The present study explored the types of errors found in Google Translate (GT) Chinese-to-English translations and, based on those error types, proposes strategies for optimizing the performance of GT. Seven abstracts written in both Chinese and English from seven articles published in English Teaching and Learning in 2017 were selected as the materials. The researchers compared the GT translations to the English abstracts written by the original author(s) and analyzed the problems in the translations. The problematic translations consisted of grammatical errors (35%) and lexical errors (65%). Relatedly, we propose nine specific strategies to employ when writing Chinese abstracts to be translated into English using GT. According to the strategies, we suggest that users (1) avoid native language-specific expressions, (2) maintain the use of original English terminologies in composing Chinese abstracts, and (3) enhance logical relations and expressions for the discipline-specific discourse community. Further analyses revealed that 99% of the 69 identified problems in the GT translations of the seven abstracts could be avoided by using the proposed strategies. A conceptual framework for the collaboration between GT and GT users is proposed and pedagogical implications are discussed.

As information technology is fast developing, it brings great convenience to our everyday life. However, hackers may also access confidential data illegally and easily from computers over the Internet. Therefore, how to protect property rights against infringement is an essential issue. Digital watermarking is a method that adds personal information to an intellectual property to protect one's ownership rights. Should the intellectual property be disputed, the owner can retrieve the watermark and prove ownership rights. Based on the principles of visual cryptography and the law of large numbers, our study generates shares by comparing pseudorandomly selected value pairs during the processes of embedding and verifying the hidden watermark. The wavelet transformation coefficients of the LL 3 region are used as the sample population. Experimental results indicate that our method has good robustness against darkening, lightening, blurring, sharpening, noising, distortion, jitter, JPEG, and cropping attacks. There are 3 advantages to our method: 1) robustness is retained when the protected image suffers from attacks; 2) unexpanded shares are created to reduce the size of every share; 3) the embedded watermark is decoded by the human visual system during the verification process.

No abstract