Title加TextRank抽取关键句的情感分类研究

考虑到不同句子对判断文档情感倾向的重要程度不同，因而区分文档的关键句和细节句将有助于提高情感分类的性能。同时，考虑到Title和上下文信息，提出了一种基于Title和加权TextRank抽取关键句的情感分析方法SKTT，实现了高效的情感分析。根据文档Title的情感权重计算Title贡献度，考虑到标点和语义规则对情感倾向的影响；根据加权TextRank算法思想，在文档正文中构建了一个情感句有向图来提取关键句；计算所有关键句的情感倾向进行情感分类。在4个领域上进行实验，实验结果表明，该SKTT方法性能明显优于Baseline，具有高效性。     

面向智能客服的句子相似度计算方法

针对金融领域中智能客服的句子相似度计算方法进行了研究。利用基于词性的分词纠正模型减少中文歧义词、金融相关词汇的分词错误；通过词向量方法和循环神经网络分别提取词语级和句子级的语义特征，并且得到句子向量；用融合层计算出句子向量间的差异特征；对差异特征进行降维和归一化得到句子相似度计算结果。实验结果表明，该方法具有较高的准确率和[F1]值。     

论我国区际法律冲突下"一事两诉"问题的解决

“一事两诉”问题在中国的区际法律冲突中经常发生。台湾地区与内地是通过承认与执行法院判决而间接解决这一问题的,但仍遗留了不少问题。目前,内地与香港既无承认与执行法院判决的安排,也无解决“一事两诉”问题的协议。因此,如何解决中国区际法律冲突下“一事两诉”,是实践中迫切需要解决的问题。     

Inference control of open relational queries under closed-world semantics based on theorem proving

Relational database systems may serve to evaluate an open query under closed-world semantics. The evaluation returns an explicit output relation complemented with an often implicit statement about the completeness of that relation. The output relation is formed from all those tuples that both fit the format and satisfy the properties expressed in the query. Using first-order logic for specifying formal semantics, the output relation can be seen as a set of (ground) sentences obtained from the query formula by suitable substitutions of free variables by constants. A statement about the completeness of a relation can also explicitly be formalized as a sentence of first-order logic. Inference control for enforcing a confidentiality policy has to inspect and to possibly distort not only the sentences representing the tuples of the output relation but also the completeness sentences. Previously designed and formally verified control procedures employ theorem-proving for such inspections while iteratively considering candidates for those sentences and determining termination conditions, respectively. In this article, we outline an implementation of these control procedures and treat improvements of their runtime efficiency, in particular to overcome shortcomings of the underlying theorem prover, which is repeatedly called with an input comprising a completeness sentence of increasing size. The improvements are obtained by an equivalent rewriting of completeness sentences, exploiting the active domain or introducing new constants for combinations of the original constants, respectively, as well as by optimizing the number of such calls. Besides theoretical complexity considerations, we also present practical evaluations for some examples. These examples include queries that—without control—would return the whole underlying database relations and—with control—can be used for confidentiality-preserving data publishing.     

AN OPINION FEATURE EXTRACTION APPROACH BASED ON A MULTIDIMENSIONAL SENTENCE ANALYSIS MODEL

With Web 2.0 applications being widely used, social networking services, including web blogs, forums, and other online communities, have become informative tools that help individuals to easily gauge the pulse of the electronic consuming market. As a substitute for traditional public media, the related site provides unique mechanisms to instantly reveal the degree of public product acceptance by either statistically aggregating the rating results or archiving opinions shared by experienced customers. However, the growth of user-generated information and its scattered unstructured contents is overwhelming to users, thereby triggering the demand for a more efficient system that can offer concise information. Most existing efforts dedicated to these issues may neglect vital aspects of the sentence-level context. This article aims to explore the critical features hidden in the sentential structure of opinion articles and expects that the detected patterns may contribute to the enhancement of related applications. Accordingly, a multidimensional sentence modeling algorithm (MSMA) is designed to evaluate various sentential characteristics and adopt a genetic algorithm to optimize the weighting scheme while determining feature importance. The study also makes use of the public knowledge resource Wikipedia as a global reference to fine-tune the feature set's effectiveness and enhance the overall performance of this framework. The results of experiments on an electronic product data set demonstrate that the proposed method is promising and provides significant improvement over previous studies.     

Phonological typicality influences sentence processing in predictive contexts: Reply to Staub, Grant, Clifton, and Rayner (2009).

In 2 separate self-paced reading experiments, Farmer, Christiansen, and Monaghan (2006) found that the degree to which a word's phonology is typical of other words in its lexical category influences online processing of nouns and verbs in predictive contexts. Staub, Grant, Clifton, and Rayner (2009) failed to find an effect of phonological typicality when they combined stimuli from the separate experiments into a single experiment. We replicated Staub et al.'s experiment and found that the combination of stimulus sets affects the predictiveness of the syntactic context; this reduces the phonological typicality effect as the experiment proceeds, although the phonological typicality effect was still evident early in the experiment. Although an ambiguous context may diminish sensitivity to the probabilistic relationship between the sound of a word and its lexical category, phonological typicality does influence online sentence processing during normal reading when the syntactic context is predictive of the lexical category of upcoming words. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

Still no phonological typicality effect on word reading time (and no good explanation of one, either): A rejoinder to Farmer, Monaghan, Misyak, and Christiansen.

In this brief rejoinder, we respond to Farmer, Monaghan, Misyak, and Christiansen (2011). We argue that the data still do not support the claim that reading time is affected by the phonological typicality of a word for its part of speech. We also question Farmer et al.'s claim that interleaving syntactic structures in an experiment modifies grammatically based syntactic expectations. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

Response latencies in auditory sentence comprehension: Effects of linguistic versus perceptual challenge.

Older adults with good hearing and with mild-to-moderate hearing loss were tested for comprehension of spoken sentences that required perceptual effort (hearing speech at lower sound levels), and two degrees of cognitive load (sentences with simpler or more complex syntax). Although comprehension accuracy was equivalent for both participant groups and for young adults with good hearing, hearing loss was associated with longer response latencies to the correct comprehension judgments, especially for complex sentences heard at relatively low amplitudes. These findings demonstrate the need to take into account both sensory and cognitive demands of speech materials in older adults' language comprehension. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Strategies for processing unbounded dependencies: Lexical information and verb–argument assignment.

This article describes 2 experiments about verb–argument relations in sentence processing in which there is no ambiguity involving the subcategorization of the verb but in which the role that the argument serves is initially unclear. Specifically, a self-paced reading experiment and an eye-tracking experiment investigated the way in which readers form unbounded dependencies when the verb is looking for both a direct object and a clause and when the filler either could be the direct object or could form part of the clause. The results suggested that readers treated the filler as the verb's direct object and probably also considered the clausal analysis at the same time. The results are interpreted with respect to current accounts of parsing. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Idiom literality judgments in younger and older adults: Age-related effects in resolving semantic interference.

To study age effects in the resolution of idiomatic semantic ambiguity, we focus on decomposability, the extent to which a literal reading of an idiom's words shares meaning with its figurative interpretation. Younger and older adults judged whether decomposable and nondecomposable idioms and nonidioms had a literal interpretation. Older adults were slower at making literality judgments and more sensitive to conflicts between literal and figurative meanings. The results support claims of decompositional analysis of idioms during later processing stages and of obligatory activation of figurative meanings. They also lend support to research that has shown age-related effects in ambiguity resolution. (PsycINFO Database Record (c) 2011 APA, all rights reserved)