How Much Information Does a Human Translator Add to the Original?

Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 889–898, Lisbon, Portugal, 17-21 September 2015. c⃝2015 Association for Computational Linguistics. How Much Information Does a Human Translator Add to the Original? Barret Zoph, Marjan Ghazvininejad and Kevin Knight Information Sciences Institute Department of Computer Science University of Southern California {zoph,ghazvini,knight}@isi.edu Abstract We ask how much information a human translator adds to an original text, and we provide a bound. We address this ques- tion in the context of bilingual text com- pression: given a source text, how many bits of additional information are required to specify the target text produced by a hu- man translator? We develop new compres- sion algorithms and establish a benchmark task. 1 Introduction Text compression exploits redundancy in human language to store documents compactly, and trans- mit them quickly. It is natural to think about com- pressing bilingual texts, which have even more re- dundancy: “From an information theoretic point of view, accurately translated copies of the original text would be expected to con- tain almost no extra information if the original text is available, so in princi- ple it should be possible to store and transmit these texts with very little ex- tra cost.” (Nevill and Bell, 1992) Of course, if we look at actual translation data (Figure 1), we see that there is quite a bit of unpre- dictability. But the intuition is sound. If there were a million equally-likely translations of a short sen- tence, it would only take us log2(1m) = 20 bits to specify which one. By finding and exploiting patterns in bilingual data, we want to provide an upper bound for this question: How much information does a human translator add to the original? We do this in the context of building a practical compressor for bilingual text. 上个星期的战斗至少夺取12个人的生命。 At least 12 people were killed in the battle last week. Last week’s fight took at least 12 lives. The fighting last week killed at least 12. The battle of last week killed at least 12 persons. At least 12 people lost their lives in last week’s fighting. At least 12 persons died in the fighting last week. At least 12 died in the battle last week. At least 12 people were killed in the fighting last week. During last week’s fighting, at least 12 people died. Last week at least twelve people died in the fighting. Last week’s fighting took the lives of twelve people. Figure 1: Eleven human translations of the same source sentence (LDC2002T01). We adopt the same scheme used in mono- lingual text compression benchmark evaluations, such as the Hutter Prize (Hutter, 2006), a com- petition to compress a 100m-word extract of En- glish Wikipedia. A valid entry is an executable, or self-extracting archive, that prints out Wikipedia, byte-for-byte. Decompression code, dictionaries, and/or other resources must be embedded in the executable—we cannot assume that the recipient of the compressed file has access to those re- sources. This view of compression goes by …

Content truncated to 500 words