Issue link: https://epubs.iltanet.org/i/13683
document) that represents an important concept in the case. The search engine analyzes this text and returns documents that contain similar concepts. The search results are accompanied by a score that tells the user how closely the returned documents relate to the search. For example, a reviewer is searching for information about the recall of a product. She runs a concept search by selecting a paragraph from a document in the case that discusses the recall. The search retrieves conceptually related documents, even if they don’t specifically contain the product name. Some concept engines will allow a reviewer to submit an entire document as a search query. Thus, when a key document is found, that document can be used to search the database and return all conceptually similar documents. This can be useful in prioritizing documents for review and leveraging previous work product. As an example, a reviewer completes a review of one custodian, and then she moves to another custodian. She can select several important documents from the previous custodian’s collection and use them as searches to prioritize her review of the next custodian. This approach is comparable to using a GPS instead of a compass; it gets reviewers where they need to go faster and with the smallest number of detours. Finally, through keyword analysis, concept search engines can accelerate a case team’s understanding of the facts of a case. For example, if a reviewer wants 26 Case/Matter Management ILTA White Paper “By using text analytics to structure a review, litigation professionals can significantly reduce discovery costs by increasing document review speeds.” to know more about the keyword “apple,” the concept engine can provide a list of closely related keywords from the keyword index. A reviewer might expect to see logical results from this search that relate closely to her idea of an apple, just as she would if her search were submitted to Google or Wikipedia. However, it is important to understand that the concept engine understands the term “apple” only in the context of her review database. The search engine does not reach out to external sources such as dictionaries or thesauri to learn about “apple.” It will return the terms that are specific to her document collection and that are closely related to “apple.” The value is that the reviewer gets a much better understanding of how the keyword “apple” was used in her matter, and what other terms should be evaluated, including terms of art. The reviewer can then add these related keywords to subsequent keyword searches, which increases the likelihood of locating responsive documents. This type of information, once gleaned only after many hours of rigorous document review, is now available immediately. CLASSIFICATION In addition to concept search, classification can reduce review costs by organizing electronic documents into conceptually related subsets. Classification can be done with or without user input. Automatic classification, also called “clustering,” groups documents together based on shared