Bilingual Alignment Of Emotion Words

With the rise of “affective computing”,people became more aware of the importance of subjective texts and attempted to use computers to automatically analyze the emotions expressed in these texts.This has given birth to a new,cross-disciplinary field of research,namely emotional analysis and opinion mining.The former focuses on the positive and negative polarity expressed in the text,while the latter pays more attention to people's opinions or comments on entities,such as products,services,and institutions.In the field of bilingual emotional analysis and opinion mining,the emotion dictionary is a very important tool serving as a translation aid.The most widely-used emotion dictionaries are those collecting commendatory and/or derogatory entries.Up to now,lexicographers at home and abroad have compiled a series of such dictionaries.For example,Chinese emotional dictionaries include the NTUSD(National Taiwan University Semantic Dictionary),the emotional word dictionary constructed by Tsinghua University and so on.English ones include WordNet,SentiWordNet,LIWC(Linguistic Inquiry and Word Count),ANEW(Affective norms for English words),and MPQA(Multi-Perspective Question Answering).However,these dictionaries are fundamentally flawed in that all the entries listed in the dictionaries tend to express the polarity or emotion tendency in a fixed way,although the polarity of these entries is likely to shift in different contexts.In addition,most of these dictionaries are monolingual,and there is still a lack of bilingual emotional dictionaries on the market.Therefore,it becomes a hot topic for linguists,lexicographers and computer scientists to compile a bilingual emotional lexicon for multilingual emotional analysis and opinion mining.From the perspective of lexicography and translation studies,the core issue of compiling bilingual dictionaries is the search for equivalent words between two languages,namely translating a Chinese entry into its equivalent English glossaries.Taking adjectives for example,the paper ventures to study the translation of Chineseadjectives into their equivalent English ones,with the focus laid on the justification of the English adjective as the optimum equivalent word for the Chinese adjective under discussion.The theoretical framework adopted in this paper is the Attribute-Value structure(namely AVS)in cognitive linguistic.“A” represents the attributes of an entity,such as a product;“V” represents the value of these features of the product,which is often realized by adjectives.On the one hand,AVS has been regarded as a basic framework to represent the meaning of adjectives;on the other hand,“entity” and “attributes” that are connected with “value” serve as the minimal context for the occurrence of the given adjective,and play an important role in the specification of the meaning of the adjective.The corpora in this paper were customers'comments on Taobao and Amazon of United States.Procedurally,we first specified the product as a mobile phone.Then,we mined the relevant comments on Taobao to obtain a Chinese AVS including the attributes and values of the phone,such as“??-?”.Meanwhile,we obtained an English AVS through the mining on the Amazon such as“memory-big”.Later,we aligned Chinese and English AVS structures to discover that“big”is the most equivalent word for“?”in this domain.Experiments showed that the accuracy of data mining and AVS alignment is 80%,which verified the validity of the method.The creativity of the paper lies in two aspects.Theoretically,the adoption of AVS as the theoretical framework was followed by an illustration from the perspective of lexicography That is to say,the AVS is actually a minimal context for the occurrence of adjectives which express the people's comments of a product.Practically,we adopted Taobao and Amazon as two comparable corpora since there is no free parallel corpora on the market up to now.The above research shed light on the compiling of bilingual lexicography.Generally speaking,a common approach to compile a bilingual dictionary is to translate an entry into several equivalent words in another language without further hints of the best equivalent one among them.In this paper,however,we can find the optimum equivalent word using the method discussed above,therefore eliminating the number ofcandidate equivalent words.Furthermore,the comparable corpora and their AVS-based mining help construct a query platform for customers in different countries about the evaluation of related products before deciding whether or not to purchase the product.This research is undoubtedly consistent with the current “One Belt and One Road”.There are some shortcomings in this study,which needs further exploration.First,this research is limited to the corpus-based alignment of adjectives.Whether the approach is suitable for the research of other word classes remains to be discussed.Secondly,the corpora adopted were Taobao and Amazon.Whether there are more suitable comparable corpora remains to explore.