favorite1We encode a target numeral in a tweet with character- and word-based schemes with the matrix composed of three parts as shown in Fig. The structure of the proposed CNN model consists of one convolutional layer, one max pooling layer, one denselyconnected layer with 64 hidden dimensions, one dropout layer with 0.5 dropout rate, one rectified linear unit (ReLU) layer, and the softmax output layer..
favorite11DATA ANNOTATION We extract 707 unique tweets containing numerals from the dataset of SemEval-2017 Task5 , which is collected from Twitter and StockTwits (a popular social media platform for investors to share their ideas and strategies).
favorite1In this paper, we classify the numerals in financial tweets into 7 categories, including Monetary, Percentage, Option, Indicator, Temporal, Quantity, and Product/Version Number.
favorite7Numerals play important roles in financial analysis processes such as determining the intrinsic values of financial instruments and forecasting the movement of asset prices based on the past market data.
favorite14This work is the first attempt to understand numerals in financial social media data, and we provide the first comparison of fine-grained opinion of individual investors and analysts based on their forecast price.
favorite0This paper is aimed at understanding the meanings of numerals in financial tweets for fine-grained crowd-based forecasting.
favorite28A translation-based language modeling framework on top of the proposed word embedding method, which can also be integrated with classic word embedding methods, is introduced to the extractive speech summarization task..
favorite11Beyond the efforts to improve the representation of words, we also present a novel and efficient translation-based language modeling framework on top of the proposed word embedding method for extractive speech summarization.
favorite0Orthogonal to the existing commonly-used methods, we explore in this paper the use of various word embedding methods [9-11] in extractive speech summarization, which have recently demonstrated excellent performance in many natural language processing (NLP) related tasks, such as machine translation , sentiment analysis  and sentence completion .
favorite6Built upon the proposed word embedding method, we further formulate a translation-based language modeling framework for the extractive speech summarization task.
favorite17Word embedding methods revolve around learning continuous distributed vector representations of words with neural networks, which can capture semantic and/or syntactic cues, and in turn be used to induce similarity measures among words, sentences and documents in context.
favorite13In the next set of experiments, we evaluate the capability of the triplet learning model for improving the measurement of similarity when applying word embedding methods in speech summarization.
favorite4The Triplet Learning Model Inspired by the vector space model (VSM), a straightforward way to leverage the word embedding methods for extractive SDS is to represent a sentence Si (and a document D to be summarized) by averaging the vector representations of words occurring in the sentence Si (and the document D) [23, 25]:.
favorite12The central idea of these methods is to learn continuously distributed vector representations of words using neural networks, which can probe latent semantic and/or syntactic cues that can in turn be used to induce similarity measures among words, sentences, and documents.
favorite0Beyond the continued efforts made to improve the representation of words, this paper focuses on building novel and efficient ranking models based on the general word embedding methods for extractive speech summarization.