Search results

Intent detection (ID) and slot filling (SF) are two important tasks in natural language understanding. ID is to identify the main intent of a paragraph of text. The goal of SF is to extract the information that is important to the intent from the input sentence. However, most of the existing methods use sentence-level intention recognition, which has the risk of error propagation, and the relationship between intention recognition and SF is not explicitly modeled. Aiming at this problem, this paper proposes a collaborative model of ID and SF for intelligent spoken language understanding called ID-SF-Fusion.

ID-SF-Fusion uses Bidirectional Encoder Representation from Transformers (BERT) and Bidirectional Long Short-Term Memory (BiLSTM) to extract effective word embedding and context vectors containing the whole sentence information respectively. Fusion layer is used to provide intent–slot fusion information for SF task. In this way, the relationship between ID and SF task is fully explicitly modeled. This layer takes the result of ID and slot context vectors as input to obtain the fusion information which contains both ID result and slot information. Meanwhile, to further reduce error propagation, we use word-level ID for the ID-SF-Fusion model. Finally, two tasks of ID and SF are realized by joint optimization training.

We conducted experiments on two public datasets, Airline Travel Information Systems (ATIS) and Snips. The results show that the Intent ACC score and Slot F1 score of ID-SF-Fusion on ATIS and Snips are 98.0 per cent and 95.8 per cent, respectively, and the two indicators on Snips dataset are 98.6 per cent and 96.7 per cent, respectively. These models are superior to slot-gated, SF-ID NetWork, stack-Prop and other models. In addition, ablation experiments were performed to further analyze and discuss the proposed model.

This paper uses word-level intent recognition and introduces intent information into the SF process, which is a significant improvement on both data sets.

This study aims to explore motives behind teachers' and students' use of translanguaging and how they use it in Tanzanian public secondary school classrooms.

Data were collected using interviews and non-participant observations.

The findings indicate that translanguaging was used to facilitate content comprehension, promote classroom interaction and increase students' motivation to learn. Translanguaging was implemented using three strategies: paraphrasing an English text into Kiswahili, translating an English text into its Kiswahili equivalent and word-level translanguaging.

By highlighting the motivations for translanguaging and corresponding strategies associated with translanguaging pedagogy in the Tanzanian context, this study has significant practical implications for teachers and students to showcase their linguistic and multimodal knowledge, while fostering a safe learning space that relates to students' daily experiences.

The study offers new insights into previous research on the role of language-supportive pedagogy appropriate for teachers and students working within bi-/multilingual education settings.

Hate speech is an expression of intense hatred. Twitter has become a popular analytical tool for the prediction and monitoring of abusive behaviors. Hate speech detection with social media data has witnessed special research attention in recent studies, hence, the need to design a generic metadata architecture and efficient feature extraction technique to enhance hate speech detection.

This study proposes a hybrid embeddings enhanced with a topic inference method and an improved cuckoo search neural network for hate speech detection in Twitter data. The proposed method uses a hybrid embeddings technique that includes Term Frequency-Inverse Document Frequency (TF-IDF) for word-level feature extraction and Long Short Term Memory (LSTM) which is a variant of recurrent neural networks architecture for sentence-level feature extraction. The extracted features from the hybrid embeddings then serve as input into the improved cuckoo search neural network for the prediction of a tweet as hate speech, offensive language or neither.

The proposed method showed better results when tested on the collected Twitter datasets compared to other related methods. In order to validate the performances of the proposed method, t-test and post hoc multiple comparisons were used to compare the significance and means of the proposed method with other related methods for hate speech detection. Furthermore, Paired Sample t-Test was also conducted to validate the performances of the proposed method with other related methods.

Finally, the evaluation results showed that the proposed method outperforms other related methods with mean F1-score of 91.3.

The main novelty of this study is the use of an automatic topic spotting measure based on naïve Bayes model to improve features representation.

The purpose of this research is to compare several machine learning techniques on the task of Asian language text classification, such as Chinese and Japanese where no word boundary information is available in written text. The paper advocates a simple language modeling based approach for this task.

