Topic-specific Image Caption Generation

来源 :第十六届全国计算语言学学术会议暨第五届基于自然标注大数据的自然语言处理国际学术研讨会 | 被引量 : 0次 | 上传用户：jiaomengni

【摘要】

：

Recently,image caption which aims to generate a textual description for an image automatically has attracted researchers from various fields.Encouraging performance has been achieved by applying deep

【作者】

：

Chang Zhou Yuzhao Mao Xiao jie Wang

【机构】

：

School of Computer,Beijing University of Posts and Telecommunications,Beijing,China

【出处】

：

第十六届全国计算语言学学术会议暨第五届基于自然标注大数据的自然语言处理国际学术研讨会

【发表日期】

：

2017年7期

【关键词】

：

Image caption Topic model Encoder-decoder

下载到本地 , 更方便阅读

下载此文赞助VIP

声明 : 本文档内容版权归属内容提供方 , 如果您对本文有版权争议 , 可与客服联系进行内容授权或下架

论文部分内容阅读

　　Recently,image caption which aims to generate a textual description for an image automatically has attracted researchers from various fields.Encouraging performance has been achieved by applying deep neural networks.Most of these works aim at generating a single caption which may be incomprehensive,especially for complex images.This paper proposes a topic-specific multi-caption generator,which in-fer topics from image first and then generate a variety of topic-specific captions,each of which depicts the image from a particular topic.We per-form experiments on flickr8k,flickr30k and MSCOCO.The results show that the proposed model performs better than single-caption generator when generating topic-specific captions.The proposed model effectively generates diversity of captions under reasonable topics and they differ from each other in topic level.

其他文献

Collective Entity Linking on Relational Graph Model with Mentions

Given a source document with extracted mentions,entity linking callsfor map-ping the mention to an entity in reference knowledge base.Previous en-tity linking approaches mainly focus on generic statis

会议

Collective Entity LinkingEntity DisambiguationRelational Graph

Cost-aware Learning Rate for Neural Machine Translation

Neural Machine Translation(NMT)has drawn much attention due to its promising translation performance in recent years.The conventional optimiza-tion algorithm for NMT sets a unified learning rate for e

会议

Neural Machine TranslationCost-aware Learning Rate

Improving Word Embeddings for Low Frequency Words by Pseudo Contexts

This paper investigates relations between word semantic den-sity and word frequency.A distributed representations based word av-erage similarity is defined as the measure of word semantic density.We f

会议

Word EmbeddingLow Freuqcy Word

Question Answering with Character-Level LSTM Encoders and Model-Based Data Augmentation

This paper presents a character-level encoder-decoder mod-eling method for question answering(QA)from large-scale knowledge bases(KB).This method improves the existing approach [9] from three aspects.

会议

Question AnsweringKnowledge BaseLong Short-TermMemoryEncoder-Decoder

A pipelined Pre-training algorithm for DBNs

Deep networks have been widely used in many domains in recentyears.However,the pre-training of deep networks is time consuming with greedy layer-wise algorithm,and the scalability of this algorithm is

会议

componentdeep networkspre-traininggreedy layer-wiseRBMpipelined

Natural Logic Inference for Emotion Detection

Current research on emotion detection focuses on the recognizingexplicit emotion expressions in text.In this paper,we propose an approach based on textual inference to detect implicit emotion expressi

会议

Natural LogicTextual InferenceEmotion DetectionImplicit Emotional Expression

Semantic Dependency Labeling of Chinese Noun Phrases Based on Semantic Lexicon

We have presented a simple algorithm to noun phrases interpretation based on hand-crafted knowledge-base containing detailed semantic information.The main idea is to define a set of relations that can

会议

Noun relationsSemantic dependencyNoun phrases

Conceptual Multi-Layer Neural Network Model for Headline Generation

Neural attention-based models have been widely used recently in head-line generation by mapping source document to target headline.However,the traditional neural headline generation models utilize the

会议

Attention-basedConceptMulti-layer Bi-LSTM

End-to-End Neural Text Classification for Tibetan

As a minority language,Tibetan has received relatively little atten-tion in the field of natural language processing(NLP),especially in current var-ious neural network models.In this paper,we investig

会议

Arabic Collocation Extraction Based on Hybrid Methods

Collocation Extraction plays an important role in machine transla-tion,information retrieval,secondary language learning,etc.,and has obtained significant achievements in other languages,e.g.English a

会议

Arabic collocation extractiondependency relationhybrid method

Topic-specific Image Caption Generation

其他学术论文