基于深度学习的中文专利自动分类方法研究

[目的/意义] 面向当前国内专利审查和专利情报分析工作中对于海量专利分类的客观需求，设计了7种基于深度学习的专利自动分类方法，对比各种方法的分类效果，从而助力专利分类效率和效果的提升。[方法/过程] 针对传统机器学习方法存在的缺陷，基于Word2Vec、CNN、RNN、Attention机制等深度学习技术，考虑专利文本语序特征、上下文特征以及分类关键特征，设计Word2Vec+TextCNN、Word2Vec+GRU、Word2Vec+BiGRU、Word2Vec+BiGRU+TextCNN等7种深度学习模型，以中国专利为例，选取IPC主分类号的"部"作为分类依据，对比这7种模型与3种传统分类模型在中文专利分类任务中的效果。[结果/结论] 实证研究效果显示，采用考虑语序特征、上下文特征及强化关键特征的深度学习方法进行中文专利分类具有更优的分类效果。     

网络口语语体初探

根据网络语言运用的现状,描述分析网络言语交际中分化出的一种新的口语形式,归纳其语言材料选用特点,以及人们使用这种口语的心理动因,表明对不断出现的语言新现象应持的态度.     

试论英语中双关特色及其修辞作用

本文提出了双关语的两种不同使用方法,意在强调,我们不仅要理解双关的内在含义,还要学会领悟其奥妙所在.     

科技期刊中数字“1”和“一”及“2”和“两”“二”的用法

使用阿拉伯数字具有如一优点:1)它是世界通用的数字,科技书刊中使用它便于国际交流;2)科技工作者整天与数字打交道,对于阿拉伯数字的视觉反应最为敏感;3)笔画简单,组数简短,便于计算机录入.     

用户自然和社会属性对网络搜索中语言使用行为的影响

以用户的自然、社会属性对用户在搜索中使用检索语言的影响为研究对象,旨在对影响用户行为的因素作探索性研究。综合网络调查问卷的分析结果和用户参与对比实验法,得出性别、年龄、学历和专业教育对用户语言使用影响的结论。对于此问题的研究,有助于认清搜索过程的影响因素,进而可以以此建模,改善搜索引擎的服务质量。     

政治上正确的语言

“政治上正确的”由一政治术语演变为语言文化语汇,有其积极的意义,它是民权运动和多元文化的产物。政治上正确的语言也是复杂的,其使用受到言论自由和时代的影响,滥用之也会造成很多问题。     

Employ Game Teaching Method in the Primary School English Classroom

on the basis of linguistics, psychology and other related theory, we should carry out game teaching method to make students learn easily and happily because of the lack of interest and other problems i...     

Usage Comparison between an English-Language and a French-Language E-Monograph Collection in a Bilingual Canadian University

Abstract

This article reports the results of a quantitative and systematic analysis comparing usage rates between English-language and French-language e-monograph titles at Laurentian University, Canada. Full-text viewings were collected as usage metrics. English-language e-monographs exhibited medium strength linear relationships between collection size and usage. On the other hand, French-language e-monographs exhibited weak linear relationships between collection size and usage. The French-language e-monograph collection experienced a higher rate of full-text use, relative to size. The cost-per-viewing ratio for the English-language e-monograph collection was higher than those that calculated for the French-language collection.     

How We Used to Build the Future: 30 Years of Collection Development Trends

This session highlighted trends in continuing resource collections and collection management throughout the 30-year period from 1985 through 2015. Electronic resources became available in the 1990s and grew in prevalence from the 2000s through the present, changing the make-up of library collections and also how libraries measure collections and collection usage. Two librarians and two vendors shared their perspectives on these changes in collection development, describing how library collections now contain a combination of older formats and older ways of collecting data, and new, electronic content and automated ways of collecting and disseminating data about collections.