類神經網路在行銷主軸與產品文案應用

Abstract

一個好的行銷文案計劃需要許多專業的角色共同完成，創意的策略及設計通常需要長時間生活的內化， 所以文案沒有對與錯而是能不能勾動閱讀者的感受。 對於管理者的難題，常常也在於無法量化或有快速有效的方式獲得提案，產品的週期往往也會在反覆的討論與無法判斷選擇提案的可行性中讓產品時程有所滯延。 2018科技話題圍繞著人工智慧AI (Artificial Intelligence) ，AI提供世界許多解決方案，用AI來生成行銷語言，使用自然語言處理和生成比人力撰稿花費的時間短，也有可快速量化、生成和優化的優勢。 本研究著重探討商業行為裡行銷活動中的文案是否可利用深度學習與自然語言處理來生成文案之可行性的評估與相關實驗，以期利用該類系統應用於文案生成的結果來有效減少傳統人力作業模式的資源耗損，藉由增加效率來解決管理上的問題。Marketing related activities have got more and more importance in the business world nowadays, and to write a decent copywriting gets even more considerable among all kinds of work related to digital marketing. There is no good or bad copywriting, the only thing matters is: Can it resonate the most with target audience? Inspired by sophisticated AI provides all kind of solutions and both of a very trendy AI application, NLP and highly appreciation of creating touching copywriting when running a business, the theme of this thesis is to study whether AI can automatically generate useful copywriting by training a designated RNN model with a large amount of existing copywriting data and then having a business category chosen as a keyword to input. This kind of technology could significantly reduce human effort of manually constructing slogans or promoting sentences. In addtion, efficiency brought by such application would help business managers to avoid back and forth discussion which could delay projects and not able to compete in current rapid business environment.