disorders and a group without. In study GG, twenty participants used the developed application and filled in a questionnaire to rate the usability afterward.
Five studies validated the discussed OST system with an experiment. The participants’ characteristics can be found in Appendix C . Study E evaluated the performances of the system with the help of forty 5- to 6-year-old children. Twenty children attended classical therapy sessions, and the other 20 used the developed system. Statistical tests were done to compare the difference between groups after 24 meetings. For investigating the performance of the system in study I, participants were asked to carry out tasks under the supervision of an SLP in a quiet room. During the experiment, responses were recorded by observation and video. For quantitative analysis, the total number of correct responses was calculated with the help of a language assessment tool. Additionally, the researchers asked the speech therapist whether they agreed with the performance score that was assigned by each tool to each participant. In study Q, one-to-one, 45 min sessions were performed with the participants and the SLP. Additionally, semi-structured interviews were conducted with both therapist and parents. Each session was examined by an observer who noted the interaction between the child and the SLP with conversational analysis. Study T administered the gaze targets of an experimental group and the control group. Based on the time interval for which the child was looking at the robot, the level of engagement was estimated. To test the effectiveness of the tool of study BB, an experiment was conducted in two schools for special education, considering objective measurements from the statistical analysis of the results stored by the tool and the subjective measurements from a therapy evaluation form for each user proposed by the therapist.
Study B tested the system with two 5-year-old children. The researchers observed both sessions of the children with the SLP and gave a description of what they saw in the paper. The participants of study J were observed while executing a task with the systems. Usability tests were executed to evaluate the interaction of the participants while working directly with the system. Both studies mentioned the characteristics of their participants ( Appendix C ).
The usability of the proposed application of study A was validated with a software test. Study D tested its modules on target datasets. Study H adopted and implemented the most extensive evaluation amongst all selected studies. The researchers evaluated the proposed system with a database that consisted of speech samples collected from 1114 evaluations with 1077 children, resulting in a database containing 93,576 audio samples. The game application of study CC was tested by using a Torgo Dataset that contained audio data of people with and without dysarthria. Study FF evaluated the performances of their trained algorithms using three experiments. For study EE, a proof-of-concept prototype was set up, which seemed to work fine but needs further clinical evaluation.
Study F mentioned in the discussion that they are planning in the future to evaluate the proposed system. Likewise, study N explained at the end that their proposed application would be tested later. In study W, they mentioned that their system has been validated experimentally, but the authors did not provide further explanations on the experiment. Noteworthy for study HH is that the authors mentioned the procedures that are part of a testing stage but did not explain any further.
The evaluation metrics that were encountered in the literature are displayed in Figure 9 . Five articles did not evaluate the performances of their models, whereas some articles used multiple metrics. Table 10 provides an overview of the metrics found in the articles. The metrics that were used the most were accuracy and efficiency, which are explained in the next paragraph. The metrics used by the studies differed, as studies tested different aspects of their systems. Two main categories evaluation metrics were used, namely, machine learning-based (ML) and general evaluation metrics. Below, we discuss the metrics under the two main categories.
The evaluation metrics adopted by the selected articles.
Evaluation metrics used in the selected literature.
Metrics | Study |
---|---|
Accuracy | H, M, Z, CC, FF |
Recall | H, FF |
F1-Score/ F1-Measure | H, FF |
Precision | FF |
Pearson’s r | I |
RMSE | H, EE |
Kappa | I |
Error | H, FF, II |
Usability | A, L, GG |
Satisfaction | AA, GG |
Efficiency | L, AA, DD, GG |
Effectiveness | J, AA |
Reliability | L, T |
Sensitivity | O |
Coherence | X |
Completeness | X |
Relevance | X |
Ease of learning memorization | GG |
ML Evaluation Metrics:
General Evaluation Metrics:
The review included studies that looked at various types of disorders, outcome measures and levels of evidence. There seems to be a need for authors that are both experts in the communication disorders field and in the software engineering field. However, no publication has been found that addresses both of these concerns clearly. Even though papers and definitions were checked by a certified SLP, the different categorization systems and broad variety of disorders make it difficult to put all the systems in a clear group that everybody would agree with. Thus, one of the challenging aspects of this review was the high heterogeneity among the discussed papers, as in the study from Chen et al. [ 11 ], making it challenging to draw general conclusions. The studies showed broad variations in study designs and methodological quality. Some experimental studies had small sample sizes—for example, only five participants—making it risky to draw any overall conclusions on the systems, as the OST Systems were proposed for various communication disorders. The classification of communication disorders is somewhat complex and versatile, and although we provided a brief summary in the background ( Section 2 ), it is essential to realize that each communication disorder has its own therapy approach. Some of the developed systems were targeted at SSDs, which is a rather broad category. We also found that many papers lacked sufficient details about the architecture used, which can be useful for researchers as inspiration for building their own systems. The findings from this study could have suffered from some limitations, which should be acknowledged when interpreting the results. A common threat to the validity of an SLR can be the so-called publication bias. We tried to minimize this limitation by developing a research protocol and writing down our methodology on a detailed level. Even though the search query used to perform the SLR was rather extensive, there is always a risk that some relevant literature is overlooked. In an ideal case, all the 4481 articles found within the query would be scanned more thoroughly, and not just their titles and abstracts. However, this would have been not feasible regarding the given time for this research. The data extraction was conducted as objectively as possible, yet there is a chance that some details were overlooked. To validate the extraction process, the second and third author did random cross-checks. Future research could look at how many of the researched OST systems are being used. Some of the studies reported that the discussed systems were also introduced to the market. However, it would be interesting to see if some of the systems are still being used. We also acknowledge that other evaluation metrics which are useful were not reported in our study. Metrics that could discern if the child’s speech has improved are not used by researchers who developed tools and these kind of metrics were missing from our study. It would be interesting to examine or propose more practical metrics that assesses the performances of children who use these developed tools.
Speech therapy is a very essential procedure for children with communication disorders. However, not all children with communication disorders have access to the limited number of speech–language pathologists (SLP). Fortunately, several online speech therapy (OST) systems have been designed and proposed. Previous systematic reviews on OST systems for children with speech sound disorders (SSDs) are limited and discuss a wide variety of features. Through a systematic literature review, this paper examined currently existing automated speech therapy programs that have been discussed in prior literature. Eight research questions were set up to obtain further information on the existing OST systems and to obtain a deeper understanding of the current challenges of the OST systems. Out of the 4481 papers found by our search strategy, 35 of the papers primarily focused on OST systems for children with speech disorders. Our analysis shows that there is a wide variety of systems that have already been developed. The main goal of most designed OST systems was to support the SLP in their tasks. Systems are available in different languages and for different target disorders. It is challenging to understand how some of these OST systems are set up, as most studies did not describe a reference architecture (RA). The studies that mainly did used a client–server approach, which provides the clients with speech therapy services with the help of a database. Additionally, the number of studies that adopted and used machine learning techniques was lower than the number that did not. This finding explains why there are so many OST systems designs, yet only a few are eventually developed and implemented for practical use.
Quality assessment score.
Reference | Q1. Are the Aims of the Study Clearly Stated? | Q2. Are the Scope and Context of the Study Clearly Defined? | Q3. Is the Proposed Solution Clearly Explained and Validated by an Empirical Study? | Q4. Are the Variables Used in the Study Likely to Be Valid and Reliable? | Q5. Is the Research Process Documented Adequately? | Q6. Are All Study Questions Answered? | Q7. Are the Negative Findings Presented? | TOTAL SCORE |
---|---|---|---|---|---|---|---|---|
[ ] | 1 | 2 | 2 | 2 | 2 | 2 | 1 | 12 |
