Search

With the increasing scale of mental health problems in modern society, the scarcity of professional assistance is alarming, especially in developing countries. To address this, some online forums have emerged to provide users with useful information and help. However, a user grappling with mental health problems often struggles to find the needed information and assistance on these forums. This is primarily due to the limitations of existing search approaches that often fail to take the characteristics of mental health text into account. In this paper, we propose a new task of mental-health-oriented question-answering (MHQA) which aims to retrieve the appropriate responses for a question post by incorporating the important criteria related to mental health. Our proposed approach, JMS-QA, matches the question post and candidate responses while jointly detecting their latent mental health signals. This enables the method to incorporate mental health signals into its representations. To test the effectiveness of our approach, we create a new dataset for MHQA and conduct experiments on it. The experimental results show that JMS-QA outperforms existing state-of-the-art methods.

Extracting various types of event relations in financial texts can benefit many downstream applications supporting financial analysis. This paper addresses the multi-type event relation extraction problem in the finance domain focusing on handling several issues in existing studies, including (1) limited event relation types involved, (2) insufficient feasibility when handling non-annotated data, (3) ineffectiveness in recognizing multi-type event relation, and (4) the asynchronous event and event relation extraction process. To tackle these limitations, we carefully define six types of event relations based on the characteristics of financial texts (e.g., abundant numerical words) and further devise an integral framework for Chinese financial event relation extraction. The framework is capable of handling unsupervised event extraction and event relation recognition jointly. Specifically, according to linguistic characteristics, a Core Verb Chain is employed for the event identification. Then, by constructing a Syntactic Semantic Dependency Parsing graph, scattered events are combined into pairs, and event ellipsis elements can be completed to prevent event information loss. Also, to capture more sentence semantics, we formulate an Event Restore module that converts the structured event pairs into event restore sentence pairs and pour the pairs into the BERT model for relation type identification. Finally, to enhance the embeddings for event core elements, an Event Core Embeddings layer is augmented in BERT, and we fine-tune the model on our annotated financial corpus. Extensive experiments are conducted to validate the effectiveness of our method, and the results confirm its superiority over the state-of-the-art baselines. • We define six event relation types based on practical needs in finance. • We propose a framework for multi-type event relation extraction, called CFERE. • Event sentence pairs are used to minimize the loss of input information. • An event core embedding layer is added to enhance the encoding of BERT. • We present a novel corpus and the effectiveness of CFERE on two datasets. [ABSTRACT FROM AUTHOR]

Online mental health communities have become a major platform where individuals can talk about their mental health problems and obtain social support. This study aims to understand the antecedents of perceived usefulness among members in an online mental health community, while providing reference for the managers and users of online mental health communities. We obtained a total of 143,190 posts from ReachOut.com released by the CLPsych2017 shared task. Then, we used text mining to derive the independent and dependent variables. Next, a structural equation model observing the perceived usefulness of online mental health community members was constructed from the perspective of an information adoption model. The informativeness of help‐seeking posts had a significant positive relationship with the information quality of reply posts; the information quality of reply posts was a significant positive predictor of perceived usefulness, with the information quality of reply posts partially mediating the relationship between the informativeness of help‐seeking posts and perceived usefulness. The information provided by online mental health community members' help‐seeking posts and the quality of replies were found to be the factors that influenced perceived usefulness. This study highlights the importance of the information quality of reply posts and provides useful insights for administrators who can help users to improve their response quality and obtain the support they need.

