As two words with opposite meanings, dou1 and shao3 are expected to be similar and different in various environments. In this work, we looked into the construction of [numeral + measure word/classifier + noun] (hereafter [Num + MW/CL + N]) and explored in what ways both words present an asymmetric phenomenon. We found that dou1 carries a numeral meaning while shao3 lacks this use. Based on the analysis of these two words in the Sinica Corpus, this paper argued that dou1 is better categorized as ‘Neu’ and suggested that dou1 in [Num + MW/CL + N] serves two functions: one is for counting numbers; the other is for the expression of quantities. These findings can be related to the use of dou1 as complement and as numeral. 關鍵詞:「多」和「少」、數量名結構、反義詞、數量概念
In recent years, community-based question and answer (CQA) sites have grown rapidly in number and size. These sites represent a valuable source of online knowledge; however, they often suffer from the problem of duplicate questions. The task of question retrieval (QR) aims to find previously answered semantically similar questions in CQA archives. Nevertheless, synonymous lexical variations pose a big challenge for question retrieval. Some QR approaches address this issue by calculating the probability of correlation between new questions and archived questions. Much recent research has also focused on surface string similarity among questions. In this paper, we propose a method that first builds a continuous bag-of-words (CBoW) model with data from Asus’s Republic of Gamers (ROG) forum and then determines the similarity between a given new question and the Q&As in our database. Unlike most other methods, we calculate the similarity between the given question and the archived questions and descriptions separately with two different features. In addition, we factor user reputation into our ranking model. Our experimental results on the ROG forum dataset show that our CBoW model with reputation features outperforms other top methods.
Due to ever-increasing amounts of publicly available multimedia associated with speech information, spoken document retrieval (SDR) has been an active area of research that captures significant interest from both academic and industrial communities. Beyond the continuing effort in the development of robust indexing and effective retrieval methods to quantify the relevance degree between a pair of query and spoken document, how to accurately and efficiently model the query content plays a vital role for improving SDR performance. In view of this, we present in this paper a novel neural relevance-aware model (NRM) to infer an enhanced query representation, extricating the conventional time-consuming pseudo-relevance feedback (PRF) process. In addition, we incorporate the notion of query intent classification into our proposed NRM modeling framework to obtain more sophisticated query representations. Preliminary experiments conducted on the TDT-2 collection confirm the utility of our methods in relation to a few state-of-the-art ones. Keyword : Spoken Document Retrieval, Query Intent, Neural Network, Pseudo-Relevance Feedback 1. 緒論 (INTRODUCTION) 伴隨著網際網路的發展與多媒體資訊的大量增長,影音的瀏覽與傳遞也逐漸成為我們的 日常生活的一部分。在這環境下,如何利用語音的資訊,快速檢索符合資訊需求的內容, 變成了一項新興的需求。因此,在過去的二十年(Chelba, Hazen & Saraclar, 2008) (Lee & Chen, 2005) (Huang, Ma, Li & Wu, 2011) (Chen, Chen, Chen & Chen, 2012),語音文件檢索 成為一個十分有魅力的研究主題。在語音文件檢索的任務上,過往有許多顯著成功的方 法,如向量空間模型(Vector Space Model) (Salton, Wong & Yang, 1975)、Okapi BM25 model (Jones, Walker & Robertson, 2000),以及主題模型(Topic Model) (Blei, Ng & Jordan, 2003)等。另一方面,將統計式語言模型(Statistical Language Model)應用在文字檢索 (Information Retrieval)和語音文件檢索,在檢索任務上取得了嶄新的突破(Ponte & Croft, 1998) (Song & Croft, 1999) (Croft & Lafferty, 2003),因此吸引了不少研究者的目光。在這 樣的概念下,查詢對每個文件計算似然機率後作排名,我們稱這樣的排序方法為查詢似 然測量(Query Likelihood Model Measure, QLM) (Manning, Raghavan & Schutze, 2008)。另 一個知名的評估方式為 KL 散度測量(Kullback-Leibler Divergence Measure, KLM) (Zhai & Lafferty, 2001),將查詢與文件皆表示為單元語法的語言模型(Unigram Language Model), 查詢與文件的相似程度即為兩個機率分佈的散度距離(Divergence Distance)。 最近,隨著深層類神經網路架構的流行,這類的方法也被大量應用在檢索的任務上。 主要的研究方向為利用不同網路架構與訓練方法,以此來學習查詢與文件間的相似關係 (Guo, Fan, Ai & Croft, 2016) (Mitra, Diaz & Craswell, 2017)。值得注意的是,大部分方法
Text readability refers to the degree to which a text can be understood by its readers: the higher the readability of a text for readers, the better the the 國立臺灣師範大學資訊工程系 Department of Computer Science & Information Engineering, National Taiwan Normal University E-mail: ouartz99@gmail.com, berlin@cise.ntnu.edu.tw 國立臺灣師範大學教育心理與輔導學系 Department of Educational Psychology and Counseling, National Taiwan Normal University E-mail: sungtc@ntnu.edu.tw
To eliminate acoustic echo, the convergence rate and low residual echo are very important to adaptive echo cancelers. Meanwhile, an affordable computational complexity has to be considered as well. In this paper, we proposed the improved vector space adaptive filter (IVAF)and Improved Vector-space Affine Projection Sign Algorithm (IVAPSA). The proposed can be divided into two phases: offline and online. In the offline phase, IVAF constructs a vector space to incorporate the prior knowledge of adaptive filter coefficients from a wide range of different channel characteristics. Then, in the online phase, the IVAF combines the conventional APSA and IVAPSA algorithms, where IVAPSA computes the filter coefficients based on the vector space obtained in the offline phase. By leveraging the constructed vector space, the proposed IVAF is able to fast converge and achieve a better echo return loss enhancement performance. Moreover, the computational complexity is less than a comparable work. 關鍵詞:回聲消除系統,可適性濾波器,向量空間可適性濾波器,機器學習,組合演算法,仿 射投影符號演算法
Speech is one of the most natural form of human communication. Recognizing emotion from speech continues to be an important research venue to advance human-machine interface design and human behavior understanding. In this work, we propose a novel set of features, termed trajectory-based spatial-temporal spectral features, to recognize emotions from speech. The core idea centers on deriving descriptors both spatially and temporally on speech spectrograms over a sub-utterance frame (e.g., 250ms) an inspiration from dense trajectory-based video descriptors. We conduct categorical and dimensional emotion recognition experiments and compare our proposed features to both the well-established set of prosodic and spectral features and the state-of-the-art exhaustive feature extraction. Our experiment demonstrate that our features by itself achieves comparable accuracies in the 4-class emotion recognition and valence detection task, and it obtains a significant improvement in the activation detection. We additionally show that there exists complementary information in our proposed features to the existing acoustic features set, which can be used to obtain an improved emotion recognition accuracy.