[ ] | 1 | 2 | 1 | 1 | 1 | 1 | 1 | 8 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 1 | 13 |
[ ] | 2 | 2 | 0 | 2 | 1 | 2 | 2 | 11 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 14 |
[ ] | 1 | 2 | 0 | 1 | 1 | 1 | 2 | 8 |
[ ] | 1 | 1 | 2 | 1 | 1 | 2 | 0 | 8 |
[ ] | 1 | 2 | 2 | 2 | 2 | 2 | 1 | 12 |
[ ] | 1 | 2 | 2 | 2 | 2 | 2 | 2 | 13 |
[ ] | 2 | 2 | 2 | 2 | 2 | 1 | 2 | 13 |
[ ] | 1 | 2 | 0 | 1 | 1 | 1 | 1 | 7 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 14 |
[ ] | 1 | 2 | 2 | 2 | 1 | 1 | 1 | 10 |
[ ] | 1 | 2 | 2 | 1 | 1 | 2 | 1 | 10 |
[ ] | 1 | 2 | 0 | 2 | 2 | 1 | 2 | 10 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 14 |
[ ] | 1 | 1 | 0 | 1 | 0 | 1 | 1 | 5 |
[ ] | 2 | 2 | 1 | 1 | 1 | 2 | 2 | 11 |
[ ] | 1 | 1 | 2 | 2 | 2 | 1 | 2 | 11 |
[ ] | 2 | 2 | 1 | 1 | 1 | 2 | 2 | 11 |
[ ] | 2 | 2 | 2 | 1 | 1 | 1 | 1 | 10 |
[ ] | 2 | 1 | 2 | 2 | 2 | 1 | 2 | 12 |
[ ] | 1 | 2 | 0 | 1 | 2 | 2 | 2 | 10 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 14 |
[ ] | 2 | 2 | 1 | 2 | 2 | 1 | 0 | 10 |
[ ] | 1 | 2 | 2 | 2 | 2 | 1 | 0 | 10 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 14 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 14 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 14 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 14 |
[ ] | 2 | 1 | 1 | 1 | 2 | 2 | 2 | 11 |
[ , ] | 2 | 1 | 0 | 1 | 2 | 2 | 2 | 10 |
[ ] | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 14 |
[ ] | 2 | 2 | 2 | 1 | 1 | 1 | 2 | 11 |
[ ] | 2 | 2 | 0 | 1 | 2 | 2 | 0 | 9 |
[ ] | 1 | 1 | 1 | 1 | 2 | 2 | 2 | 10 |
[ ] | 2 | 1 | 0 | 1 | 0 | 2 | 0 | 6 |
[ ] | 1 | 1 | 0 | 1 | 1 | 1 | 1 | 6 |
Data extraction form.
# | Extraction Element | Contents |
---|---|---|
1 | ID | |
2 | Reference | |
3 | SLR Category | Include vs Exclude |
4 | Title | The full title of the article |
5 | Year | The publication year |
6 | Repository | ACM, IEEE, Scopus, Web of Science |
7 | Type | Journal vs article |
8 | Intervention target | |
9 | Disorder (Target group) | |
10 | Target language | |
11 | Sample Size | |
12 | Participant characteristics | |
13 | Evaluation | |
14 | Outcome measure | |
15 | OST Name | |
16 | System objective | |
17 | Architecture design | |
18 | ML approach | |
19 | OST Technology details |
Sample size characteristics of experimental studies.
Author | Sample Size | Participant Characteristics |
---|---|---|
B | 2 | (1M, 1F): One 5-year-old Turkish speaking boy with no language/speech problem and one 5-year-old Turkish speaking girl with a language disorder because of hearing impairment. |
C | 356 | 5 to 9 years old Portuguese speaking children |
E | 40 | 5 to 6 years old Romanian speaking children, boys and girls with difficulties in pronunciation of R and S sounds. |
G | 5 | 2 to 6 years old English speaking children with hearing impairment. |
H | 1077 | 3 to 8 years old Portuguese speaking children. |
I | 60 | (22M, 20F): 42 14 to 60 years old Dutch-speaking children and adults without disabilities. (9M, 9F): 18 19 to 75 months old. Dutch-speaking children with severe cerebral palsy. |
J | 4 | (2M, 2F): 8 to 10 years old Portuguese speaking children |
K | 21 | (13M, 1F): fourteen 4 to 12 year old with diagnosed SSDs ranging from mild to severe (7 motor-speech and 7 phonological impairments). (4M, 3F): seven 5 to 12 years old children typically developing |
L | 30 | IT professionals |
M | 10 | Deaf, hard of hearing, implanted children and those who had a speech impediment. |
O | 11 | (6M, 5F): 5,2 to 6,9 years old German-speaking children suffering from a specific articulation disorder, i.e., [s]-misarticulation |
P | 5 | Portuguese speaking. 2 females (30 and 46 years) and 3 males (ages: 13, 33, 36). The younger participant is the only one doing speech therapy. |
Q | 18 | 16 boys and 2 girls were recruited from three psychology offices. Their mean age was 10.54 years (range 2–16; std 4.34). |
R | 1 | A 4-year-old and a 6-year-old. |
S | 32 | Children of regular schools |
T | 12 | Italian speaking children |
U | Unknown | Children with disabilities and typically developing children. |
V | 8 | (3M, 1F): 3 to 7 years old children clinically diagnosed with apraxia of speech. |
X | 22 | 2-year-old children |
Y | 53 | Children with different types of disabilities and cognitive ages from 0 to 7 years |
Z | 53 | Children with different types of disabilities and cognitive ages from 0 to 7 years |
AA | 35 | (13M, 7F): 20 7 to 10 years old children with no speech or language impairments (9M, 6F): 15 7 to 9 year old with speech and language impairments |
BB | 27 | 11 to 34 years old children and adults with mild to severe mental delay or a communication disorder. |
DD | 143 | 43 parents and 100 teachers (kindergarten and primary school |
FF | Unknown | Children and adults with different levels of dysarthria. |
GG | 20 | (11M, 9F): 15 to 55 years old CP volunteers with speech difficulties and motor impairment. |
II | 14 | (7M, 7F): 11 to 21 years old children and adults with physical and psychical handicaps like cerebral palsy, Down’s syndrome and similar impairments. |
This research received no external funding.
