Improving Matching Process with Expanding and Classifying Criterial Keywords leveraging Word Embedding and Hierarchical
- PDF / 673,272 Bytes
- 12 Pages / 439.37 x 666.142 pts Page_size
- 99 Downloads / 174 Views
Improving Matching Process with Expanding and Classifying Criterial Keywords leveraging Word Embedding and Hierarchical Clustering Methods Yutaka Iwakami1 · Hironori Takuma2 · Motoi Iwashita3 Received: 14 May 2020 / Accepted: 26 August 2020 / Published online: 14 September 2020 © Springer Japan KK, part of Springer Nature 2020
Abstract Matching processes, such as the selection of producers of advertising content cor‑ responding to specific products or the screening of job applicants based on prede‑ fined requirements, have become important operations required by enterprises. Such problems generally include several keywords representing the matching criteria, but it is difficult for enterprises to expand and classify criterial keywords properly to improve the matching performance. This study proposes solutions to this issue by extracting criterial keywords from social networking services (SNSs) based on word embedding and by classifying the obtained keywords via hierarchical cluster‑ ing. This approach will enable enterprises to gather and prioritize criterial keywords more accurately to improve their matching processes. Keywords Matching process · Word2Vec · Hierarchical clustering · NLP · SNS · Semantic analysis
1 Introduction Over the recent years, matching processes have become an important component of commercial success. For example, it is vital for enterprises to advertise on the web to increase public recognition of their products and services. Product recommenda‑ tions or introductions by individual content creators have also garnered attention of late. However, the number of available content creators is very high. It is, therefore, * Yutaka Iwakami [email protected] 1
Department of Research & Analysis for IT Industry, Nork Research Co., Ltd, 2‑13‑10 Shinjuku, Shinjuku‑Ku, Tokyo 160‑0022, Japan
2
Department of Project Management, Chiba Institute of Technology, 2‑17‑1 Tsudanuma Narashino, Chiba 275‑0016, Japan
3
Department of Management Information Science, Chiba Institute of Technology, 2‑17‑1 Tsudanuma Narashino, Chiba 275‑0016, Japan
13
Vol.:(0123456789)
194
The Review of Socionetwork Strategies (2020) 14:193–204
critical for enterprises to identify the content producers that are most suitable to adver‑ tise their products and services. Several techniques have been proposed to resolve this issue [1–3]. On the other hand, employment is an essential requirement for enterprises. Owing to the rapid digitalization of society, job descriptions are becoming progres‑ sively more diverse and complicated making recruitment difficult to optimize. This also comprises a topic of active research [4, 5]. Matching of inquiries with solutions in cus‑ tomer support is yet another challenging matching problem [6]. In the aforementioned processes, enterprises usually provide a set of keywords as criteria for their requirements. For example, if a beer-brewer considers that “stout” well represents a characteristic of their product, they would wish to include the word as a criterion for selecting produce
Data Loading...