The Latent Dirichlet Allocation (LDA) model, which is a document-level probabilistic model, has been widely used in topic modeling. However, an essential issue of the LDA is its shortage in identifying co-occurrence relationships (e.g., aspect-aspect, aspect-opinion, etc.) in sentences. To address the problem, we propose an association constrained LDA (AC-LDA) for effectively capturing the co-occurrence relationships. Specifically, based on the basic features of the syntactic structure in product reviews, we formalize three major types of word association combinations and then carefully design corresponding identifications. For reducing the influence of global aspect words on the local distribution, we apply an important constraint on global aspects. Finally, the constraint and related association combinations are merged into the LDA to guide the topic-words allocation in the learning process. Based on the experiments on real-world product review data, we demonstrate that our model can effectively capture the relationships hidden in local sentences and further increase the extraction rate of fine-grained aspects and opinion words. Our results confirm the superiority of the AC-LDA over the state-of-the-art methods in terms of the extraction accuracy. We also verify the strength of our method in identifying irregularly appeared terms, such as non-aspect opinions, low-frequency words, and secondary aspects. [ABSTRACT FROM AUTHOR]

Social media data, e.g. Tweets, are usually geo-tagged, embedded with creation or posting time, and associated with texts. Nowadays, there is an increasing need for querying such spatio-temporal-text data. In this work, we propose a new type of query, top- k spatio-temporal keyword query ( k -STKQ in short), over Twitter-like social media data. A k -STKQ takes a location, a timestamp and a set of keywords as argument, and returns top- k objects that are near the location, close to the timestamp, and relevant to the set of keywords. An example of k -STKQ is to search the tweets mentioning “garage sale” recently sent from some places nearby. The massive amount and dynamic nature of social media data are the primary obstacles towards efficient processing of k -STKQs. In order to return the answers efficiently, we propose a novel index, TiST, for the processing of k -STKQs. TiST partitions the incoming data into subsets, and builds an R-tree index on each subset. The timestamps and texts of the objects are also integrated with the R-trees. To further boost the indexing performance, we propose a routing R-tree based R-tree insertion method, which is inspired by the observation that many sets of objects are similar in their locations. For the texts of objects, we propose a hybrid bitmap-based index, which is space-saving and supports relevance computation. The query processing algorithm is also presented based on the TiST index. We conduct extensive experiments to demonstrate that our solution is capable of providing excellent indexing performance and good query performance. [ABSTRACT FROM AUTHOR]

Abstract: Text-centric (or document-centric) XML document retrieval aims to rank search results according to their relevance to a given query. To do this, most existing methods mainly rely on content terms and often ignore an important factor – the XML tags and paths, which are useful in determining the important contents of a document. In some previous studies, each unique tag/path is assigned a weight based on domain (expert) knowledge. However, such a manual assignment is both inefficient and subjective. In this paper, we propose an automatic method to infer the weights of tags/paths according to the topical relationship between the corresponding elements and the whole documents. The more the corresponding element can generalize the document’s topic, the more the tag/path is considered to be important. We define a model based on Average Topic Generalization (ATG), which integrates several features used in previous studies. We evaluate the performance of the ATG-based model on two real data sets, the IEEECS collection and the Wikipedia collection, from two different perspectives: the correlation between the weights generated by ATG and those set by experts, and the performance of XML retrieval based on ATG. Experimental results show that the tag/path weights generated by ATG are highly correlated with the manually assigned weights, and the ATG model significantly improves XML retrieval effectiveness. [Copyright &y& Elsevier]

Abstract: Keyword search on XML documents has received considerable research interests recently. Most existing methods put their emphases on the document side, and focus on how to utilize structural properties of XML documents to produce better search results, more effective ranking methods, or more efficient algorithms. However, effective XML search requires a full understanding of not only XML documents but also XML keyword queries, whereas little attention has been paid to the latter. In this paper, we focus on the query side of XML keyword search instead of the document side. We show that keyword queries have structures, and define a concept called keyword query with structure (QWS) to capture query structure. As query structure provides hints about the intent of the query, it can be used to improve the quality of the search results. We exploit some key observations to characterize the structure in a keyword query and show how to refine search results with the assistance of query structure. In order to take the benefits of query structure, we design a query processing approach to produce results given a keyword query. It first derives some QWSs based on heuristics, and computes results of these queries, then expands the results if needed. We implement the proposed methods and conduct comprehensive experiments. Experimental results verify the effectiveness of our methods. [Copyright &y& Elsevier]