Classification of search queries

Starting point

• A market research company would like to classify a million search queries from Internet search engines
• The Machine Learning (ML) algorithm developed in the project is to be applicable to further search queries
• Classification categories as well as design and training of the algorithm by the mgm technology partners GmbH

Objectives

• Manual classification of a training sample of 8,000 search queries
• Documentation of the categories “industry” and “user intent“ – adapted to the specifics of the available search queries
• Automated classification of one million search queries by the developed ML algorithm

Course of action

• Technical definition of suitable categories
• Selection of a representative training set
• Manual classification by means of crowd workers
• Training the ML algorithm
• Automatic classification of industries with the help of newsgroups, forums, Wikidata and Wiktionary as ontologies
• Visualization for quality assurance and improvement
• Calculation of KPIs, accuracy, F-measure

Results and benefits

• Technical specification of the individual categories for industry and user intent
• Automated classification of one million search queries
• ML algorithm including parameters