Rescoring approaches for parsing aims to re-rank and change the order of parse trees produced by a general parser for a given sentence. The re-ranking performance depends on whether or not the rescoring function is able to precisely estimate the quality of parse trees by using more complex features from the whole parse tree. However it is a challenge to design an appropriate rescoring function since complex features usually face the severe problem of data sparseness. And it is also difficult to obtain sufficient information requisite in re-estimatation of tree structures because existing annotated Treebanks are generally small-sized. To address the issue, in this paper, we utilize a large amount of auto-parsed trees to learn the syntactic and sememtic information. And we propose a simple but effective score function in order to integrate the scores provided by the baseline parser and dependency association scores based on dependency-based word embeddings, learned from auto-parsed trees. The dependency association scores can relieve the problem of data sparseness, since they can be still calculated by word embeddings even without occurrence of a dependency word pair in a corpus. Moreover, semantic role labels are also considered to distinct semantic relation of word pairs. Experimental results show that our proposed model improves the base Chinese parser significantly.
Chinese Segmentation Ambiguity (CSA) is a fundamental problem confronted when processing Chinese language, where a sentence can generate more than one segmentation paths. Two techniques are commonly used to identify CSA: Omni-segmentation and Bi-directional Maximum Matching (BiMM). Due to the high computational complexity, Omni-segmentation is difficult to be applied for big data. BiMM is easier to be implemented and has a higher speed. However, recall of BiMM is much lower. In this paper, a Segmentation Matrix (SM) method is presented, which encodes each sentence as a matrix, then maps string operation into set operations. To identify CSA, instead of scanning a whole sentence, only specific areas of the matrix are checked. SM has a computational complexity close to BiMM with recall the same as Omni-segmentation. In addition to CSA identification, SM also supports lexicon-based Chinese word segmentation. In our experiments, based on SM, several issues about CSA are explored. The result shows that SM is useful for CSA analysis.
Previous studies on emotion classification mainly focus on the emotional state of the writer. By contrast, our research emphasizes emotion detection from the readers' perspective. The classification of documents into reader-emotion categories can be applied in several ways, and one of the applications is to retain only the documents that trigger desired emotions to enable users to retrieve documents that contain relevant contents and at the same time instill proper emotions. However, current information retrieval (IR) systems lack the ability to discern emotions within texts, and the detection of reader's emotion has yet to achieve a comparable performance. Moreover, previous machine learning-based approaches generally use statistical models that are not in a human-readable form. Thereby, it is difficult to pinpoint the reason for recognition failures and understand the types of emotions that the articles inspired on their readers. In this paper, we propose a flexible emotion template-based approach (TBA) for reader-emotion detection that simulates such process in a human perceptive manner. TBA is a highly automated process that incorporates various knowledge sources to learn an emotion template from raw text that characterize an emotion and are comprehensible for humans. Generated templates are adopted to predict reader's emotion through an alignment-based matching algorithm that allows an emotion template to be partially matched through a statistical scoring scheme. Experimental results demonstrate that our approach can effectively detect reader's emotions by exploiting the syntactic structures and semantic associations in the context, while outperforming currently well-known statistical text classification methods and the stat-of-the-art reader-emotion detection method.
Conventional search engines usually consider a search query corresponding only to a simple task. Nevertheless, due to the explosive growth of web usage in recent years, more and more queries are driven by complex tasks. A complex task may consist of multiple sub-tasks. To accomplish a complex task, users may need to obtain information of various task-related entities corresponding to the sub-tasks. Users usually have to issue a series of queries for each entity during searching a complex search task. For example, the complex task ＂travel to Beijing＂ may involve several task-related entities, such as ＂hotel room,＂ ＂flight tickets,＂ and ＂maps＂. Understanding complex tasks with task-related entities can allow a search engine to suggest integrated search results for each sub-task simultaneously. To understand and improve user behavior when searching a complex task, we propose an entity-driven complex task model (ECTM) based on exploiting microblogs and query logs. Experimental results show that our ECTM is effective in identifying the comprehensive task-related entities for a complex task and generates good quality complex task names based on the identified task-related entities.