Naïve Bayes, maximum entropy model, support vector machines, and language modeling approaches were implemented and were applied to Chinese and Japanese text classification. To investigate the influence of word segmentation, different word segmentation approaches were investigated and applied to Chinese text. A segmentation‐based approach was compared with the non‐segmentation‐based approach.

There were two findings: the experiments show that statistical language modeling can significantly outperform standard techniques, given the same set of features; and it was found that classification with word level features normally yields improved classification performance, but that classification performance is not monotonically related to segmentation accuracy. In particular, classification performance may initially improve with increased segmentation accuracy, but eventually classification performance stops improving, and can in fact even decrease, after a certain level of segmentation accuracy.

Apply the findings to real web text classification is ongoing work.

The paper is very relevant to Chinese and Japanese information processing, e.g. webpage classification, web search.

Dysarthria is a neuromotor speech disorder caused by neuromuscular disturbances that affect one or more articulators resulting in unintelligible speech. Though inter-phoneme articulatory variations are well captured by formant frequency-based acoustic features, these variations are expected to be much higher for dysarthric speakers than normal. These substantial variations can be well captured by placing sensors in appropriate articulatory position. This study focuses to determine a set of articulatory sensors and parameters in order to assess articulatory dysfunctions in dysarthric speech.

The current work aims to determine significant sensors and parameters associated using motion path and correlation analyzes on the TORGO database of dysarthric speech. Among eight informative sensor channels and six parameters per channel in positional data, the sensors such as tongue middle, back and tip, lower and upper lips and parameters (y, z, φ) are found to contribute significantly toward capturing the articulatory information. Acoustic and positional data analyzes are performed to validate these identified significant sensors. Furthermore, a convolutional neural network-based classifier is developed for both phone-and word-level classification of dysarthric speech using acoustic and positional data.

The average phone error rate is observed to be lower, up to 15.54% for positional data when compared with acoustic-only data. Further, word-level classification using a combination of both acoustic and positional information is performed to study that the positional data acquired using significant sensors will boost the performance of classification even for severe dysarthric speakers.

The proposed work shows that the significant sensors and parameters can be used to assess dysfunctions in dysarthric speech effectively. The articulatory sensor data helps in better assessment than the acoustic data even for severe dysarthric speakers.

The purpose of this paper is to propose a similarity-based approach to accurately retrieve reference solutions for the intelligent handling of online complaints.

This approach uses a case-based reasoning framework and firstly formalizes existing online complaints and their solutions, new online complaints, and complaint products, problems and content as source cases, target cases and distinctive features of each case, respectively. Then the process of using existing word-level, sense-level and text-level measures to assess the similarities between complaint products, problems and contents is explained. Based on these similarities, a measure with high accuracy in assessing the overall similarity between cases is designed. The effectiveness of the approach is evaluated by numerical and empirical experiments.

The evaluation results show that a measure simultaneously considering the features of similarity at word, sense and text levels can obtain higher accuracy than those measures that consider only one level feature of similarity; and that the designed measure is more accurate than all of its linear combinations.

The approach offers a feasible way to reduce manual intervention in online complaint handling. Complaint products, problems and content should be synthetically considered when handling an online complaint. The designed procedure of the measure with high accuracy can be applied in other applications that consider multiple similarity features or linguistic levels.

A method for linearly combining the similarities at all linguistic levels to accurately assess the overall similarities between online complaint cases is presented. This method is experimentally verified to be helpful to improve the accuracy of online complaint case retrieval. This is the first study that considers the accuracy of the similarity measures for online complaint case retrieval.

Course recommendations are important for improving learner satisfaction and reducing dropout rates on massive open online course (MOOC) platforms. This study aims to propose an interpretable method of analyzing students' learning behaviors and recommending MOOCs by integrating multiple data sources.

The study proposes a deep learning method of recommending MOOCs to students based on a multi-attention mechanism comprising learning records attention, word-level review attention, sentence-level review attention and course description attention. The proposed model is validated using real-world data consisting of the learning records of 6,628 students for 1,789 courses and 65,155 reviews.

