Coreference resolution is the problem of determining which mention in a text refer to the same entities, and is a crucial and difficult step in every natural language processing task. Despite the efforts that have been made in the past to solve this problem, its performance still does not meet today’s applications requirements. Given the importance of the verbs in sentences, in this work we tried to incorporate three types of their information on coreference resolution problem, namely, selectional restriction of verbs on their arguments, semantic relation between verb pairs, and the truth that arguments of a verb cannot be coreferent of each other. As a needed resource for supporting our model, we generate a repository of semantic relations between verb pairs automatically using Distributional Memory (DM), a state-of-the-art framework for distributional semantics. This resource consists of pairs of verbs associated with their probable arguments, their role mapping, and significance scores based on our measures. Our proposed model for coreference resolution encodes verbs’ knowledge with Markov logic network rules on top of deterministic Stanford coreference resolution system. Experiment results show that this semantic layer can improve the recall of the Stanford system while preserves its precision and improves it slightly. Manuscript profile

The prior relation extraction approaches were relation-specific and supervised, yielding new instances of relations known a priori. While effective, this model is not applicable in case when the number of relations is high or where the relations are not known a priori. Open Information Extraction (OIE) is a relation-independent extraction paradigm designed to extract relations directly from massive and heterogeneous corpora such as Web. One of the main challenges for an Open IE system is estimating the probability that its extracted relation is correct. A confidence measure shows that how an extracted relation is a correct instance of a relation among entities. This paper proposes a new method of confidence estimation for OIE called Relation Confidence Estimator for Open Information Extraction (RCE-OIE). It investigates the incorporation of some proposed features in assigning confidence metric using logistic regression. These features consider diverse lexical, syntactic and semantic knowledge and also some extraction properties such as number of distinct documents from which extractions are drawn, number of relation arguments and their types. We implemented proposed confidence measure on the Open IE systems’ extractions and examined how it affects the performance of results. Evaluations show that incorporation of designed features is promising and the accuracy of our method is higher than the base methods while keeping almost the same performance as them. We also demonstrate how semantic information such as coherence measures can be used in feature-based confidence estimation of Open Relation Extraction (ORE) to further improve the performance. Manuscript profile

در این مقاله اولین سیستم تبدیل متن به گفتار پیاده‌سازی شده برای زبان کردی معرفی شده است. زبان کردی دارای دو رسم‌الخط رایج بر اساس الفبای عربی و لاتین است. در قسمت تحلیل متن، علاوه بر رفع ابهامات رایج در متون مختلف، مشکلات مربوط به هر دو رسم‌الخط بررسی شده است. همچنین نمادهای استانداردی تعریف شده‌اند كه سيستم قادر است متن ورودي به هر يك از رسم‌الخط‌هاي فوق را به رشته‌اي از نمادهاي استاندارد تبديل نمايد. همچنین منحنی‌های تغییرات گام برای انواع مختلف جمله‌ها در این زبان برای اولین بار بررسی شده است. در قسمت تولید گفتار، سه سیستم مختلف بر اساس واحدهای واج‌گونه، هجا و دایفون پیاده‌سازی شده است. برای بررسی کیفیت این سیستم‌ها و مقایسه آنها با همدیگر از چهار آزمون MOS، قابلیت فهم، DRT و MRT استفاده شده است. نتایج این آزمون‌ها نشان‌دهنده قابلیت فهم بالای این سیستم‌ها و به‌ویژه سیستم مبتنی بر دایفون است. Manuscript profile