Detecting Content-Heavy Sentences: A Cross-Language Case Study

Abstract

The information conveyed by some sentences would be more easily understood by a reader if it were expressed in multiple sentences. We call such sentences content heavy: these are possibly grammatical but difficult to comprehend, cumbersome sentences. In this paper we introduce the task of detecting content-heavy sentences in cross-lingual context. Specifically we develop methods to identify sentences in Chinese for which English speakers would prefer translations consisting of more than one sentence. We base our analysis and definitions on evidence from multiple human translations and reader preferences on flow and understandability. We show that machine translation quality when translating content heavy sentences is markedly worse than overall quality and that this type of sentence are fairly common in Chinese news. We demonstrate that sentence length and punctuation usage in Chinese are not sufficient clues for accurately detecting heavy sentences and present a richer classification model that accurately identifies these sentences.

Does splitting make sentence easier?

“…We further rearranged component texts so that the flow of events they depict comes in line with that of the complex sentence. We emphasize that in contrast to Li and Nenkova ( 2015 ), this study is not about identifying conditions under which people favor a split sentence.…”

Estimating Text Intelligibility via Information Packaging Analysis

“…Failure in doing so may result in a translated sentence being hard to process for speakers in the target language. To understand whether this phenomenon is important enough for human and system translators to consider, I showed that sentences need to be translated into multiple English sentences cause significant quality drop for MT, while the number of words in them has little correlation with MT quality (Li, Carpuat, and Nenkova 2014;Li and Nenkova 2015a). From a translator's point of view, more than 15% of the sentences were translated into multiple sentences in English by at least three out of four human translators.…”

“…To identify sentences that need a multi-sentence translation, I designed a system that achieves more than 80% accuracy (Li and Nenkova 2015a). Next I plan to explore methods to improve the flow of MT outputs for these sentences.…”

The Case for Claim Difficulty Assessment in Automatic Fact Checking

“…While claim difficulty prediction appears novel in the context of fact-checking, we note that the value of modeling and predicting difficulty of different task instances is already recognized in other areas, such as machine translation [14,13], syntactic parsing [8], or search engine switching behaviors [29], among others. From this perspective, we argue instance difficulty prediction can bring similar value to fact-checking.…”