Discriminability objective for training descriptive captions

نویسندگان

  • Ruotian Luo
  • Brian Price
  • Scott Cohen
  • Gregory Shakhnarovich
چکیده

•ATTN models better than FC models, and discriminability objective works for both. •ATTN+CIDEr+* combination is our best choice •Moderate λ = 1 produces good tradeoff between discriminability and fluency •Higher λ make captions more discriminative to machine and to humans, but at the cost of fluency •With moderate λ, non-discriminative scores like BLEU, METEOR, CIDEr improve as well! • especially surprising result: CIDEr (ostensibly we focus less on maximizing it during training.)

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

SentiCap: Generating Image Descriptions with Sentiments

The recent progress on image recognition and language modeling is making automatic description of image content a reality. However, stylized, non-factual aspects of the written description are missing from the current systems. One such style is descriptions with emotions, which is commonplace in everyday communication, and influences decision-making and interpersonal relationships. We design a ...

متن کامل

On the generality of optimal versus objective classifier feedback effects on decision criterion learning in perceptual categorization.

Biased category payoff matrices engender separate reward- and accuracy-maximizing decision criteria Although instructed to maximize reward, observers use suboptimal decision criteria that place greater emphasis on accuracy than is optimal. In this study, objective classifier feedback (the objectively correct response) was compared with optimal classifier feedback (the optimal classifier's respo...

متن کامل

Generating Image Descriptions using Multilingual Data

In this paper we explore several neural network architectures for the WMT 2017 multimodal translation sub-task on multilingual image caption generation. The goal of the task is to generate image captions in German, using a training corpus of images with captions in both English and German. We explore several models which attempt to generate captions for both languages, ignoring the English outp...

متن کامل

Laughter extracted from television closed captions as speech recognizer training data

Closed captions in television broadcasts, intended to aid the hearing impaired, also have potential as training data for speech-recognition software. Use of closed captions for automatic extraction of virtually unlimited training data has already been demonstrated [1]. This paper reports some preliminary work on the use of non-speech sound tokens included in closed captions to extract training ...

متن کامل

Effect of Amnesia Mild Cognitive Impairment and Alzheimer’s Diseaseon Recognition Memory of Elderly Peoplein Shiraz Verbal Learning Test: Differences in Recognition Discriminability and Response Bias

Background: Most studies have investigated the effect of brain pathological aging on information recall and recognition memory performance of patients (by using a yes/no procedure), and for this reason, provide a partial picture of memory deficits and other factors involved in recognition memory such as discriminability and response bias are not considered. In this regard, the aim of present st...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2018