GrammAds: Keyword and Ad Creative Generator for Online Advertising Campaigns Author : Stamatina...

GrammAds: Keyword and Ad Creative Generator for Online Advertising CampaignsAuthor : Stamatina Thomaidou, Konstantinos Leymonis, and Michalis VazirgiannisPresenter: Wayne Yang

Agenda

Introduction

Related Work

Keyword Generation

Ad Creative Generation

Campaign Organization and Use Cases

System Architecture and Communication

Experimental Evaluation

Conclusion and future work

Introduction - Baseline Campaign Creation

Process

ProductA landing web

page(price, buy)

Finding keyword

Advertising

text

Introduction - Motivation for an

Automated ApplicationAn advertiser decides

Usually choose specific 1-3 keywords per product

Short headline and ad text

maximum cost-per-click of each keyword

=>effective keyword selection is one of the most important success factors for online advertising

Related Work

Google Adwords - They start from an ini- tial key phrase and they are based on past queries that contain these search terms.

Other commercial tools - determine an advertiser’s top competitors and then actively search for the keywords they are targeting.

TermsNet and Wordy - With their methods they find other semantically similar terms.

Keyword Generation

Keyword Extraction

Landing page parsed => stop words removed => trigram, bigrams, unigram => calculate relevance keyword

Keyword Suggestion

Seed keywords => top 30 result pages => corpus => lucene score

Ad Creative Generation

generic summary - shows the main topics and key contents covered in the source text

query-relevant summary - locates the contents pertinent to user’s seeking goals(our system provided)

extract text => Bayesian classifier to summarize

Campaign Organization and Use Cases

Step 1 – initial settings

1.no optimization - where the system just uploads automatically the generated keywords, ad-texts, and bids along with their organized structure without continuing to be responsible for an automated optimization strategy

2.traffic optimization - where the advertiser considers the profit to be the amount of clicks at the ad-texts

3.profit optimization - where the profit is the actual monetary profit from offline product sales or online conversions to a specific landing page that is defined in a next step

Step 2 – Crawler module

Step3 – select keyword

Campaign Organization and Use Cases

Step 4 - Automatically generated keywords and recommended bids

Step 5 - the user can see through the Google AdWords interface the uploaded settings of his constructed campaign.

System Architecture and Communication

The proposed system structure resulted as a need for an appropriate user inter- face interaction of the Adomaton subsystem, which optimizes Google AdWords Campaigns through a novel bidding strategy.

Adopt model view controller for information managment

System Architecture and Communication

Experimental Evaluation

Conclusion and future work

Automating the task of finding the appropriate keywords

Recommending multiword terms (n-grams) with high specificity without the need to capitalize on usage data such as query and web traffic logs

Generating fast snippets of ad texts

An automated and fast way of uploading campaign and AdGroup settings (keywords and ad-texts per AdGroup) into the AdWords service

A developed web application with an initial experimentation on real world campaigns