Concept searching

You can use Concept searching to find information without a precisely phrased query by applying a block of text against the database to find documents that have similar conceptual content. This can help prioritize or find important documents.

Concept searching is very different from keyword or metadata search. A concept search performed in Relativity Analytics reveals conceptual matches between the query and the document quickly and efficiently, so you can focus on the concepts that you deem important. The following table illustrates the differences between standard searching and concept searching.

Standard Method Analytics Method

Finds the presence (or absence) of a query (term or block of text)

Derives the potential meaning of a query

Simply looks for matches of query and indexed docs

Attempts to understand semantic meaning and context of terms

Incorporates Boolean logic

Incorporates mathematics

With standard Keyword Search, people frequently experience “term mismatch,” where they use different terms to describe the same thing.

  • “Phillipines” and “Philippines”—misspelling
  • “Jive” and “jibe”—misuse of the word
  • “Student” and “pupil”—synonyms
  • “Pop” and “soda”— regional variation

Using concept searching, you can submit a query of any size and return documents that contain the concept the query expresses. The match isn't based on any one specific term in the query or the document. The query and document may share terms, or they may not, but they share conceptual meaning.

Every term in a conceptual index has a position vector in the concept space. Every searchable document also has a vector in the concept space. These vectors, when close together, share a correlation or conceptual relationship. Increased distance indicates a decrease in correlation or shared conceptuality. Two items that are close together share conceptuality, regardless of any specific shared terms.

During concept searching, you create text that illustrates a single concept (called the concept query), and then submit it to the index for temporary mapping into the concept space. The conceptual analytics index uses the same mapping logic to position the query into the concept space as with the searchable documents.

Once the position of the query is established, the Analytics index locates documents that are close to it and returns those as conceptual matches. The document that is closest to the query is returned with the highest conceptual score. This indicates distance from the query, not percentage of relevancy—a higher score means the document is closer to the query, thus it is more conceptually related.

You can use concept searches in conjunction with keyword searches. Since a keyword can have multiple meanings, you can use a concept search to limit keyword search results by returning only documents that contain the keyword used in similar conceptual contexts.

This page contains the following content:

Benefits of concept searching

The following are benefits of concept searching:

  • Language contains many obstacles that prevent keyword search from being effective. These issues revolve around term mismatch—in that a single word can have many meanings (e.g., mean), multiple words can have the same meaning (e.g., “cold” and “frigid”), and some words have specialized meaning within a certain context or group of people. Concept search is more accurate and circumvents these issues because it does not depend upon term matching to find relevant documents.
  • Communications can often be intentionally distorted or obfuscated. However, because concept search can see a document holistically, it can still find conceptual meaning in a document with intentional obfuscation.
  • Concept searching forces the focus on conceptual relevancy rather than on any single term.
  • Concept searching encourages the user to express a concept in the way that people are used to describing ideas and concepts. Concept searching can handle very long queries.
  • Concept searching ultimately functions by finding term correlations within a document and across a document set. Therefore, in the context of the data set, Conceptual Analytics can provide a very accurate term correlation list (dynamic synonym list).

Running a concept search in the viewer

To run a concept search from the viewer, perform the following steps:

  1. Select a document from the document list, and then open it in the Relativity viewer.
  2. Highlight a section of text, and then right-click the text to display a set of menu options.
  3. Hover over Analytics, and then click Concept Search.

The conceptual hits display in the Search Results Related Items pane. The results are sorted by rank. The minimum concept rank used for the right-click concept search is 60. This value is not configurable.

Note: The rank measures the conceptual distance between the query text and the searchable documents in the conceptual index. The higher the rank, the higher the relevance to the query. A rank of 1.00 represents the closest possible distance. The rank doesn't indicate the percentage of shared terms or the percentage of the document that isn't relevant.

Running a concept search from the Documents tab

To run a concept search, perform the following steps:

  1. Navigate to the search panel.
  2. Click Add Condition.
  3. Select (Index Search) from the Add Condition drop-down.
  4. The (Index Search) window opens.

  5. Select an conceptual analytics index.
  6. Perform one or more of the following tasks:
  7. In the Concepts box, enter a paragraph, document, or long phrase for a conceptual search. 

    Note: Enter a block of text, rather than a single word to get better results. Single word entries return broad, unreliable results. A good example source might be a hot document from your workspace, a complaint for the case, or an excerpted paragraph from a relevant web article.

  8. Select any of these optional settings to control how your results are displayed:
    • Select Sort by rank to order the documents in the result set by relevance. The most relevant documents are displayed at the top of the list.
    • Select Min rank to set the ranking for a minimum level of conceptual correlation. The resulting set contains only documents that meet this minimum correlation level.
  9. Click Apply.
  10. Note: To stop a long running search, click Cancel Request.