

Note: Relativity partitions a single index into multiple smaller indexes (called sub-indexes), which multiple workers build simultaneously. This increases performance by spreading out the work over a configurable number of agents. When you perform a search, Relativity runs your query on the smaller indexes in parallel. The application then federates and returns your results. For more details, see the Ask The Expert Training content: Searching: Best Practices for dtSearch Builds.
See these related pages:
Also see these related recipes:
There are roughly three million files relevant to a case you are working on, including emails, email attachments, invoices, and technical manuals related to construction practices and material handling. It is early in the case, and you need to gain an understanding of the data set. You also need to retrieve certain text related to five substances that you know are prevalent in this data, as an employee from the construction company mentioned these specifically in an email to you. To do this, you need to be able to perform proximity, stemming, and fuzzy searches on your data set. So you create a new dtSearch index.
You call the index Hazardous Materials dtSearch so that you can identify it in the Search drop-down menu on the Documents list. You might also create an Analytics index for this case with a similar name, so make sure to differentiate them clearly. For the Searchable set field, you select a saved search that you have already created called Hazardous Materials searchable set, which has documents to which you have already applied keywords related to the substances mentioned in your client's email.
Because many of the invoices and emails in your data set contain references to various purchases of building materials made by various departments in the construction company you are helping to represent, you set the Auto-recognize date, email, and credit card numbers field to Yes.
You leave all other fields at their default settings and save the index. You then build and activate the index so that you can select it in the Search drop-down menu.
When you select the index and search your document set on it, you run proximity searches to see how close terms relating to hazardous substances occur to the names of the building materials that may or may not contain them. The searches you run include the following:
As you keep running these proximity searches, you get down to a small group of intriguing emails between a prospective buyer, your client, and a prospective seller, which may prove that the seller had knowledge of the fact that those building materials were potentially dangerous when they were negotiating a price with your client. This discovery turns out to be crucial to the case.
You can build custom dtSearch indexes for a subset of documents or for certain document fields in a workspace. You must have the appropriate permissions to complete this task.
Before you begin, you need to create a saved search that includes the fields that you want to include in the index. You can then name the index based on the document search set used to create it.
Note: Within a field, dtSearch truncates any string longer than 32 characters that does not contain a space character. It indexes only the first 32 characters of the string. For more information, see Searching for words longer than 32 characters.
To create a new dtSearch index:
Click OK to build your index.
Note: Network problems can slow down your dtSearch builds. If a dtSearch manager or worker agent encounters a network-related error during the build process, it executes up to three retry attempts at 30-second intervals.
Once the index builds, the console enables additional options. See dtSearch console.
By default, Relativity builds an accent-insensitive index. In an accent-sensitive index, some characters translate to the base character, which causes those characters and any terms containing those characters to be treated the same in a Search Terms Report.
Note: dtSearch uses .ABC files, but only for characters in the range from 33 to 127. Relativity handles all other characters according to the definitions in the Unicode character tables.
Example: Relativity converts accented characters like á or ñ to the unaccented versions, a or n.
Example: If you search for the term fröhlich, searching that term as fröhlich or frohlich would both return the hit. However, highlighting in the Viewer may not display both variations.
The dtSearch index page includes the following fields:
dtSearch Index Information
Note: When creating a saved search for a dtSearch index, the best practice is to use long text fields. You will see a warning message, when creating a dtSearch index, if you select a saved search that does not contain at least one long text field.
Advanced Settings
Noise Words
Alphabet
Note: If you search for long, uninterrupted strings that have no spaces or word breaks, such as when you have made a character searchable, dtSearch truncates the string after 32 characters and inserts a wildcard. For more information, see Searching for words longer than 32 characters.
The dtSearch index console includes the following options:
Note: Canceling the build stops the indexing thread, leaving the index in an unstable state. Relativity deletes these indexes from the population table and gives them an inactive status. You cannot search on an index with an inactive status until you run a full build. Canceling also deletes the index files in the index share.
Note: The incremental build process copies each sub-index that requires modification, updates the copy, then replaces existing sub-indexes with the updated copies. When run, the
Note: The Compress Index button only runs compression on sub-indexes that have a fragmentation level greater than zero. Canceling compression returns the index to its original fragmented state before compression began.
Selecting the index from the drop-down list and clicking OK completes the index swap. You cannot reverse the swap results in the current dialog box. You must close this swap and run it again to swap back or swap another time. This functionality is useful in limited cases, for example, if you are performing a full rebuild on a very large index. Since dtSearch incremental builds are online, you can search documents once indexed.
Note: The Swap Index function updates anything in the Views table, which affects batches, saved searches, nested searches, and more.
After you create and build a dtSearch index, the dtSearch page has several sections where you can view details about your index.
You can view the state of your dtSearch index from the Index Status section. The name of the Index Status section populates with the name of your dtSearch index. When you are building an index, this section changes to a progress bar where you can track your index's progress in real-time. When the index is no longer in progress, this section changes to a static field that displays the below fields.
The dtSearch Index Information section provides general details about the settings applied to your dtSearch index. This section has the following information:
The Advanced Settings section provides sub-index details about your dtSearch index. This section has the following information:
The Temporary Index Details section only appears during an incremental build. This table displays sub-indexes that copy from your original index and are in the process of modification during the incremental build. Once the sub-indexes in this table update, they replace the original sub-indexes from which they were copied. This section has the following information:
The Current Index Details section displays the sub-indexes that make up your dtSearch index. This section has the following information:
Using the View Audit button, you can see when a user modified dtSearch index settings. The View Audit layout has the following fields:
If you specify a temporary storage location, dtSearch builds the index in this directory and then copies the index over to the final index share when the build completes. Using a temporary storage location could speed up the build time and reduce network contention
Why was this not helpful?
Check one that applies.
Thank you for your feedback.
Want to tell us more?
Great!