Conceptualization, G.A.A. and B.T.; methodology, G.A.A.; validation, G.A.A., K.E.B. and B.T.; writing—original draft preparation, G.A.A. and K.E.B.; writing—review and editing, G.A.A., K.E.B. and B.T.; supervision, K.E.B. and B.T. All authors have read and agreed to the published version of the manuscript.
Conflicts of interest.
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.
An official website of the United States government
The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.
The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.
Email citation, add to collections.
Your saved search, create a file for external citation management software, your rss feed.
Affiliations.
This research was aimed to investigate changes in the reading technique and in terms of its semantic charge in primary schoolers diagnosed with dyslexia, which occur as a result of the integrated use of speech therapy techniques. The study was performed between 2016 and 2019 in 6 schools of Moscow and Almaty. It enrolled 194 and 200 children, respectively, who were examined with form I to III inclusive. The study revealed that 13% of children had reading speed disorders; they were constituted group 1. Another 11% had reading comprehension disorders; they constituted group 2. In group 1, by form III, the number of reading repetitions increased twofold. In group 2, the number of children, who read in words and phrases, increased by half; in group 1, it doubled. This research showed clear progress in children with technical dyslexia vs. those with semantic dyslexia. Based on the results, it is possible to develop a methodology for speech therapy techniques that can be suitable not only for speech therapists, but also for primary school teachers, as well as for parents of dyslectic children.
© 2023. The Author(s).
PubMed Disclaimer
The authors declare no competing interests.
Linkout - more resources, full text sources.
NCBI Literature Resources
MeSH PMC Bookshelf Disclaimer
The PubMed wordmark and PubMed logo are registered trademarks of the U.S. Department of Health and Human Services (HHS). Unauthorized use of these marks is strictly prohibited.
Physical Therapy
Occupational Therapy
Gritman Therapy Solutions is proud to offer comprehensive, evidence-based, multidisciplinary treatment for children from birth to 18 years of age. Our expertly trained pediatric therapists work closely with a child’s family to build a personalized treatment plan to address each child’s individual needs. We provide occupational, physical and speech-language services that are focused on helping a child develop and succeed.
Whether you or your loved one are recovering or developing through life’s milestones, our compassionate, expert staff can assist you in a variety of ways. Our multi-disciplinary, team approach to therapy provides our pediatric patients with a full range of therapy expertise.
Speech and Language Disorders
If you are a new patient to Gritman Pediatric Therapy or if it has been more than 6 months since your last visit please print one of our patient intake packets and answer each question as completely as possible before arriving for your appointment.
You can use your computer to edit and save the PDFs. Please email the PDFs with a copy of your insurance card(s) to [email protected] .
You can alternately fill in the forms on your computer and print them and mail them or bring them in to your initial appointment.
See open therapy positions
COMMENTS
Below is a list of topics I talked about in my essay, in order! This is obviously very personalized to me, and my life, but will hopefully help in deciding how and where to start! 1) Intro: I began with a personal story about my brother, what originally sparked my interest in speech-language pathology. 2) Paragraph 2: My work experience during ...
Description of the condition. Speech and/or language disorders are amongst the most common developmental difficulties in childhood. Such difficulties are termed 'primary' if they have no known aetiology, and 'secondary' if they are caused by another condition such as autism, hearing impairment, general developmental difficulties, behavioural or emotional difficulties or neurological impairment ...
In each folder, create a copy of your "base draft" of the essay. In a new word document, put that school's specific essay requirements at the top. Into the document, paste chunks of your "base essay" together and start tweaking it. One thing to help customize your essay is simply to mention the school's name.
Speech pathology statement of purpose examples #1: "My passion for helping others communicate more effectively, combined with my academic background in linguistics, makes me a strong candidate for this program. Ever since I assisted in a local school where children with speech difficulties were often left behind, I realized the profound ...
29 days ago. Sure, your speech impediment can make for a compelling topic for your college essay. You want to show colleges parts of your identity that don't necessarily appear elsewhere in your application, and your experience with a speech impediment can fall under that. The critical part is how you draw connections between your speech ...
