Cognitive Search

Cognitive search applies Artificial Intelligence (AI)/Machine Learning ( ML) on user’s search term to contextual document search. IntelliService leverages trained bi-directional, transformer-based ML model along with unsupervised learning to improve the context specific information. The built-in low/no code connector framework provides the ability to quickly connect to many data sources and store knowledge articles such as product FAQs, service bulletins, troubleshooting guides, and user care manuals into a centralized data repository for automated indexing and content processing. The sources can be CRM, ticketing system, SharePoint site, knowledge base, web pages or shared folders.

Configure Source

Before you perform search, make sure to configure the document source. You can configure the data source through one of the following ways:

Uploading documents
Web-crawling

Upload Documents

IntelliService allows direct upload of the unstructured documents like PDF, word and text files, from your local folder or from the SharePoint site. In the project, click ADD > Documents, selectother documents and upload the required documents either by dragging and dropping the files or by browsing the content and upload them. On successful upload of document, the documents are ready for search.

Within a search service, the Synonyms and Acronyms are resources that associate equivalent terms without the user having to actually provide the term. In the Project page, on the Synonyms and Acronyms tab, click Add Synonym/Acronym, and enter the word and its corresponding Synonym/Acronym.

Web-crawler

Web crawler downloads the content of website and performs the context search on the downloaded content. In the Setup Crawler page under Settings, add the base URL and keywords either to exclude or include the content from mining the website. Also, you have the option to download the embedded documents available in the website. The embedded documents can be PDF, excel, word and text files. Click Start Crawling to crawl all available urls available in the base url. From the list of crawled URLs, you can either select all the urls or specific urls and click Download Data. In the Downloaded Data page, the downloaded urls will be listed along with the status.

Note: The web-crawling will be limited to download the content of up to 500 urls and index the data.

Figure 1. Sample screenshot of web-crawler mining page for google cloud products

Search Documents

In the cognitive search page, you can enter the queries to be searched in the uploaded unstructed documents and the crawled website, and the responses will be listed based on the relevancy of the context being searched along with the hyperlink to the source data.