Scholarlink

On the practical side, the chapter trains embeddings in R (using word2vec) on product reviews to: find nearest neighbors for brand names (Apple/Samsung), visualize terms with UMAP, cluster them via cosine distance, and do vector arithmetic (e.g., shifting “kindle” toward positive/negative attributes). It then shows transformer applications with the text package: abstractive summarization (T5) of reviews and question-answering (RoBERTa SQuAD2) to pull facts from review corpora. Throughout, it ties methods to marketing uses—brand association mapping, competitive intel, recommendation cues, and concise insight extraction—while noting limits: models pattern-match and can misinterpret meaning without human judgment.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Word Embeddings and Transformers for Marketing Text Analytics

  • 章节

  • 发表日期：

    2026年

查看全文
  • Daniel Dan,
  • Thomas Reutterer

摘要

On the practical side, the chapter trains embeddings in R (using word2vec) on product reviews to: find nearest neighbors for brand names (Apple/Samsung), visualize terms with UMAP, cluster them via cosine distance, and do vector arithmetic (e.g., shifting “kindle” toward positive/negative attributes). It then shows transformer applications with the text package: abstractive summarization (T5) of reviews and question-answering (RoBERTa SQuAD2) to pull facts from review corpora. Throughout, it ties methods to marketing uses—brand association mapping, competitive intel, recommendation cues, and concise insight extraction—while noting limits: models pattern-match and can misinterpret meaning without human judgment.