Speech-language therapy has been shown to be useful for some though not all disorders of language and speech disorders (53, 57). Nonetheless, based on this research, clinical consensus is that children should be referred for language and/or speech therapy, either in isolation or as part of an early intervention or special education program.
Introduction. Speech development disorders are common problems or difficulties people have when communicating. They are grouped into two broad types namely: Congenital and acquired. Congenital disorders normally are present from birth meaning that a child is born with it. In most cases, mothers who have problems during the delivery process do ...
Program:Speech-Language Pathology. Posted March 22, 2013. On 3/22/2013 at 12:23 PM, midnight streetlight said: Ha, a large part of my SOP was about my mother-in-law's aphasia and acting as her caretaker over the summer! Her struggles with language have absolutely informed my decision to become an SLP.
speech therapy OR speech and language therapy OR speech pathology OR speech and language pathology OR speech ... The final papers will be read in-depth several times by the first author (JM) to facilitate initial data extraction. The date of the study, author(s) as well as the sample population or context will be extracted from all included ...
American Speech-Language-Hearing Association 2200 Research Blvd., Rockville, MD 20850 Members: 800-498-2071 Non-Member: 800-638-8255. MORE WAYS TO CONNECT
I specialize in providing speech therapy to help children, teens, and young adults to improve language, reading, writing, and executive functioning skills. Now offering in-person sessions in Chicago! Learn more about me on my About Hollis page. Whether you are writing an essay for your high school class or college course, you need to look at ...
Speech Language Pathology Essay Examples. 734 Words3 Pages. The profession of Speech Language Pathology enables others to be heard and gives them the ability to have a voice. As a Communication Disorders major, I found my voice through education and personal experiences. During my undergraduate career, I have balanced extracurricular activities ...
Narrative intervention uses an authentic, functional context, which entails the formation of a genuine connection between adult and child, where meaningful information is conveyed, and language processing, pragmatics, and social-emotional learning are integrated (Brinton & Fujiki, 2019).
Participants in the study by Hegarty et al. (2020) tended to use long-standing approaches such as minimal pairs and speech discrimination therapy, despite recognizing that newer and more complex approaches could be more appropriate for specific presentations (Hegarty et al., 2020). These combined findings raise questions around how information ...
Social Communication and Parent Verbal Responsiveness Across Interaction Contexts in Toddlers on the Autism Spectrum. Abigail Delehanty, Ciera M. Lorio, Mollie Romano, Jennifer A. Brown, Juliann J. Woods and. Amy M. Wetherby. American Journal of Speech-Language PathologyResearch Article1 May 2024.
Four principles of. clinical ethics may guide ethical decision-making in speech-language pathology: (1) autonomy; (2) beneficence; (3) nonmaleficence; and (4) justice. Ethical decisions require ...
Abstract. This research was aimed to investigate changes in the reading technique and in terms of its semantic charge in primary schoolers diagnosed with dyslexia, which occur as a result of the ...
New York University (NYU) Application Processing Center. Speech@NYU. PO Box 30096 010-003. College Station, TX 77842. To be considered an official transcript, the transcript must be sent directly from your institution (s) or through an electronic transcript vendor retained by that institution.
1. Introduction. Young children judge each other based on their communication skills, and therefore, a communication disorder can harm someone's social status at a young age [].Children enrolled in therapy before the age of five experience more positive outcomes than children that enroll after this age [].Even when access to a speech-language pathologist (SLP) is possible, SLPs often ...
Speech. Speech Therapy*. This research was aimed to investigate changes in the reading technique and in terms of its semantic charge in primary schoolers diagnosed with dyslexia, which occur as a result of the integrated use of speech therapy techniques. The study was performed between 2016 and 2019 in 6 schools of Moscow a ….
Find Out More Information. We are happy to answer questions and share personal experiences about the services we provide. 208-883-1522. Gritman therapists offer solutions for a variety of therapy needs. Care through therapy is essential for proper healing, injury relief and recovery.
Recommendations from the preeminent models of evidence-based practice (EBP) in speech-language pathology (American Speech-Language-Hearing Association [ASHA], n.d.-a, 2004a, 2004b; Dollaghan, 2007) suggest that clinicians should identify and critically appraise evidence from research, clinical, and patient sources, and then integrate these to ...
Gritman Therapy Solutions is proud to offer comprehensive, evidence-based, multidisciplinary treatment for children from birth to 18 years of age. Our expertly trained pediatric therapists work closely with a child's family to build a personalized treatment plan to address each child's individual needs. We provide occupational, physical and ...