The main contribution of this study is its exploration of multiple unstructured information using the proposed multi-attention network model. It provides an interpretable strategy for analyzing students' learning behaviors and conducting personalized MOOC recommendations.

The findings suggest that MOOC platforms must fully utilize the information implied in course reviews to extract personalized learning preferences.

This study is the first attempt to recommend MOOCs by exploring students' preferences in course reviews. The proposed multi-attention mechanism improves the interpretability of MOOC recommendations.

This chapter discusses the findings of a qualitative study conducted on the US–Mexico border to investigate preservice bilingual teachers’ understandings of the effective practices needed to teach content in bilingual classrooms. Specifically, participants’ understandings of teaching language through content to emergent bilinguals and the role of academic language in a content methods course taught in Spanish for preservice bilingual teachers were explored. The results of the study show that preservice bilingual teachers struggled to internalize how to develop language objectives that embed the four language domains as well as the three levels of academic language into their content lessons. Although participants emphasized vocabulary development, they integrated multiple scaffolding strategies to support emergent bilinguals. Moreover, although preservice bilingual teachers struggled with standard Spanish, they used translanguaging to navigate the discourse of education in their content lessons. The use of academic Spanish was also evident in participants’ planning of instruction. The authors contend that bilingual teacher preparation would benefit from the implementation of a dynamic bilingual curriculum that: (a) incorporates sustained opportunities across coursework for preservice bilingual teachers to strengthen their understanding of content teaching and academic language development for emergent bilinguals; (b) values preservice bilingual teachers’ language varieties, develops metalinguistic awareness, and fosters the ability to navigate between language registers for teaching and learning; and (c) values translanguaging as a pedagogical strategy that provides access to content and language development.

Entrepreneurship is an activity with far-reaching economic and cultural implications. Research seeking to understand the cognition and behavior of entrepreneurs is devoting increasing attention to how entrepreneurs construct and utilize discourse. However, word-level analysis of the specific language used by entrepreneurs has not received significant attention. The purpose of this paper is to identify the words that comprise entrepreneurship discourse and describe how word-usage has changed over time.

To examine the language of entrepreneurship, the authors use modified MapReduce algorithms in conjunction with text mining techniques to construct a longitudinal corpus of approximately 2.5m words. The authors identify the most frequently used words in the entrepreneurship lexicon and then use content analysis to chart the evolution of word-use.

The findings reveal that entrepreneurs’ lexicon is complex and fluid. The most commonly used words suggest new trends in entrepreneurship.

The findings and methodological procedures contribute to research on entrepreneurs and the entrepreneurial process and, specifically, to work on entrepreneurial discourse, language-use and new venture communication. The findings also have implications for entrepreneurs and policymakers.

Being an important data exchange and information storage standard, XML has generated a great deal of interest and particular attention has been paid to the issue of XML indexing. Clear use cases for structured search in XML have been established. However, most of the research in the area is either based on relational database systems or specialized semi‐structured data management systems. This paper aims to propose a method for XML indexing based on the information retrieval (IR) system Okapi.

First, the paper reviews the structure of inverted files and gives an overview of the issues of why this indexing mechanism cannot properly support XML retrieval, using the underlying data structures of Okapi as an example. Then the paper explores a revised method implemented on Okapi using path indexing structures. The paper evaluates these index structures through the metrics of indexing run time, path search run time and space costs using the INEX and Reuters RVC1 collections.

Initial results on the INEX collections show that there is a substantial overhead in space costs for the method, but this increase does not affect run time adversely. Indexing results on differing sized Reuters RVC1 sub‐collections show that the increase in space costs with increasing the size of a collection is significant, but in terms of run time the increase is linear. Path search results show sub‐millisecond run times, demonstrating minimal overhead for XML search.

Overall, the results show the method implemented to support XML search in a traditional IR system such as Okapi is viable.

The paper provides useful information on a method for XML indexing based on the IR system Okapi.