Overview
Similarity Terms bring the power of AI to your fingertips, enabling you to instantly search and track similar clause language across your LinkSquares Analyze repository.
These values automate contract language audits and track important language upon ingestion of new agreements.
Similarity Terms enable you to instantly search and track similar clause language across your Analyze repository.
Administrators and Managers can define their own custom-extracted values with a single example of clause language. From there, LinkSquares AI finds and tracks similar clauses to support niche language use cases.
These values automate contract language audits and track important language upon ingestion of new agreements. Run reports using Similarity Terms to identify agreements containing or missing your unique language.
To watch the webinar on Similarity Terms, reference the link HERE.
If you would like this feature enabled in your LinkSquares account, contact your Customer Success Manager.
Feature Focus
Process
Analyze Administrators and Managers can create Similarity Terms to track their clause language of choice within their repository. Currently, this can be done by navigating to Analyze App settings and manually entering reference text, or highlighting a string of reference text within the individual agreement view. The LinkSquares AI algorithm handles the rest.
Similarity Terms are based on the syntax of your agreements. Syntax is defined as the arrangement of words and phrases within a sentence. An example of syntactically similar language is, “The contract shall renew on October 31, 2023” and “The contract shall not renew on October 31, 2023.” Within these language examples, the only difference is one word. Otherwise, the wording and word order are very similar. LinkSquares AI would view these results as similar and flag them as Similarity Terms.
An example of semantically similar but syntactically different language is, “This agreement shall terminate on the last day of the first quarter in 2023” and “termination date March 31, 2023.” This language means the same thing but has different wording. LinkSquares AI would view these results as less similar to the previous example.
When leveraging Similarity Terms, LinkSquares AI will compare the similarity of a set of words (i.e., the reference text) to another set of words (i.e., the index text of your agreements) and find the best match. Once a Similarity Term is created, LinkSquares AI will search through your repository and identify instances where words and word order are similar.
LinkSquares AI will allow for some words within these instances to be completely different, assuming that the majority of the original sequence is the same. This means that LinkSquares AI will identify syntactic similarity within your agreements and the reference text while still allowing for some syntactic variance. If the variation is within the threshold, the variant language will be populated.
LinkSquares AI will not perform as well when the majority of words differ between the reference text and the index text within your agreements (e.g., completely different sentences or when using many synonyms). For example, “This agreement shall not be assigned without notice” and “This contract cannot be transferred without written notice” are too dissimilar and would not produce a match.
Limitations
- There is a maximum of 10 Similarity Terms per an organization's Analyze instance.
- There is a reference text maximum of 600 words per Similarity Term.
- Including more than one reference language example per Similarity Term is unsupported.
- Each Similarity Term will return one match on a given agreement.
- Similarity Terms will not be applicable on the Governing Summaries level.
Use Cases
- Create Similarity Terms based on your standard language and run reports to identify where it is present or missing.
- Monitor agreements containing your non-standard language to proactively mitigate risk.
- Monitor language surrounding price increases.
- Monitor radius provisions within property agreements.
- Monitor territory provisions (e.g., where companies are based, where governing law provisions prevail, and the language that follows publicity clauses).
- Monitor limitation of liability clauses.
- Monitor retention of title clauses.
- Monitor FAR clauses in government contracts.
- Monitor fee schedules in insurance contracts.
Additionally, consider the following:
- The pain points surrounding manually inputting data
- The limitations of your standard content searches
- The information not presently captured within your Smart Values
Best Practices
- Provide the longest version of the language you wish to track which is known to you as that provides LinkSquares AI with the greatest surface area to match against.
- When tracking a clause (i.e., one or more sentences), begin with a capitalized word and include a period at the end. This helps LinkSquares AI snap to sentence boundaries. If the clause is longer than 20 words, it is assumed that we want to snap to sentence boundaries irrespective of beginning and end.
- When tracking a clause (i.e., one or more sentences), do not include number headings at the beginning of a clause as those could vary in other documents.
- When tracking short queries (i.e., four words or fewer), content searches from the main Agreements page will provide results more quickly than Similarity Terms. It will also provide multiple results per document. However, this method lacks the flexibility of Similarity Terms (e.g., spelling errors will not prevent a match) as well as the fact that it is a permanently saved search applied to all incoming documents.
