運用GPT-3語言模型探究網路口碑與知覺價值對社群影響者聯名產品之影響
dc.contributor | 蔣旭政 | zh_TW |
dc.contributor | Chiang, Hsu-Cheng | en_US |
dc.contributor.author | 陳語彤 | zh_TW |
dc.contributor.author | Chen,Yu-Tong | en_US |
dc.date.accessioned | 2024-12-17T02:56:12Z | |
dc.date.available | 2024-04-15 | |
dc.date.issued | 2024 | |
dc.description.abstract | 心理學家馬斯洛的需求層次理論(Maslow's hierarchy of needs)中,提及生理需求為人類最基本的滿足需求;《尚書.洪範》曾列舉施政制度「八政」,將「食」列為第一,「貨」也就是商業列為第二。由此可鑒,無論自古往今來還是東西地域,民以食為天,飲食皆為人類維生基本需求。 聯名行銷發展至今,以多種形式與商業有著緊密的合作,其中運用名人資本與社群影響者合作模式,已然成為社群的行銷趨勢。過去使用問卷調查法對社群影響者和聯名商品進行研究已行之有年,但以大數據瞭解消費者所關切之內容面向卻付之闕如。因而,本研究選擇——韓勾ㄟ金針菇與Joeman兩位社群影響者,在匿名平台Dcard上的15175則留言中,以大數據取徑,並藉由詞頻與文字雲進行文本分析。同時,運用OpenAI公司開發之GPT-3語言模型,根據學者所提出之消費者行為價值理論,將文本分類為「功能價值」、「社會價值」、「條件價值」和「與聯名商品無關」四個類目。 最終的研究結果顯示,金針菇與全家所推出的聯名商品相較於Joeman和7-11合作的聯名商品,呈現更顯著的正面聲量。知覺價值分類結果中,發現兩位社群影響的出生背景與就職經驗,皆使消費者對聯名商品的期待與評價產生影響。此外,儘管社群影響者可能因為自身不良事件而造成負面影響,但與之合作的關鍵還是取決於其個人特質與目標受眾之間的契合度是否相符。整體而言,本研究通過留言瞭解消費者行為價值與網路口碑與情緒,提供品牌與企業對於聯名商品特色、價格策略和品牌形象等調整策略。 | zh_TW |
dc.description.abstract | Maslow's hierarchy of needs theory, physiological needs are regarded as the most fundamental for human satisfaction. "The Book of Documents - Hong Fan" enumerates the "Eight Principles of Governance." "Food" is seen as the most important, and " commodity " are ranked second, representing commerce. Thus, we discover that regardless of historical or East-West region, "Bread is the staff of life," diets are fundamental needs for human survival. Co-branding has evolved to incorporate various forms of collaboration with businesses, particularly through celebrity endorsements and partnerships with social media influencers, or"SMIs", which have become a prevailing marketing trend. While past research on SMIs and co-brand products has often relied on questionnaire methods, there has been a lack of understanding of consumer concerns through big data analytics. Therefore, this study focuses on two social media influencers, " Jin Jhen Gu, " and "Joeman," analyzing 15,175 comments on the anonymous platform "Dcard" using big data analytics and textual analysis techniques such as term frequency and word clouds. Additionally, leveraging OpenAI's GPT-3 language model, the text is classified into four categories based on consumer behavior value theory proposed by scholars: "functional value," "social value," "conditional value," and "unrelated to co-brand products." The final research results indicate that the co-branded product featuring Jin Jhen Gu and FamilyMart exhibits significantly a more positive sentiment compared to the co-branded product collaboration between Joeman and 7-Eleven. Within the perceived value classification results, it was found that the birth backgrounds and professional experience of the two SMIs influenced consumers' expectations and evaluations of the co-branded products. Furthermore, despite the potential negative impact from adverse events associated with SMIs, the key to collaboration still lies between their personal traits and the target audience's preferences. Overall, this study provides insights into consumer behavior value and online sentiment through comments, offering brands and businesses adjustment strategies for co-branded product features, pricing strategies, brand image. | en_US |
dc.description.sponsorship | 大眾傳播研究所 | zh_TW |
dc.identifier | 61088004I-44949 | |
dc.identifier.uri | https://etds.lib.ntnu.edu.tw/thesis/detail/0268ccc1a70d27701fec0f16bfd3a583/ | |
dc.identifier.uri | http://rportal.lib.ntnu.edu.tw/handle/20.500.12235/122125 | |
dc.language | 中文 | |
dc.subject | 社群影響者 | zh_TW |
dc.subject | 聯名商品 | zh_TW |
dc.subject | 知覺價值 | zh_TW |
dc.subject | 網路口碑 | zh_TW |
dc.subject | GPT-3 | zh_TW |
dc.subject | social media influencers(SMIs) | en_US |
dc.subject | co-branded products | en_US |
dc.subject | perceived value | en_US |
dc.subject | e-WOM | en_US |
dc.subject | GPT-3 | en_US |
dc.title | 運用GPT-3語言模型探究網路口碑與知覺價值對社群影響者聯名產品之影響 | zh_TW |
dc.title | Using GPT-3 Model To Explore The Impact of e-WOM and Perceived Value on Social Media Influencers' Co-Brand Products | en_US |
dc.type | 學術論文 |
Files
Original bundle
1 - 1 of 1
No Thumbnail Available
- Name:
- 202400044949-107287.pdf
- Size:
- 3.64 MB
- Format:
- Adobe Portable Document Format
- Description:
- 學術論文