Creating Similarity Terms
There are two workflows for creating Similarity Terms:
- Analyze Settings
- Individual Agreement View
Analyze Settings
1. To begin, go to Settings from the app selector.
2. Select Analyze App.
3. Go to the Terms tab.
4. Click CREATE NEW TERM.
5. Select the SIMILARITY TERM tab within the modal.
6. Complete the fields as follows.
Text for Comparison
Enter an example of clause language. This will be the reference text of the language you are looking to track.
Note:
- There is a reference text minimum of three words and a maximum of 600 words.
- The reference text cannot be edited once the term is created. The entire term will need to be deleted and recreated with the edits.
- Each Similarity Term supports one example of reference language.
Name
Enter a unique name for your Similarity Term.
Description
Although optional, a description is recommended to offer context of what the Similarity Term is tracking to other users in your workspace.
This description will be displayed as a tooltip for the term within the individual agreement view.
SFDC Field Name
Enter the field name for how this Similarity Term will appear in Salesforce. This field is optional.
Click CREATE once complete.
After creating the term, there will be a delay before users can view the AI-generated matching text within their repository. LinkSquares estimates processing will take several minutes. This delay will differ depending on the number of agreements within the account.
Similarity Terms will be automatically populated as new agreements are added to your repository. They can be exported, leveraged for reporting, and added as a column on the main Agreements page so you can easily surface your custom language.
Main Agreements Page
Once a Similarity Term has been added as a column, use the sorting icons next to the column to organize your agreement view.
The sorting icons sort all agreements by how closely their Similarity Term matches the reference text.
- Ascending order: Lowest to highest similarity with the Not Present values at the end of the results.
- Descending order: Highest to lowest similarity score with the Not Present values at the end of the results.
Note: The sorting order will apply to any exports of the main Agreements page.
Individual Agreement View
While reviewing an agreement within the individual agreement view, highlight a string of text to track the language with Similarity Terms and initiate the term creation workflow.
Note: There is a reference text minimum of three words and a maximum of 600 words.
Once the language has been highlighted, select + Create term.
Select the SIMILARITY TERM tab within the modal.
The Text for Comparison field will be pre-populated with the highlighted language.
To learn more about completing the fields within the modal and creating new Similarity Terms, reference the Analyze Settings section of this article.
Modification
Editing Similarity Terms
To edit an existing Similarity Term, click the ellipsis icon next to the relevant Similarity Term within the Terms tab of Analyze App settings. Leverage the Similarity Terms filter within the All Terms drop-down and the search bar if needed.
Click EDIT from the drop-down to modify the term.
From here, users can edit the name of the Similarity Term, the description, and the Salesforce field name.
The reference text can be found within the Text for Comparison section of the modal. Since the reference text is being used to surface the value, it cannot be modified after creation.
To modify the reference text, delete the Similarity Term and create a new one with the appropriate reference text.
Deleting Similarity Terms
To delete an existing Similarity Term, click the ellipsis icon next to the relevant Similarity Term within the Terms tab of Analyze App settings. Leverage the Similarity Terms filter within the All Terms drop-down and the search bar if needed.
Click DELETE from the drop-down to delete the term.
Once actioned, there will be a delay for Analyze to fully delete the term from the application. All data within the deleted Similarity Term will be lost.
Saved reports that include a filter on the deleted Similarity Term will provide unexpected results since the filter will not be tied to an active Analyze term.
Reporting
Similarity Terms can be leveraged as data points within the advanced filtering workflow. Run reports using Similarity Terms to identify agreements containing or missing your unique language.
To report on Similarity Terms, start at the main Agreements page. Click + Advanced Filters.
From here, select Similarity Term from the drop-down. Next, select the Similarity Term. Lastly, determine whether the Similarity Term is present, not present, etc.
The page will automatically populate with the results based on your search criteria.
Similarity Terms Values
Reference the Similarity Terms within the individual agreement view by locating the Similarity Terms section at the bottom of the Global Terms panel. From here, identify the language that LinkSquares AI surfaced.
Click the highlighter icon next to a Similarity Term to view where similar language can be found within the agreement.
Similarity Terms can also be added as data points within column settings. Click the Columns button to customize the main Agreements page for easy reference.
The value of the Similarity Term is dependent on the search results. The agreement text similar to the term’s reference text will be the value returned. This means that there will be variations across agreements.
- Present: If language similar to the reference text is detected, the closest matching text will populate within the Similarity Term field. It will be highlighted in the agreement text.
- Processing: This will be displayed until a result is or is not found.
- Not Present: If language similar enough to the threshold is not found, Not Present will be populated as the value for the term.